Drug-induced thrombocytopenia: an evidence-based review

Introduction

A low platelet count in a patient taking drugs raises one question before any other: is a drug the cause? The answer changes management at once, because the treatment of immune drug-induced thrombocytopenia is to stop the drug, and the patient must then avoid it for life.^4,5 Getting the answer wrong has costs in both directions. Missing it leaves the patient exposed to a drug that may cause severe bleeding. Blaming the wrong drug removes a useful treatment and may lead to unnecessary treatment for immune thrombocytopenia; in one cohort, 3 of 28 patients later found to have drug-induced thrombocytopenia had had a splenectomy before the drug was recognised.⁵

Three features make the question hard. The reaction is rare for any single drug. The laboratory test that would settle it is available in only a few reference laboratories and misses many true cases.^3,6 And most of the evidence for individual drugs comes from case reports, judged by explicit causality criteria, and from laboratory series of patients whose serum was referred for testing. Randomised evidence exists only where thrombocytopenia is common enough to count in a trial: the glycoprotein IIb/IIIa inhibitors, the heparins and some cancer drugs. The level of evidence is stated with each claim.

The first part of this review covers what is common to all drugs: frequency, causality, mechanisms, drug-induced thrombotic microangiopathy, recognition, laboratory confirmation and management. Heparin-induced thrombocytopenia follows as its own section. The implicated drugs are then arranged by system, each with a table.

Epidemiology and how causality is judged

How often it happens

Population estimates are few and old. A review of the epidemiological studies, which differ in method, put the incidence of immune drug-induced thrombocytopenia at about 10 cases per million people a year,¹ and a later expert review gives the same figure and suggests it may be higher in older and hospitalised patients.¹³ The first case–control study to give relative risks for individual drugs was published in 1993. It compared 62 patients with acute thrombocytopenic purpura against 2625 hospital controls. The overall incidence of acute thrombocytopenic purpura from any cause was 18 per million a year. For drugs used in the week before onset, the relative risk was 124 for trimethoprim–sulfamethoxazole, 101 for quinidine or quinine, 14 for dipyridamole, 4·8 for sulfonylureas and 2·6 for salicylates. In absolute terms the excess risk was 38 cases per million users a week for trimethoprim–sulfamethoxazole and 26 for quinidine or quinine.¹⁷

Later surveillance confirms that drugs are a common cause of acute immune thrombocytopenia in adults. In the Berlin Case-Control Surveillance Study (2000–2009), 90 of 169 validated cases of acute immune thrombocytopenia were at least possibly related to a drug. Eighty-five drugs were involved, 30 of them with certain or probable causality. Tirofiban (10 cases), abciximab (4), trimethoprim–sulfamethoxazole (4), influenza vaccine (3) and citalopram (2) were each implicated in at least two probable or certain cases.¹⁸ An earlier analysis from the same programme judged 20 cases (25%) of immune thrombocytopenia at least possibly drug-related.¹⁹ A Dutch population case–control study of 705 cases found a raised risk with current beta-lactam use (adjusted OR 7·4, 95% CI 1·8–29·6), but no significant increase for anticonvulsants, cinchona alkaloids, diuretics, NSAIDs or antituberculous drugs, for which it had limited power.²⁰

Children are affected too. A systematic review found 32 substances with definite or probable evidence in children, and in 2008–2012, 21 of 91 children tested had drug-dependent platelet-reactive antibodies, involving 6 substances.²¹ Thrombocytopenia in intensive care is common and has many causes; the share that is drug-induced is discussed under anaesthesia and intensive care.

How many drugs

Lists of implicated drugs are long, and their length reflects reporting more than risk. The Oklahoma systematic review of 1998 assessed 515 case reports from 561 articles describing 774 patients. Evidence for a causal role was definite or probable in 247 (48%), involving 98 drugs. Quinidine was named in 38 of these reports, gold in 11 and trimethoprim–sulfamethoxazole in 10. Heparin, cytotoxic drugs and children were excluded.²² By 2009 the database held 1185 individual patient reports on 317 drugs, of which 451 (35%) were definite or probable, and 21 drugs had five or more definite or probable reports. Both analyses state that counts of published reports cannot estimate how often a drug causes the reaction.^5,22

Two studies narrowed the list by adding independent evidence. Reese and colleagues (2010) combined three methods. Published case reports had implicated 253 drugs, of which 87 (34%) met the clinical criteria. The BloodCenter of Wisconsin had tested 202 drugs and found drug-dependent platelet-reactive antibodies for 67 (33%). Mining the US Food and Drug Administration adverse event database flagged 573 of 1444 drugs (40%). Twenty-four drugs had evidence from both case reports and antibody testing, and 23 of these were also flagged by data mining; amiodarone was the exception. The authors suggest that these 23 drugs should be considered a likely cause of otherwise unexplained thrombocytopenia in any patient taking them.²³ Arnold and colleagues (2013) then regraded the published laboratory evidence. Of 153 drugs clinically implicated, 72 (47%) had a positive laboratory test, and only 16 met the criteria for a definite laboratory diagnosis (Table 1).³ The 47% is a proportion of drugs, not of patients. No study found in this search gives the proportion of patients with suspected drug-induced thrombocytopenia in whom an antibody is eventually found.

Judging causality in a single patient

Most statements that a drug "causes" thrombocytopenia rest on case reports graded by the criteria that George and colleagues published in 1998 (Table 2).^5,22 The criteria are simple and can be applied at the bedside. Their weakness is the fourth criterion, re-exposure, which is rarely done and is often unsafe.

Mechanisms

Six immune mechanisms are recognised, alongside non-immune suppression of the marrow (Table 3).^4,13 The distinction matters at the bedside for three reasons. It predicts timing: most drug-dependent antibodies need a week or more to form, but the antibodies against glycoprotein IIb/IIIa inhibitors are often already present. It predicts recovery: a drug-dependent antibody stops working when the drug is cleared, but a drug-induced autoantibody may not. And it predicts the response to dose: myelosuppression can be managed by lowering the dose, whereas an immune reaction cannot.

Several laboratory findings change how the classic mechanism should be understood. Drug-dependent antibodies may come from a natural pool of weakly self-reactive antibodies; the drug is thought to supply the missing contact that lets them bind at ordinary drug concentrations. The authors state that this model has not been directly validated.^5,13 Patients with quinine-induced thrombocytopenia carry two antibodies: one that binds platelets only with the drug present, and a second that recognises the drug alone.²⁹ The antibodies that follow tirofiban or eptifibatide recognise drug-specific shapes of GPIIb/IIIa clustered around the RGD site, which differ from the classic ligand-induced binding sites.²⁵ For paracetamol and naproxen, the antibody may recognise a metabolite rather than the parent drug, so a test with the parent drug can be negative.³⁰

Drug-induced thrombotic microangiopathy

A few drugs cause a thrombotic microangiopathy rather than isolated thrombocytopenia. The patient has a low platelet count, microangiopathic haemolytic anaemia and organ injury. The picture is easily taken for thrombotic thrombocytopenic purpura (TTP) or haemolytic uraemic syndrome (HUS). In a registry series of 19 patients with quinine-induced thrombotic microangiopathy, all had acute kidney injury, and every one was first diagnosed as TTP or HUS.³¹

The evidence is a systematic review of published reports and a population-based registry combined with one reference laboratory. The systematic review assessed 344 articles with individual patient data on 586 patients and described 78 drugs. Only 22 drugs (28%) had definite evidence of causing thrombotic microangiopathy, and for 11 of those 22 the evidence was a single patient. Definite evidence was present in 104 of the 586 patients (18%). Three drugs accounted for 61 of the 104 definite reports: quinine (34), ciclosporin (15) and tacrolimus (12). Another 20 drugs had probable evidence.³² The authors warn that earlier reports can cause diagnostic suspicion bias. After the first report of mitomycin-associated thrombotic microangiopathy in 1971, 61 further patients were reported, but the association was only possible in 54, mostly because 5-fluorouracil was given at the same time; no report blamed 5-fluorouracil. Of 85 patients reported with gemcitabine, the association was only possible in 73.³² In the systematic review, ADAMTS13 deficiency counted as evidence for another cause, and drug-dependent inhibition of ADAMTS13 has not been reported.³²

Immune and toxic forms

The Oklahoma group separates two mechanisms by their clinical pattern (Table 4).^15,32 The distinction matters most for the immune form: the authors stress that failing to recognise the drug may have critical consequences if the patient is given it again.³² For almost all drugs the mechanism is unknown. Thrombotic microangiopathy caused by inhibitors of vascular endothelial growth factor involves injury to renal podocytes, and quinine may cause immune injury to endothelial cells.³²

Quinine

When causality criteria are applied strictly, one drug dominates. The Oklahoma TTP–HUS Registry enrols every patient in a defined region referred for plasma exchange. Of 58 registry patients once thought to have a drug-induced cause, involving 15 suspected drugs, 21 had definite evidence, and these involved only three drugs: gemcitabine, pentostatin and quinine. Nineteen of the 21 (90%) had taken quinine. Of 40 patients whose serum was tested at the BloodCenter of Wisconsin for 8 drugs, 30 had drug-dependent antibodies, involving oxaliplatin, quinine and vancomycin; 28 of the 30 (93%) had taken quinine. Across both sources, 51 patients had definite evidence, and quinine was the cause in 47 (92%).¹⁵

Quinine-induced thrombotic microangiopathy is severe. In an earlier registry analysis, 17 of 225 consecutive patients with TTP–HUS had taken quinine; 4 died and 7 survivors had chronic kidney failure. Among the 132 patients who were asked directly about drugs, 14 (11%) had taken quinine.³³ In the 2017 series of 19 patients, all had microangiopathic haemolytic anaemia, thrombocytopenia and acute kidney injury. Eighteen were women. The exposure was quinine tablets in 18 and tonic water in 1. Seventeen of the 18 survivors needed dialysis, 14 developed chronic kidney disease, 3 reached end-stage kidney disease, and 8 patients died.³¹ A systematic review of 142 patients with acute immune reactions to quinine found that 72% had taken it as tablets and 20% in drinks. Apart from purely skin reactions, 91% needed hospital care, 30 needed renal replacement therapy and 3 died.³⁴ Quinine is also present in tonic water and bitter lemon.⁵ Patients may not think that drinks, foods, herbal remedies or medicines they choose for themselves are relevant, and may not mention them.⁵ Asking about quinine, in tablets and in drinks, belongs in the assessment of every patient thought to have TTP or HUS.¹⁵

The thienopyridines are a separate case and are discussed with the cardiovascular drugs. Most patients with ticlopidine-associated TTP had severe ADAMTS13 deficiency with antibodies; patients with clopidogrel-associated TTP did not.^35,36

Clinical recognition and timing

The typical patient with immune drug-induced thrombocytopenia has a sudden, severe fall in the platelet count, with petechiae, bruising or nosebleeds.³⁷ The count is usually below 20 × 10⁹/L,^4,5 and with quinine and quinidine it is classically below 10 × 10⁹/L.⁴ In one comparison the median nadir was 10 × 10⁹/L in drug-induced immune thrombocytopenia and 60 × 10⁹/L in heparin-induced thrombocytopenia.⁴ Bleeding is common but seldom fatal. Among 247 published patients with definite or probable evidence, 23 (9%) had major bleeding and 2 (0·8%) died of bleeding.^2,22 Series for single drugs can be worse. In 29 patients with vancomycin-dependent antibodies, the mean nadir was 13·6 × 10⁹/L and 10 (34%) had severe bleeding.³⁸

A patient who already has the antibody may feel unwell within hours of a dose, with faintness, chills, fever and nausea; hypotension and syncope have occurred with high-titre antibodies. A few patients also have immune haemolytic anaemia or neutropenia.¹³

Timing

Timing is the most useful clue (Table 5). On first exposure to a drug taken daily, the count falls 5–10 days after the start.^2,4 George and Aster give the same window as 1–2 weeks,⁵ and Aster and colleagues note that the patient has, as a rule, taken the drug for at least a week.¹³ These are the same statement at different precision. The fall is sudden when a patient who has taken a drug on and off takes it again, sometimes within hours of a single dose.^4,5 The glycoprotein IIb/IIIa inhibitors are the exception to the first-exposure rule: they can cause severe thrombocytopenia within hours of the first dose, because the antibodies are often naturally occurring.^2,13

Re-exposure to abciximab has been measured directly. In the ReoPro readministration registry of 1342 repeat treatments, the count fell below 100 × 10⁹/L in 5% and below 20 × 10⁹/L in 2%. When the repeat dose came within a month of the first (115 treatments), the rates were 16·5% and 12·2%. A positive human anti-chimeric antibody test before the repeat dose was linked to more thrombocytopenia, 14·1% against 4·4%.³⁹

Recovery

Once the drug is stopped, bleeding usually settles in 1–2 days and the count starts to rise within 1–2 days.^5,13 Recovery is usually complete within a week,⁵ and the count is often normal in 4–8 days, though it occasionally stays low for several weeks.¹³ Bakchoul and Marini put it in terms of clearance: recovery starts after 4–5 half-lives of the drug or its metabolite.² The rule therefore assumes that the drug is cleared normally. A 65-year-old woman with liver and kidney failure had a fall from about 70 × 10⁹/L to 3 × 10⁹/L two days after a 5-day course of ceftriaxone. Her count stayed at 1–6 × 10⁹/L until she died 13 days later, despite intravenous immunoglobulin, steroids and platelet transfusion. Ceftriaxone was still detectable in her serum at least 8 days after the last dose.⁴⁰ A drug-induced autoantibody is the other reason for slow recovery, because it acts without the drug and the thrombocytopenia may continue after withdrawal.^4,13 Slow recovery should prompt a check of drug clearance and of the other causes, not only a longer wait.

Hidden exposures and the mistaken diagnosis of ITP

A drug cause is often missed when the exposure is not seen as a drug. In a cohort of 343 patients registered with immune thrombocytopenia (ITP), 28 (8%) were later found to have drug-induced thrombocytopenia. Quinine, including quinine in tonic water, was the cause in 13 of the 28.⁵ In another report, 5 patients with persistent thrombocytopenia were treated as relapsing ITP, with several treatments including splenectomy, before their use of quinine tablets for leg cramps came to light.⁴ Drug-dependent antibodies can last for many years, and quinine has caused acute thrombocytopenia in patients who had not taken it for more than 10 years.⁵ The history therefore needs to cover drugs taken in the past as well as now, drugs taken only now and then, and drinks, foods and remedies the patient does not count as medicines.

Diagnosis and laboratory confirmation

The diagnosis is clinical in most patients. Arnold and colleagues give a practical set of features: a nadir below 20 × 10⁹/L, bleeding, onset 5–10 days after first exposure or within hours of a later exposure or of a first dose of a glycoprotein IIb/IIIa inhibitor, and a drug that has been implicated before.⁴ The George criteria (Table 2) then grade the case: the drug came first and the count recovered fully after it was stopped; other drugs were continued or restarted without a fall; other causes were excluded; and, rarely, re-exposure reproduced the fall.^5,22 The diagnosis is confirmed by a positive rechallenge, which is often impractical, or by a positive test for drug-dependent antibodies.⁴ George and Aster add one practical test of the diagnosis. If steroids were started because ITP could not be excluded, they can be stopped once the count is normal. If thrombocytopenia returns and the patient has certainly not taken the drug again, drug-induced thrombocytopenia is excluded.⁵

The differential in hospitalised and critically ill patients

The diagnosis is hardest in hospital. Most inpatients take several drugs and have other illnesses that lower the count.^2,4,41 Thrombocytopenia is common in intensive care: across 24 studies of 6894 patients, it was present on admission in 8·3–67·6% and developed during the stay in 13·0–44·1%.⁴² The shape of the platelet curve is a useful guide to the cause. A moderate fall in the first 3–4 days after major surgery is expected. A slow fall over 5–7 days suggests consumption or failure of production. An abrupt fall, within 1–2 days, after the count had begun to recover, at about 1–2 weeks after surgery, strongly suggests an immune cause: heparin-induced thrombocytopenia, another drug-induced thrombocytopenia, or post-transfusion purpura.⁴³ A fall of more than 50% within 1–2 days, especially after an earlier rise, carries the same meaning.⁴⁴ When the patient has received heparin, heparin-induced thrombocytopenia is considered as described in its own section.

Before any of this, the low count must be shown to be real. Pseudothrombocytopenia is an effect in the sample tube, and it has to be separated from true drug-induced thrombocytopenia.¹ The point is not academic. In four trials of abciximab in coronary intervention, pseudothrombocytopenia occurred in 2·1% of patients given abciximab and 0·6% of those given placebo, and it made up 36·3% of the low counts on abciximab. Unlike true thrombocytopenia, it was not linked to bleeding, transfusion, stroke or repeat revascularisation.¹⁴ A look at the blood film for platelet clumps comes before stopping any drug.

Laboratory confirmation

A test for drug-dependent antibodies must show three things: the drug, or its metabolite, is needed for the reaction in the test tube; immunoglobulin binds; and the target is platelets. Ideally the result is also reproduced in a second laboratory.^3,6 Several methods meet these criteria. Flow cytometry is the most used, because it is sensitive and gives a quantitative result.⁴ The tests are available in only a few reference laboratories, and the methods have not been standardised.^3,4 In most hospitals, therefore, the result will not be back in time to guide the first decision, which is whether to stop the drug. That decision rests on clinical judgement.¹³

In 2015 the Platelet Immunology Subcommittee of the International Society on Thrombosis and Haemostasis (ISTH) published recommendations to standardise testing. They are based on the experience of three reference laboratories, in the United States, Canada and Germany, and they cover the four prescription drugs most often implicated: quinine, vancomycin, trimethoprim–sulfamethoxazole and piperacillin–tazobactam (Table 6). The authors state that the recommendations still need prospective evaluation.⁶

Why a negative test does not exclude the diagnosis

The test is specific but insensitive for some drugs, and the ISTH document gives three reasons for a false-negative result.⁶ The first is that a metabolite, not the parent drug, caused the reaction, while the laboratory tests the parent drug. In 5 patients with thrombocytopenia after naproxen or acetaminophen (paracetamol), no antibody was found with the parent drug, but antibodies were found in the presence of naproxen glucuronide or acetaminophen sulfate.³⁰ The second is poor solubility. Sulfamethoxazole and trimethoprim are only slightly soluble in water at neutral pH, which limits the concentration that can be tested.⁶ The third is a sample taken too late, after the count recovered and the antibody fell or disappeared.⁶

The sources differ on how long the antibody lasts. The ISTH document allows testing up to 3 weeks after the event.⁶ Aster and colleagues say the antibodies remain for at least a month and sometimes indefinitely,¹³ and George and Aster say they can persist for many years.⁵ A sample from the acute episode avoids the question.

Case reports show what this means in practice. A 76-year-old man whose count fell from 187 to 5 × 10⁹/L over 4 days after a single 750 mg dose of levofloxacin, six months after an earlier course, had no detectable levofloxacin-dependent antibody. He recovered after the drug was stopped and he was given intravenous immunoglobulin and dexamethasone. The authors put the negative test down to the low sensitivity of the assays and a probable metabolite-dependent antibody.⁴⁵

Rechallenge

A positive rechallenge confirms the diagnosis, but it is often impractical and can be dangerous.⁴ The ISTH document suggests that a carefully supervised rechallenge may be considered when the test is negative and the diagnosis is still strongly suspected, particularly when the drug is very common, such as acetaminophen, or when there is no alternative.⁶ Aster and colleagues reserve it for exceptional cases in which the drug is essential, once the count is normal again. They advise starting with very small doses, 1–2 mg, with close monitoring of the count, because a full dose can cause life-threatening bleeding.¹³ The evidence for this advice is expert opinion.

Management

No randomised trial has tested any treatment for immune drug-induced thrombocytopenia. What follows rests on expert reviews and case series (Table 7).

The treatment is to stop the drug. The decision cannot wait for a laboratory result, because the test takes time and few centres offer it.¹³ When the patient takes many drugs, Bakchoul and Marini advise stopping every drug started in the last 2 weeks, antibiotics above all, and replacing those that are needed.² Arnold and colleagues describe the same approach in a worked case: the likely drugs were stopped or replaced, and a drug judged very unlikely to be the cause was continued.⁴ A pharmacologically equivalent drug with a different chemical structure can usually be substituted safely.¹³ The count usually starts to recover within 1–2 days.^4,5 If it does not, the causes of slow recovery described above apply: impaired clearance of the drug, a drug-induced autoantibody, or a different diagnosis.

Doses are as stated in the cited sources; check local protocols and product labelling.

Transfusion, immunoglobulin and steroids

Platelet transfusion is reasonable in a patient with a very low count and active bleeding, because of the risk of intracranial haemorrhage, but its effect has not been formally studied.¹³ Bakchoul and Marini add that transfusion is generally ineffective while the drug or its metabolite is still in the plasma.² High-dose intravenous immunoglobulin can be given for severe thrombocytopenia with bleeding or a high risk of bleeding, but the recommendation is based only on case reports.² Corticosteroids are commonly given because ITP cannot be excluded at first; whether they help in drug-induced thrombocytopenia is not established.^5,13 A review written for intensive care reaches the same view of steroids, immunoglobulin and plasma exchange: their role is uncertain.⁴¹ The same review notes that the prognosis is excellent once the drug is stopped,⁴¹ which makes the added effect of any other treatment hard to measure.

Drug-induced thrombotic microangiopathy

The drug is stopped here too. Whether plasma exchange adds anything has not been tested in a controlled comparison. The Oklahoma Registry enrols patients who were referred for plasma exchange, so it has no untreated comparison group.¹⁵ All 19 patients with quinine-induced thrombotic microangiopathy received it, and the outcomes described above, with dialysis in most survivors and 8 deaths, were recorded in treated patients.³¹ The best comparative data are for the thienopyridines and are observational. In 60 cases of ticlopidine-associated TTP, mortality was 50% without plasmapheresis and 24% with it.⁴⁶ A later analysis found that the apparent benefit depended on timing. When TTP began more than 2 weeks after starting a thienopyridine, survival was 84% with plasma exchange and 38% without. When it began within 2 weeks, survival was 77% with and 78% without.³⁶ Ticlopidine-associated cases were more likely than clopidogrel-associated cases to begin after more than 2 weeks (90% against 26%), and patients with severe ADAMTS13 deficiency had mostly received ticlopidine.³⁶

After recovery

Drug-dependent antibodies can persist for many years, so a patient with a confirmed diagnosis must avoid the drug for life.^4,5 A positive laboratory test is a firm reason for avoidance.⁶ When the diagnosis cannot be confirmed, the advice on avoidance is less certain.⁴ The reaction should be documented in the medical record and reported according to institutional and national standards.⁴¹ George and Aster, writing in the United States, ask that confirmed cases be reported to the Food and Drug Administration's adverse event system, and that a case report be published when a drug has few or no published reports with definite or probable evidence. They give the reason: these records are the basis for judging the next patient.⁵

Heparin-induced thrombocytopenia

Heparin-induced thrombocytopenia (HIT) is a prothrombotic adverse drug reaction. In most cases it is caused by IgG antibodies against complexes of platelet factor 4 (PF4) and heparin.⁸ The antibody activates platelets, so the danger is thrombosis, not bleeding. It differs from the other drug-induced thrombocytopenias in three practical ways. The fall in count is usually moderate. Stopping the drug is not enough, because the risk of thrombosis continues and needs a non-heparin anticoagulant. And the diagnosis rests on a validated clinical score and two classes of laboratory test. The management below follows the American Society of Hematology (ASH) 2018 guideline,⁸ which remains the current general guideline; no later ASH guideline on HIT was found in the search for this review. Newer guidance from the ISTH Scientific and Standardization Committee (2026) covers only cardiac surgery with bypass.⁴⁷

Incidence by heparin type and population

About 12 million inpatients in the USA are exposed to heparin each year. The incidence of HIT ranges from under 0·1% to 7%, depending on the type of heparin, the duration of treatment and the patient population.⁸ Three factors matter most.

Type of heparin. The risk with unfractionated heparin (UFH) is about 10 times that with low-molecular-weight heparin (LMWH), and the risk with fondaparinux is negligible.⁸ A meta-analysis of 15 studies (7287 patients) found an absolute risk of 2·6% with UFH and 0·2% with LMWH. In the randomised trials that measured HIT, the odds ratio for LMWH against UFH was 0·10 (95% CI 0·01–0·2), and in the prospective studies it was also 0·10 (0·03–0·33). Most of the included studies were of patients after orthopaedic surgery, so the figures apply best to that setting.⁴⁸ The key randomised trial enrolled 665 patients after hip surgery: HIT occurred in 9 of 332 on UFH and in none of 333 on LMWH (2·7% vs 0%).⁴⁹ In critically ill patients, the PROTECT trial (3746 patients) compared dalteparin with UFH. Fewer HIT events occurred with dalteparin in the pre-specified per-protocol analysis (hazard ratio 0·27, 95% CI 0·08–0·98).⁵⁰

Duration. HIT occurs in 2–3% of patients who receive UFH for 6 or more consecutive days, and is much less common with shorter courses.⁸

Population. HIT is about three times more frequent after surgery or major trauma than in medical patients. It is rare in children and in obstetric patients.⁸ Forming the antibody is not the same as having the disease. In a study of 744 patients, platelet-activating HIT antibodies formed in 3·2% of orthopaedic patients on LMWH and in 20% of cardiac surgical patients on UFH. Yet among UFH-treated patients who formed the antibody, clinical HIT followed far more often after orthopaedic surgery than after cardiac surgery (52·6% vs 5% by the activation assay).⁵¹ A positive antibody test after cardiac surgery is therefore common and, by itself, weak evidence of HIT.

Timing patterns

HIT follows four timing patterns. Only the first matches the textbook course of a fall between days 5 and 10.

Typical onset. In 243 patients with serologically confirmed HIT, the count began to fall 4 or more days after heparin was started in 170 (70%). Previous heparin treatment did not change the timing in this group.⁵²

Rapid onset. The other 73 (30%) had a fall within a median of 10·5 hours of starting heparin. All of them had received heparin in the previous 100 days.⁵² The antibody was already present, so a new dose acted at once. In the same study the antibodies became undetectable at a median of 50–85 days, depending on the assay.⁵² The ASH guideline gives the same figures: functional assays turn negative at a median of 50 days and immunoassays at a median of 85 days. The platelet count recovers within 7 days of stopping heparin in 90% of cases.⁸

Delayed onset. Twelve patients presented with thrombocytopenia and thrombosis a mean of 9·2 days (range 5–19) after heparin had been stopped. Nine who were given heparin again had a further fall in count. They had higher anti-PF4/heparin IgG titres than controls, and their antibodies activated platelets more strongly without heparin. The authors advise suspecting HIT for up to 3 weeks after heparin exposure.⁵³ A patient readmitted with thrombosis a week or two after discharge should be asked about heparin, and should not be given heparin until HIT has been considered.

Autoimmune and spontaneous HIT. Some anti-PF4 antibodies activate platelets without heparin, though they are still inhibited by high heparin concentrations. This is called autoimmune HIT. It underlies several unusual pictures: delayed-onset HIT; HIT that persists for weeks; HIT after fondaparinux; HIT after a heparin flush; and severe HIT with a count below 20 × 10⁹/L and disseminated intravascular coagulation. In these patients fibrinogen, fibrin and D-dimer may track the disease better than the platelet count. Argatroban and bivalirudin are adjusted by the APTT and often fail here, because the coagulopathy itself prolongs the APTT. The review suggests danaparoid, fondaparinux or a direct oral anticoagulant (DOAC), and notes that high-dose intravenous immunoglobulin (IVIG) can interrupt the antibody's activation of platelets.⁵⁴ A spontaneous HIT syndrome, with no preceding heparin, was first reported in 2008. It has a surgical form, especially after total knee replacement, and a medical form, usually after an infection. The suggested treatment is a non-heparin anticoagulant, with oral factor Xa inhibitors favoured over direct thrombin inhibitors, plus high-dose immunoglobulin. Other polyanions can form the same antigen with PF4: pentosan polysulfate, hypersulfated chondroitin sulfate, fondaparinux, and natural polyanions such as bacterial lipopolysaccharide and nucleic acids.⁵⁵ These recommendations come from expert reviews of case series, not from trials.

A 2023 classification sorts anti-PF4 antibodies into three types. Type 1 antibodies do not activate platelets and cause no disease. Type 2 activate platelets only in the presence of heparin; this is classic HIT. Type 3 activate platelets without heparin; this group includes autoimmune HIT and vaccine-induced immune thrombotic thrombocytopenia, which is discussed with the vaccines. Type 3 disease needs anticoagulation plus high-dose IVIG.⁵⁶

What HIT is not. A mild, early fall in count is common on heparin and is not immune. It occurs in 10–30% of heparin-treated patients, usually before day 5. The count often stays above 80–100 × 10⁹/L and recovers even if heparin continues. There is no bleeding or thrombosis, and no treatment is needed.²

Thrombosis and outcomes

One-third to one-half of patients with HIT develop thrombosis, which may be venous, arterial or microvascular. Amputation occurs in about 1–3% and death in about 5–10%. Untreated, the initial risk of thromboembolism, amputation and death is 5–10% a day. Clinically significant bleeding at presentation is uncommon despite the low count.⁸

The best cohort data come from a 14-year study of 127 patients with serologically confirmed HIT. New venous thromboses outnumbered arterial ones by about 4 to 1 (78 against 18 patients), and one in four patients had a pulmonary embolism. About half were recognised only after a thrombosis had occurred. Of the 62 patients recognised with isolated thrombocytopenia, 52·8% developed thrombosis within 30 days. The risk was the same whether heparin was simply stopped or replaced by warfarin.⁵⁷ In this cohort, stopping heparin alone did not prevent thrombosis.

Warfarin carries its own danger in acute HIT. Venous limb gangrene was more often associated with warfarin than limb arterial thrombosis was (8 of 8 against 3 of 10 patients). The median INR in patients with gangrene was 5·8, against 3·1 in 58 without gangrene. The proposed mechanism is a warfarin-induced failure of the protein C pathway while thrombin generation is still high.⁵⁸ The numbers are small and the series is observational, but the finding matches the guideline caution on warfarin set out below.

The 4Ts score

The 4Ts score is a pretest probability score built from four features: the thrombocytopenia, its timing, thrombosis or other sequelae, and other causes of a low count (Table 9).⁷ Its value lies almost entirely in a low score, which rules HIT out.

The score was first tested in 100 consecutive patients in Hamilton, Canada, and 236 in Greifswald, Germany. With a low score, HIT was confirmed in 1 of 64 patients (1·6%) in Hamilton and in none of 55 in Greifswald. With an intermediate score the figures were 8 of 28 (28·6%) and 11 of 139 (7·9%); with a high score, 8 of 8 (100%) and 9 of 42 (21·4%). The authors concluded that a low score rules HIT out in most settings, but that the meaning of an intermediate or high score varies with the setting.⁵⁹

A later meta-analysis pooled 13 studies of 3068 patients, of whom 1712 (55·8%) had a low score. The negative predictive value of a low score was 0·998 (95% CI 0·970–1·000), and it held regardless of who scored the patient, the prevalence of HIT and the population. The positive predictive value was modest: 0·14 (0·09–0·22) for an intermediate score and 0·64 (0·40–0·82) for a high score.⁷ In practice, a low score ends the inquiry, and an intermediate or high score only starts it. Other scores, such as the HIT Expert Probability score and the Lillo-Le Louët score, still need prospective validation.²

Laboratory testing

There are two classes of test. Immunoassays detect antibodies to PF4–heparin complexes. Functional assays detect antibodies that activate platelets, and only they show that the antibody is the platelet-activating kind that causes disease.² The reference functional assays are the serotonin-release assay (SRA) and the heparin-induced platelet activation (HIPA) assay, both of which use washed platelets.²

Immunoassays. Enzyme-linked immunoassays have an excellent negative predictive value, but a specificity of only 40–80%.² A meta-analysis of 49 publications (128 test evaluations, 15 199 patients) found that only five tests combined a sensitivity above 95% with a specificity above 90%: polyspecific ELISA at an intermediate threshold, the particle gel immunoassay, the lateral flow immunoassay, polyspecific chemiluminescent immunoassay at a high threshold, and IgG-specific chemiluminescent immunoassay at a low threshold. An IgG-specific ELISA at a low threshold was borderline (sensitivity 99·6%, specificity 89·9%). Accuracy was inadequate with high thresholds, with IgG-specific tests at intermediate thresholds, with a high-dose heparin confirmation step, and with the particle immunofiltration assay.⁶⁰ The performance of an immunoassay therefore depends on the named test and the cut-off the laboratory uses. An expert review found no overall advantage of IgG-specific over polyspecific ELISAs;² the meta-analysis found that the answer depends on the threshold, as set out above. Clinicians should know which assay their laboratory runs.

Functional assays. The SRA was described in 1986 as a two-point assay, at therapeutic and at high heparin concentrations, measuring the release of serotonin rather than aggregation. It was positive in only 1 of 573 controls (specificity 99%), in all 6 patients with definite HIT, and in none of 4 in whom HIT was unlikely.⁶¹ These numbers are small. The HIPA assay, a visual test in microtitre wells, was as sensitive as the SRA and better than platelet aggregometry in 34 patients with suspected HIT.⁶² The SRA is not perfect. About 4% of sera give an indeterminate result, and false positives occur; requiring a positive immunoassay as well lowers that risk. Platelet activation in the SRA without heparin marks unusually severe HIT, such as delayed-onset HIT or HIT with consumptive coagulopathy.⁶³ In one laboratory, a cohort of 430 patients contained one case of SRA-negative HIT, detected with a PF4-enhanced SRA. This gave an estimated SRA sensitivity of 97·2% (95% CI 85·8–99·9%) for that laboratory.⁶⁴

The ASH guideline sets the sequence. Use the 4Ts score rather than clinical impression. With an intermediate or high score, do an immunoassay. If the immunoassay is positive and a functional assay is available, do one.⁸

Management: the ASH 2018 guideline

The ASH 2018 guideline makes 33 recommendations.⁸ Its own systematic review found that the evidence was mostly non-comparative studies and case series, so the certainty is low to very low for most treatment questions.⁶⁵ Even so, several recommendations are strong. The guideline divides the course of HIT into five phases, and the recommendations refer to them: suspected HIT (count falling, tests pending); acute HIT (confirmed, until the count recovers); subacute HIT A (count recovered, functional assay still positive); subacute HIT B (functional assay negative, immunoassay still positive); and remote HIT (both negative).⁸ Table 10 summarises the recommendations in order.

Doses are as stated in the cited sources; check local protocols and product labelling.

Choosing the anticoagulant. The guideline gives no single preferred drug. It prefers argatroban or bivalirudin in critical illness, at high bleeding risk, or when an urgent procedure is likely. Argatroban should be avoided or its dose reduced in Child–Pugh class B or C liver disease. Fondaparinux or a DOAC is reasonable in a stable patient at average bleeding risk. Most DOAC experience is with rivaroxaban. For acute HIT with thrombosis the dose is 15 mg twice daily for 3 weeks, then 20 mg daily. For isolated HIT it is 15 mg twice daily until the count recovers, usually to 150 × 10⁹/L or more, then 20 mg daily.⁸

Warfarin and platelet transfusion. A vitamin K antagonist started while the count is still low risks venous limb gangrene, as described above;⁵⁸ the guideline makes a strong recommendation against it.⁸ For platelet transfusion, the main evidence is an analysis of a US national inpatient database: among 6332 admissions with HIT, the 7·1% who received platelets had more arterial thrombosis (adjusted OR 3·4, 95% CI 1·2–9·5) and higher in-hospital mortality (adjusted OR 5·2, 2·6–10·5), with no difference in venous thrombosis.⁶⁶ This is an association in administrative data, not a trial, and sicker patients may have been more likely to receive platelets. The guideline's recommendation against routine transfusion is conditional, and it allows transfusion for active bleeding or high bleeding risk.⁸ Its systematic review judged that inferior vena cava filters and platelet transfusion may cause net harm.⁶⁵ The evidence against filters is small: 9 of 10 patients with acute HIT and a filter had a new thromboembolic event.⁸

Duration. In isolated HIT with no thrombosis on screening ultrasound, anticoagulation continues at least until the count recovers. The guideline advises against continuing it for 3 months or more unless HIT persists. It did not address duration in HIT with thrombosis, in which anticoagulation is typically given for 3–6 months.⁸

Re-exposure and cardiac surgery

HIT antibodies are transient, and this makes later heparin possible in selected patients. Seven patients given heparin again after their antibodies had become undetectable did not develop a new episode.⁵² In the guideline's evidence, none of 3 patients given a second course of heparin more than 100 days after acute HIT developed recurrent HIT.⁸ These numbers are very small. The guideline accepts re-exposure in one setting: heparin during cardiac surgery in subacute HIT B or remote HIT, kept to the operation itself, with platelet monitoring afterwards because delayed-onset HIT 5–10 days later has been reported.⁸ For any other anticoagulation in remote HIT, it recommends a non-heparin drug.⁸

The ISTH Scientific and Standardization Committee guidance of 2026 on cardiac surgery with cardiopulmonary bypass rests on a systematic review of 30 studies. Bivalirudin has the strongest supporting data, but it increases bleeding and has no reversal agent. Therapeutic plasma exchange, with or without IVIG, lowers anti-PF4/heparin antibody levels and permits heparin during surgery, but the protocols vary widely. Potent antiplatelet agents given with heparin show promise, at a cost of bleeding and hypotension. In patients with negative functional assays, limited heparin during surgery carries a low risk of recurrent HIT if the count is monitored afterwards.⁴⁷ This is consistent with the ASH position.

The ACCP 2012 guideline

The American College of Chest Physicians guideline of 2012 predates the ASH guideline. It suggested platelet monitoring every 2–3 days from day 4 to day 14 when the risk of HIT is above 1%, and no monitoring when it is below 1%. For HIT with or without thrombosis and normal renal function it suggested argatroban, lepirudin or danaparoid, and argatroban in renal insufficiency. It suggested bivalirudin for urgent cardiac surgery in acute or subacute HIT. All of these were Grade 2C, which marks a weak recommendation. It called for further study of fondaparinux and the new oral anticoagulants.⁶⁷ Where ACCP and ASH differ, as in the list of drugs and the monitoring thresholds, the ASH guideline is the more recent.

DOACs and fondaparinux: the evidence

No randomised trial has compared anticoagulants in HIT. Every comparison below is between non-randomised groups. The evidence is best read as showing that DOACs and fondaparinux are reasonable choices in stable patients, not that one drug is better than another.

Case series. At Hamilton, 10 of 10 patients treated with rivaroxaban recovered without new, progressive or recurrent thrombosis. Pooled with published cases, thrombosis occurred in 1 of 46 rivaroxaban-treated patients (2·2%, 95% CI 0·4–11·3%) and major bleeding in none. Experience with apixaban (12 patients) and dabigatran (11) was smaller.⁶⁸

A retrospective registry. A registry of 195 patients with a 4Ts score of 4 or more compared fondaparinux, used off-label, with the approved drugs. First-line treatment was argatroban in 46, lepirudin in 4, danaparoid in 61 and fondaparinux in 84. HIT-specific complications occurred in 11·7% on the approved drugs and in none on fondaparinux, and in-hospital mortality was 14·4% against 0%. Bleeding was similar (6·3% against 4·8%). HIT was confirmed after the event in at least 74 of the 195 patients (38%), 35 of whom had received fondaparinux.⁶⁹ Many patients in the registry may therefore not have had HIT, and the groups were not randomised, so the difference cannot be taken as a treatment effect.

Meta-analyses. A meta-analysis of 92 studies (119 treatment groups, 4698 patients) gave pooled rates by drug. Platelet recovery ranged from 74% with bivalirudin to 99% with fondaparinux; thromboembolism from 1% with fondaparinux to 7% with danaparoid; major bleeding from 1% with a DOAC to 14% with bivalirudin; and death from 7% with fondaparinux to 19% with bivalirudin. The confidence intervals mostly overlapped.⁷⁰ The guideline prefers bivalirudin in critical illness and at high bleeding risk,⁸ so its patients are likely to have been sicker, which may explain part of its higher rates. A 2026 meta-analysis of DOACs pooled 44 publications, mostly case reports, with 352 patients; 53·9% had thrombosis at diagnosis and the mean nadir was 63 × 10⁹/L. The pooled rate of thrombosis on treatment was 6·4% (95% CI 4·2–9·2%) and of bleeding 3·9% (2·2–6·2%), with no difference between DOACs or between first-line and later use. The authors rated the certainty of the evidence as very low.⁷¹ The ASH systematic review reached a similar view: there may be little difference between the parenteral non-heparin anticoagulants and DOACs in acute HIT, at low to very low certainty.⁶⁵

In autoimmune HIT and HIT with disseminated intravascular coagulation, the APTT-adjusted drugs often fail, and expert reviews favour danaparoid, fondaparinux or a DOAC, with high-dose IVIG.⁵⁴

Drugs by system

The sections that follow list the implicated drugs by therapeutic class. Each table gives the mechanism where it is known, the usual timing and severity, and the references, and the prose states how strong the evidence is for each drug. Heparin is covered in its own section above (heparin-induced thrombocytopenia) and appears here only as a pointer.

Cardiovascular drugs

Two different kinds of problem sit under this heading. The glycoprotein IIb/IIIa inhibitors cause thrombocytopenia often enough to be counted in randomised trials, and they can do so within hours of a first dose. The older cardiovascular drugs, with quinidine at their head, cause classic drug-dependent immune thrombocytopenia; for most of them the evidence is case reports, with a drug-dependent antibody shown for a few. The thienopyridines are a third case: their main haematological hazard is thrombotic thrombocytopenic purpura (TTP), not isolated thrombocytopenia.

Glycoprotein IIb/IIIa inhibitors

These drugs are the one class in which a drug can cause severe thrombocytopenia within hours of the first dose. The antibodies that react with tirofiban or eptifibatide are often present before the patient has ever received the drug. In 9 patients whose count fell below 25 × 10⁹/L within hours of tirofiban or eptifibatide, all had antibodies that bound GPIIb/IIIa only when the drug was present. Five had never had the drug before, and in two of these the antibody was already in serum taken before treatment. No such antibody was found in 100 healthy blood donors.²⁴ Severe thrombocytopenia is reported in 0·1–2% of patients given tirofiban or eptifibatide.^4,13

Abciximab works differently. It is a chimeric Fab fragment, and the antibody recognises the drug bound to the platelet. About 1–2% of patients become thrombocytopenic on first exposure.^13,26 Nine patients whose count fell below 10 × 10⁹/L within hours of a second exposure all had a strong IgG against abciximab-coated platelets. Such IgG was also present in 74% of 104 healthy people; the antibodies of patients differed in reacting more strongly with platelets coated with the parent murine antibody.²⁶ A delayed form starts 3–6 days after the drug is stopped, usually after the patient has left hospital.²⁷ It can appear up to 8 days later because the drug stays on platelets for up to 2 weeks.⁴

The ReoPro readministration registry measured the risk of a second course. Among 1342 readministrations, the count fell below 100 × 10⁹/L in 5% and below 20 × 10⁹/L in 2%. When the second course came within a month of the first (115 patients), the figures were 16·5% and 12·2%. Thrombocytopenia was more frequent in patients with human anti-chimeric antibodies before readministration (14·1% vs 4·4%).³⁹ This is registry evidence, without a control group.

Before any of this, the low count must be shown to be real. In a pooled analysis of four trials of abciximab (CAPTURE, EPIC, EPILOG and EPISTENT), pseudothrombocytopenia occurred in 2·1% of patients on abciximab and 0·6% on placebo, and true thrombocytopenia in 3·7% and 1·8%. Pseudothrombocytopenia made up 36% of the low counts on abciximab. Unlike true thrombocytopenia, it was not linked to bleeding, transfusion, stroke or repeat revascularisation.¹⁴ A blood film settles the question.¹ Stopping the drug for an artefact removes an effective treatment at a critical time.

True thrombocytopenia in these trials was a marker of poor outcome. In the pooled EPIC, EPILOG and EPISTENT data, the count fell below 100 × 10⁹/L in 178 of 7290 patients (2·4%). Older age, weight below 90 kg, a baseline count below 200 × 10⁹/L and abciximab were independent predictors. Bleeding and transfusion were more frequent, and 30-day mortality was 8·4% with thrombocytopenia against 0·6% without.⁷⁶ In the PURSUIT trial of eptifibatide in non-ST-elevation acute coronary syndrome, thrombocytopenia (below 100 × 10⁹/L or a fall of more than 50%) occurred in 7·0% of patients, with a median onset of 4 days in both arms. It more than doubled the risk of moderate or severe bleeding and was associated with death or myocardial infarction at 30 days. In that trial neither heparin nor eptifibatide independently raised the risk of thrombocytopenia.⁷⁷ These are associations within trials. They do not show that the drug caused the fall in each patient.

A 2023 critical review grades thrombocytopenia with these drugs as mild (50–100 × 10⁹/L), severe (20 to below 50 × 10⁹/L) and profound (below 20 × 10⁹/L). Profound thrombocytopenia occurs in fewer than 1% of patients given a first course. Onset can be acute, within 24 hours, or delayed up to about 14 days, and both bleeding and thrombosis occur. The diagnosis requires pseudothrombocytopenia and heparin-induced thrombocytopenia to be excluded. Management follows severity: stop the drug, and use steroids, intravenous immunoglobulin or platelet transfusion as needed.⁷⁸ This advice comes from expert review. No randomised trial has tested any treatment beyond stopping the drug.

Quinidine, procainamide and amiodarone

Quinidine has the strongest evidence of any drug in this section. It was named in 38 of the case reports with definite or probable evidence in the 1998 Oklahoma systematic review, more than any other drug.²² It is one of six drugs with ten or more definite or probable case reports and drug-dependent antibodies in ten or more patients,²³ and its laboratory diagnosis has been confirmed in two independent laboratories.³ In the 1993 case–control study, use of quinidine or quinine in the week before onset carried a relative risk of 101.¹⁷ The reaction follows the classic pattern: onset 5–10 days after the first dose, or within hours on re-exposure, and a nadir that is classically below 10 × 10⁹/L.⁴ Quinine is discussed with the anti-infectives and under drug-induced thrombotic microangiopathy.

Procainamide is placed among the drugs that induce a true autoantibody, which binds platelets without the drug. The practical point is that thrombocytopenia may not resolve promptly when the drug is stopped, unlike the drug-dependent type.⁴ No primary series of procainamide thrombocytopenia was found.

Amiodarone is an unusual entry. It is one of the 24 drugs with evidence from both case reports and antibody testing, but it is the only one of them that data mining of the US adverse event database did not flag.²³ Its laboratory grade is probable, from one laboratory.³ A 2021 case report describes delayed-onset immune thrombocytopenia on amiodarone in a 72-year-old man.⁷² The evidence is real but thin, and no figure for frequency, onset or nadir is available.

Diuretics, digoxin and other drugs

The evidence for the diuretics is weak and mixed. Thiazides have been linked to thrombocytopenia that may come from suppression of megakaryocytes rather than peripheral destruction, and chlorothiazide appears on older lists of definite case reports.⁴ In the Berlin case–control surveillance study, triamterene–hydrochlorothiazide carried a significantly increased risk.¹⁸ A 1986 case report describes thrombocytopenic purpura in a 65-year-old man about a year after he started hydrochlorothiazide 50 mg daily, resolving two weeks after it was stopped.⁷³ Against this, a Dutch population case–control study found no significant increase in risk with diuretics, although it had little power to detect one.²⁰ Furosemide has a probable laboratory grade, from one laboratory.³

Digoxin is named on older lists of definite case reports,⁴ but no antibody study or case series was found, and it should be regarded as a list entry. Dipyridamole carried a relative risk of 14 in the 1993 case–control study, with an excess risk of 3·9 cases per million users a week.¹⁷ Simvastatin is one of the drugs positive by case reports, antibody testing and data mining,²³ but no figure for frequency, onset or nadir was found for any statin.

Thienopyridines and thrombotic thrombocytopenic purpura

With ticlopidine, clopidogrel and prasugrel the concern is TTP.

Ticlopidine carries the clearest risk. In a review of 60 cases, the drug had been taken for less than a month in 80%, and most patients had had a normal count within the two weeks before onset, so routine monitoring of the count did not predict it. Mortality was 50% without plasma exchange and 24% with it.⁴⁶ In a survey of 43 322 patients stented at 63 centres over one year, the incidence was 1 in 4814 (0·02%). Onset came a mean of 22 days after starting the drug (range 5–60). Four of 19 patients died, all of them among those not given plasma exchange; none of the 13 who had plasma exchange died. Limiting ticlopidine to two weeks after stenting did not prevent TTP.⁷⁴

Clopidogrel is associated with TTP much less often. The first series described 11 patients; 10 had taken the drug for 14 days or less. At that time no case had been seen among 20 000 closely monitored patients in trials and cohort studies, against an estimated 1 case per 1600–5000 patients treated with ticlopidine.⁷⁵ Prasugrel TTP has also been reported, with 14 reports to the US Food and Drug Administration by 2011.³⁵

The two main drugs appear to cause TTP in different ways. In a series of 128 patients, ticlopidine-associated cases more often followed more than two weeks of the drug (90% vs 26% for clopidogrel), were more often severely thrombocytopenic (84% vs 60%) and more often had normal renal function (72% vs 45%). Severe ADAMTS13 deficiency was strongly associated with ticlopidine. When TTP began more than two weeks after starting a thienopyridine, plasma exchange was associated with survival of 84% against 38% without it; when it began within two weeks, survival was 77% with plasma exchange and 78% without.³⁶ A 20-year review found severe ADAMTS13 deficiency in 80% and an antibody in all patients with ticlopidine-associated TTP, and in none with clopidogrel-associated TTP.³⁵ Read together, these series suggest that late-onset TTP on a thienopyridine, most often ticlopidine, behaves like acquired TTP and responds to plasma exchange, while early-onset TTP, more typical of clopidogrel, may not. This evidence comes from case series and pharmacovigilance reports, not from trials.

Anaesthesia and intensive care

Thrombocytopenia is common in intensive care, and drugs are one cause among several. A systematic review of 24 studies with 6894 patients found thrombocytopenia on admission in 8·3–67·6% and developing during the stay in 13·0–44·1%. The studies differed too much to pool, so these are ranges, not an estimate. Severe illness, sepsis and organ dysfunction were associated with it, and 6 of 8 studies with multivariable analysis found that it increased the risk of death.⁴² In 3746 critically ill patients receiving thromboprophylaxis in a randomised trial, thrombocytopenia was mild (100–149 × 10⁹/L) in 15·3%, moderate (50–99 × 10⁹/L) in 5·1% and severe (below 50 × 10⁹/L) in 1·6%.⁷⁹ An expert review puts the frequency at about 40%.⁴³

How much of this is caused by drugs is uncertain, and the two main sources point in different directions. A 2010 expert review states that up to 25% of acutely ill patients develop drug-induced thrombocytopenia, and names the glycoprotein IIb/IIIa inhibitors, the cinchona alkaloids, antibiotics, anticonvulsants and heparin as the classes most often involved.⁴¹ The abstract gives no study for the 25% figure, so it is best read as an estimate. A 2014 single-centre case–control study tested the attribution directly in 238 patients with thrombocytopenia acquired in intensive care, each matched to one control. Quinolones and extended-spectrum beta-lactams were associated with thrombocytopenia in unadjusted analysis. After adjustment only quinolones remained significant (OR 1·70, 95% CI 1·00–2·87), and the confidence interval only just excluded no effect. No association was found for anticonvulsants, antiplatelet drugs, NSAIDs or heparins. Linezolid was not studied.⁸⁰ Both findings belong in the assessment: drugs are a real cause in intensive care, but a single study could not confirm most of the classes usually blamed.

Drugs, sepsis and consumption

The diagnosis is hard in intensive care because the patient takes many drugs and has other reasons for a low count.^2,4,41 Sepsis and organ dysfunction are themselves associated with thrombocytopenia,⁴² and causes other than drugs must be looked for and excluded before a drug is blamed.⁴¹ The first step is to confirm the count on a blood film, because pseudothrombocytopenia must be excluded.¹

The shape of the platelet curve is the most useful bedside guide. After major surgery a moderate fall in the first 3 days is expected. A slow fall over 5–7 days suggests consumption or marrow failure. An abrupt fall within 1–2 days, after the count had first recovered, at about 1–2 weeks after surgery, strongly suggests an immune cause: heparin-induced thrombocytopenia, another drug-induced thrombocytopenia, or post-transfusion purpura.⁴³ A fall of more than 50% within 1–2 days, especially after an intermittent rise, points the same way.⁴⁴ These rules come from expert reviews, not from validation studies. No study among the sources for this review compares the laboratory features of drug-induced thrombocytopenia with those of sepsis or disseminated intravascular coagulation at the bedside, so no discriminating test can be offered here.

When a drug is suspected, stopping it is the treatment.⁵ The prognosis after the drug is withdrawn is excellent; the roles of corticosteroids, immunoglobulin and plasma exchange are uncertain.⁴¹ Two drug groups need their own approach. Heparin is the drug for which the risk is thrombosis rather than bleeding; its assessment with the 4Ts score and its management are covered in the heparin section. In a randomised trial in 3746 critically ill patients, heparin-induced thrombocytopenia was less frequent with dalteparin than with unfractionated heparin in the prespecified per-protocol analysis (hazard ratio 0·27, 95% CI 0·08–0·98).⁵⁰ The glycoprotein IIb/IIIa inhibitors, which can cause a profound fall within hours and whose low counts are often artefactual, are covered under cardiovascular drugs.

Anti-infective drugs

Anti-infective drugs carry the strongest evidence of any class. The 16 drugs with a definite laboratory diagnosis include quinine, trimethoprim–sulfamethoxazole, vancomycin, penicillin, rifampicin and ceftriaxone,³ and in the only case–control study to give relative risks for single drugs, trimethoprim–sulfamethoxazole carried the highest risk of all.¹⁷ A Dutch population study found a raised risk with current beta-lactam use (adjusted OR 7·4, 95% CI 1·8–29·6),²⁰ and the Berlin surveillance study found a significantly raised risk with gentamicin.¹⁸

Two different problems sit in this class, and they are managed differently. Most of these drugs cause immune thrombocytopenia through a drug-dependent antibody: the fall is sudden and severe, bleeding is common, and the drug must be stopped and then avoided. Linezolid is different. Its effect depends on the length of treatment, the drug level and kidney function, and the answer may be a lower dose rather than a ban. A third group, the antivirals ganciclovir, valganciclovir and zidovudine, is toxic to the marrow as a whole (Table 14, Table 15).

Vancomycin

Vancomycin is easily overlooked as a cause, because it is given to patients who often have other reasons for a low count. The defining laboratory series found drug-dependent antibodies of IgG class, IgM class or both in 34 patients referred over 5 years, with follow-up on 29. The mean nadir was 13·6 × 10⁹/L, and 10 patients (34%) had severe bleeding. The count returned to baseline in all 26 survivors after vancomycin was stopped. In 15 patients the drug was continued for 1–14 days while other causes were sought. The antibodies were not found in 25 patients who received vancomycin and kept a normal count, and this control group is what makes the causal argument.³⁸ Vancomycin meets the definite laboratory grade.³

A narrative review of the case reports found a mean of 8 days to the nadir, shorter on re-exposure. Antibodies were found in 13 of 17 patients tested. In 17 patients the count was normal again within 5–6 days of stopping; in the one patient who stayed on the drug, recovery came 11 days after the nadir.⁸¹ A retrospective cohort of 1269 patients on intravenous vancomycin gives a rate for an unselected hospital population. Thrombocytopenia occurred in 3·3%, a median of 9 days (range 2–22) after starting, and 45·2% of these reached grade 3–4. Recovery, which came only after stopping, took a mean of 9 days (range 1–35). The independent risk factors were a qSOFA score of 2 or more, renal disease, treatment for 8 days or longer, a baseline count of 150 × 10⁹/L or less, and a blood urea above 12 mmol/L. The rate rose from 0·2% with none of these factors to 62·5% with all five.⁸² The report does not describe antibody testing, so the cohort probably counts every fall in the count on vancomycin, immune or not.

In practice, a patient on vancomycin whose count falls sharply after about a week should have the drug stopped rather than continued while sepsis or consumption is investigated. The 15 patients in the laboratory series who stayed on the drug show how easily this happens.³⁸

Linezolid

Thrombocytopenia on linezolid is common, and it behaves like a toxic effect, not an allergy. The summary of the clinical trial programme found thrombocytopenia, with a slight increase in anaemia, after 2 weeks or more of treatment, and described it as mild, reversible, duration-dependent myelosuppression.⁸³

The mechanism is disputed. A 2003 report of one patient found normal-looking megakaryocytes in the marrow alongside ringed sideroblasts, and the authors argued for an immune mechanism for the thrombocytopenia, citing a slower fall in the count after intravenous immunoglobulin.⁸⁴ A 2013 review states the opposite position: thrombocytopenia tends to occur with treatment longer than 2 weeks and at high drug levels, platelets are more affected than other cell lines, and neither a linezolid-dependent antiplatelet antibody nor a specific effect on megakaryocytes has been shown.⁴ The first position rests on a single case; the second is a statement that the laboratory evidence is absent. The practical evidence, on duration, drug level and kidney function, fits a dose-dependent effect.

How often it happens depends on how it is defined. A meta-analysis of 40 observational studies with 6454 patients gave a pooled rate of 37%, across definitions and populations that differ widely. Older age, renal and liver disease, low baseline count, low albumin, duration of treatment and renal replacement therapy were associated with it.⁸⁵ In a cohort of 264 patients without haematological or oncological disease, thrombocytopenia occurred in 29·2% after a mean of 11·2 days. The risk rose with duration: the adjusted odds ratio was 7·25 for 7–13 days, 19·51 for 14–20 days and 28·80 for 21 days or more. A baseline count below 150 × 10⁹/L (adjusted OR 5·08), creatinine clearance below 30 mL/min (adjusted OR 4·19) and low-dose aspirin (adjusted OR 2·99) also raised the risk, and a low baseline count predicted an early fall, within 6 days.⁸⁶

Kidney function is the risk factor that can be acted on. Among 341 patients who received linezolid for 10 days or more, thrombocytopenia occurred in 42·9% of those with renal impairment and 16·8% of those without (adjusted HR 2·37, 95% CI 1·52–3·68). A pharmacokinetic model built on 603 patients found that a lower dose improved the chance of reaching the target trough of 2–8 mg/L in simulated patients with an eGFR below 60 mL/min.⁹ Whether dosing by drug level actually prevents thrombocytopenia is less certain. A 2025 meta-analysis of four studies (208 patients with personalised dosing, 195 with standard dosing) found no significant effect in its main analysis (OR 0·648, 95% CI 0·150–2·797), with severe heterogeneity. Only after one study at high risk of bias was removed, in an analysis not planned in advance, did the result become significant (OR 0·356, 95% CI 0·179–0·708). The authors conclude that personalised dosing "may" lower the risk.⁸⁷

A retrospective single-centre cohort of 453 patients treated for at least 5 days found about four times as much thrombocytopenia with linezolid as with vancomycin (OR 4·39, 95% CI 2·38–8·08).⁹⁷ The two drugs were not assigned at random, so sicker patients may have received one more than the other.

For the clinician, this means checking the count during any course longer than a week or two, taking particular care in renal impairment, and considering a lower dose, or measuring drug levels where that is available. Because no antibody has been shown, a fall on linezolid is not handled like an immune reaction, which calls for lifelong avoidance.

Beta-lactams

Penicillin is the classic example of a hapten: the drug binds covalently to platelet proteins and the antibody is directed at the drug-protein complex. It meets the definite laboratory grade, although penicillin is better known for immune haemolysis.^3,4 Ceftriaxone is also graded definite, and piperacillin and ampicillin probable.³ Piperacillin and ampicillin are among the 24 drugs with both case-report and antibody evidence.²³

Re-exposure is the dangerous setting. Two patients who were given piperacillin–tazobactam again after a recent course developed severe thrombocytopenia within a short time; both improved quickly when it was stopped and they were given steroids and intravenous immunoglobulin.⁸⁸ A patient who has had a recent course of a beta-lactam may therefore fall within hours rather than after a week.⁴

The most useful finding for choosing the next antibiotic comes from a laboratory study of antibodies from 32 patients who became thrombocytopenic on a penam (such as piperacillin) or a cephem (such as ceftriaxone). Of 14 penam-specific antibodies, 5 cross-reacted strongly with other penams. Of 18 cephem-specific antibodies, 8 cross-reacted strongly with other cephems. Antibodies raised by a penam did not cross-react strongly with cephems, or the reverse. Cross-reactions followed the R1 side group: drugs with similar or identical R1 side groups cross-reacted. The authors conclude that a patient who has had immune thrombocytopenia on a penam can generally be switched to a cephem, and the reverse; and that if a drug from the same group must be used, it should have a clearly different R1 side group.⁸⁹ This is a single laboratory study of antibody behaviour in vitro, not a clinical trial of switching, but it addresses the question directly.

The usual rule that the count starts to recover within 1–2 days of stopping the drug⁵ assumes that the drug is cleared normally. A 65-year-old woman with liver and kidney failure had a stable count of about 70 × 10⁹/L, which fell to 3 × 10⁹/L two days after a 5-day course of ceftriaxone. It stayed at 1–6 × 10⁹/L until she died 13 days later, despite intravenous immunoglobulin, steroids and platelet transfusion. A strong ceftriaxone-dependent antibody was found 7 days after the last dose, and ceftriaxone was still in her serum at least 8 days after it.⁴⁰ In a patient with poor drug clearance, a slow recovery does not argue against the diagnosis.

Trimethoprim–sulfamethoxazole and other sulfonamides

Trimethoprim–sulfamethoxazole has the widest range of evidence of any drug in this section. In the 1993 case–control study, its use in the week before onset carried a relative risk of 124, the highest of any drug, and an excess risk of 38 cases per million users a week.¹⁷ It was named in 10 of the case reports graded in 1998,²² it is one of seven drugs with 10 or more definite or probable case reports,²³ and it accounted for 4 probable or certain cases in the Berlin surveillance study.¹⁸ The laboratory evidence is definite. Antibodies have been shown with trimethoprim, with sulfamethoxazole, with their metabolites and with urine collected after the drug was taken, so a test with the parent drug alone may miss the antibody.³ Sulfisoxazole is also among the 24 drugs with both case-report and antibody evidence, and is graded probable.^3,23

Because the fall comes 5–10 days after starting,⁴ a short course may be over before the count is measured. The drug history should cover antibiotic courses finished recently, not only current drugs.

Rifampicin and other antituberculous drugs

Rifampicin is graded definite on laboratory evidence and is one of seven drugs with 10 or more definite or probable case reports.^3,23 In one patient who developed purpura during antituberculous treatment, the drug-dependent IgG was directed at the GPIb/IX complex and bound only when rifampicin was present. The same report notes that rifampicin has been recognised as a cause of immune thrombocytopenia during intermittent high-dose treatment.⁹⁰ Ethambutol is among the 24 drugs with both case-report and antibody evidence, and is graded probable.^3,23 The Dutch population study found no significant increase in risk with antituberculous drugs as a group, but it had little power to detect one.²⁰

A patient on combination treatment who develops immune thrombocytopenia is taking several candidate drugs at once. In the laboratory grading, rifampicin is definite, ethambutol probable, and no other antituberculous drug is listed as definite or probable.³

Quinine and other antimalarials

Quinine is the model for the classic drug-dependent antibody, often called the quinine type. It classically causes severe thrombocytopenia, with a nadir below 10 × 10⁹/L.⁴ In the 1993 case–control study, quinidine or quinine carried a relative risk of 101 and an excess risk of 26 cases per million users a week.¹⁷ It is graded definite on laboratory evidence.³

The exposure is not always a prescription. Quinine is used off-label for leg cramps, and it is present in tonic water and in some aperitifs such as Dubonnet. The US Food and Drug Administration warned against its unapproved uses after 38 reports of serious adverse events in the United States between April 2005 and October 2008.⁴ A systematic review of acute immune reactions to quinine found 142 patients with definite or probable evidence. Quinine tablets caused 102 reactions (72%), quinine-containing drinks 28 (20%), and five other exposures 12 (8%). The reactions went well beyond thrombocytopenia: they included fever, hypotension, disseminated intravascular coagulation, haemolysis, neutropenia, acute kidney injury and liver toxicity. Of the 101 patients whose reactions were not limited to the skin, 92 (91%) were admitted with severe illness, 30 needed renal replacement therapy and 3 died.³⁴ Severe thrombocytopenia after drinking tonic water has been described in a case report.⁹² The drug history must therefore ask about drinks, and a patient may not think of tonic water as a drug.

Quinine is also the leading cause of drug-induced thrombotic microangiopathy: in a combined analysis of the Oklahoma Registry and the BloodCenter of Wisconsin, 47 of 51 patients (92%) with definite evidence had taken quinine.¹⁵ Any patient thought to have TTP or HUS should be asked about quinine; this is covered in the section on drug-induced thrombotic microangiopathy.

For antimalarials other than quinine, such as chloroquine, mefloquine and the artemisinins, the evidence was not assessed for this review. Hydroxychloroquine is discussed with the rheumatological drugs.

Fluoroquinolones

The evidence for fluoroquinolones is thinner but consistent. Ciprofloxacin is graded probable, on antibody testing in one laboratory.³ A 76-year-old man had severe thrombocytopenia on two separate courses of ciprofloxacin, and IgG antibodies against GPIIb/IIIa were found after the second. The authors suggested that the structural likeness of the quinolones to quinine may explain the reaction.⁹¹ Levofloxacin can cause a rapid fall on re-exposure. Another 76-year-old man, who had taken a course six months earlier, had a fall in the count from 187 to 5 × 10⁹/L over 4 days after a single 750 mg dose. He recovered after the drug was stopped and he was given intravenous immunoglobulin and dexamethasone. No levofloxacin-dependent antibody could be found. The authors put this down to the known low sensitivity of the tests and to a probable antibody against a metabolite.⁴⁵

That case makes a general point: a negative test for drug-dependent antibodies does not exclude the diagnosis. In intensive care, a single-centre case–control study found that quinolones were the only antibiotic class still associated with thrombocytopenia after adjustment (OR 1·697, 95% CI 1·002–2·873).⁸⁰ The confidence interval only just excludes no effect, and the report does not say how many of the cases were immune.

Antiviral drugs

The antivirals in this section lower the count by suppressing the marrow, not through an antibody, and the effect is not limited to platelets. After solid-organ transplantation, low white cells, platelets or haemoglobin occur in 5–50% of patients on anti-CMV treatment. Few studies address how often this happens or who is at risk, and the usual response, a lower dose of ganciclovir, may raise the risk of CMV reactivation and resistant disease.⁹³ In the main randomised trial of prophylaxis, in 364 solid-organ transplant recipients at high risk (CMV-positive donor, CMV-negative recipient), valganciclovir and oral ganciclovir had similar safety except for more neutropenia with valganciclovir (8·2% against 3·2%).⁹⁴ The abstract reports no difference in thrombocytopenia.

Zidovudine is toxic to the marrow, but the evidence is about red and white cells. In the placebo-controlled trial of 282 patients, haemoglobin fell below 7·5 g/dL in 24% on zidovudine and 4% on placebo, and neutrophils fell below 0·5 × 10⁹/L in 16% and 2%. Macrocytosis developed within weeks in most patients on the drug. The abstract gives no figure for thrombocytopenia.⁹⁵ In HIV the more common effect runs the other way: HIV-related thrombocytopenia improves on antiretroviral treatment, and studies of zidovudine have shown increased platelet production.⁹⁶

Rheumatology and immunology

The drugs in this group lower the platelet count in different ways, and the way decides what to do. Gold induces an autoantibody that does not need the drug, so the count may stay low after the drug is stopped. Low-dose methotrexate and leflunomide suppress the marrow; the reaction is often pancytopenia rather than isolated thrombocytopenia, it can come months or years into treatment, and it can kill. Azathioprine suppresses the marrow on a genetic background that can be tested for. Rituximab can cause an acute fall during the infusion and a separate, delayed fall weeks later. For sulfasalazine and hydroxychloroquine the evidence is a single case report each. The evidence for this group is mostly case series and cohort studies of patients with rheumatoid arthritis; the one randomised trial that counts thrombocytopenia as a main adverse event is a trial of a JAK inhibitor in myelofibrosis.

Gold and D-penicillamine

Gold shows something every clinician should know: a drug can cause thrombocytopenia and still give a negative drug-dependent antibody test. In a laboratory study of 10 patients with rheumatoid arthritis and profound thrombocytopenia on gold, the platelet autoantibodies in all 10 targeted glycoprotein V, and gold was not needed for them to react. No such antibodies were found in 28 patients on gold with normal counts.²⁸ The antibody is an autoantibody, not a drug-dependent one, and the picture resembles immune thrombocytopenia. The Wisconsin reference laboratory had tested for gold-dependent antibodies in the past with negative results. In the Oklahoma analysis gold had 11 case reports graded probable and none definite.²³ The laboratory review graded it probable on the evidence of one laboratory and noted that gold induces autoantibodies rather than typical drug-dependent antibodies.³ A negative test for a drug-dependent antibody therefore does not clear gold.

The older clinical series describe the same disease. Of 10 patients who became thrombocytopenic on sodium aurothiomalate without marrow aplasia, 8 had petechiae, purpura or bruising, but none had life-threatening bleeding. Eight responded to corticosteroids.⁹⁸ A series of 17 patients with gold-associated thrombocytopenia found two patterns. Ten had an early, sudden fall that looked immune, and all 10 were HLA-DR3 positive. Seven had a less dramatic, fluctuating fall; only two of them were DR3 positive, and all had also taken penicillamine. The authors concluded that HLA typing would not predict all patients at risk.⁹⁹ These are case series from the 1970s and 1980s.

D-penicillamine is poorly tolerated: in long-term studies only 30–40% of patients were still taking it at 2 years. Thrombocytopenia or leucopenia led to stopping the drug in 2–5% of patients, and patients who were HLA-B8/DR3 positive or poor sulfoxidisers were at higher risk of adverse effects.¹⁰⁰

Low-dose methotrexate and leflunomide

Both drugs have caused deaths. Low-dose methotrexate does not cause an immune thrombocytopenia. It suppresses the marrow, usually all three cell lines, and the fall is often triggered by something that raises exposure to the drug. A review of 70 published cases of methotrexate-related pancytopenia in rheumatoid arthritis found that 12 (17%) died. Most had impaired renal function, low albumin, an infection or five or more other drugs, and the smallest cumulative dose that led to fatal pancytopenia was 10 mg. In the five long-term prospective studies within that review, 7 of 511 patients (1·4%) developed pancytopenia.¹⁰² A cohort of 28 adults admitted with toxicity on 10–25 mg a week had a mean age of 70. Pancytopenia was the commonest finding (78·5%), and 7 (25%) died, all of pancytopenia followed by sepsis. The risk factors were acute renal failure, low albumin, interacting drugs and dosing errors. Serum methotrexate levels did not correlate with the counts and did not differ between those who died and those who survived, and the authors concluded that measuring drug levels has no place in low-dose toxicity.¹¹ In a series of five patients with serious pancytopenia, two were caused by prescription errors in primary care.¹⁰³ A 2026 systematic review of 108 studies reports haematological adverse effects of low-dose methotrexate in rheumatoid arthritis in 2·4–5·6% of patients, a range that covers anaemia, leucopenia, thrombocytopenia and pancytopenia together. It adds older age and inadequate folic acid supplementation to the risk factors, and describes folinic acid (leucovorin) as the intervention for pancytopenia.¹⁰⁴

Timing matters for monitoring. In a clinic cohort of 3923 courses of disease-modifying drugs in 2170 patients, most reactions that needed the drug stopped appeared within 6 months. Methotrexate was the exception: the median time to thrombocytopenia was 9·4 months and to neutropenia 16·9 months.¹⁰¹ Monitoring of the blood count cannot be confined to the first months of treatment.

Leflunomide behaves in the same way and is most dangerous with methotrexate. Of 18 Australian patients with leflunomide-associated pancytopenia, 14 were also taking methotrexate, and 5 died, 4 of them on both drugs. Onset ranged from 11 days to 4 years, with a median of 4 months. In one patient, four years of stable treatment ended in fatal pancytopenia after acute renal failure from dehydration and an NSAID.¹⁰⁶ Using spontaneous reports and national prescribing data, the estimated incidence of pancytopenia was 1 in 3698 to 1 in 4582 patients on leflunomide alone and 1 in 575 to 1 in 822 on leflunomide with methotrexate.¹⁰⁵ These estimates come from spontaneous reports, not from a cohort.

In practice, a patient on low-dose methotrexate or leflunomide who presents with a low platelet count should have a full blood count, not a platelet count alone, with renal function and albumin, a check of the dose actually taken each week, and a review of new drugs and recent illness.

Azathioprine and the thiopurines

Azathioprine and mercaptopurine suppress the marrow, and the risk depends partly on two genes. The Clinical Pharmacogenetics Implementation Consortium guideline (2018 update, published 2019) explains that TPMT variants with low enzyme activity, and loss-of-function variants in NUDT15, reduce the breakdown of active thiopurine metabolites and predispose to myelosuppression. NUDT15 variants are common in people of Asian and Hispanic ancestry. The guideline gives starting doses adjusted by genotype;¹⁰⁹ the doses themselves are in its online tables and should be taken from there. A case–control study of 491 patients with thiopurine-induced myelosuppression and 679 tolerant controls confirmed TPMT (odds ratio 2·3) and found that carrying any of three coding NUDT15 variants raised the odds 27-fold, independent of TPMT genotype and dose. It also found that TPMT testing identifies only 25% of patients of European ancestry who are at risk.¹¹⁰ That study defined myelosuppression by the white-cell count, not the platelet count. It does not give a risk of thrombocytopenia, and should not be read as one.

Sulfasalazine and hydroxychloroquine

Both rest on single case reports. Sulfasalazine caused immune thrombocytopenia in one reported patient with rheumatoid arthritis,¹⁰⁷ and it appears on the historical lists of implicated drugs.⁴ Hydroxychloroquine caused severe thrombocytopenia in one patient with systemic lupus erythematosus, with recurrence on re-exposure; the authors established the diagnosis by excluding other causes and by that recurrence.¹⁰⁸ No incidence is known for either drug.

Biologics and JAK inhibitors

With the TNF inhibitors, thrombocytopenia is very rare. A review of their non-malignant haematological effects found no serious adverse events linked to it, whereas transient neutropenia developed in up to 16% of patients.¹¹¹ In 187 patients with psoriasis treated over nine years, none had a count below 50 × 10⁹/L or any bleeding, and two had counts below 100 × 10⁹/L.¹¹²

Rituximab causes two different kinds of fall. The acute kind comes during or just after the infusion. In one patient studied in detail, TNF-α was raised, complement was reduced, and neither flow cytometry nor MAIPA found a rituximab-dependent antiplatelet antibody. Among the 15 further patients found in the literature, mantle cell lymphoma was the commonest diagnosis, 10 had splenomegaly, 6 had marrow involvement, and in 10 the fall followed a cytokine release or infusion reaction. Bleeding was usually absent.¹¹³ The mechanism is uncertain: immune complexes with complement-mediated destruction and direct binding to CD20 on platelets have both been proposed, and features of disseminated intravascular coagulation may be present.⁴ The delayed kind comes weeks later. In 77 consecutive rituximab courses in patients with non-Hodgkin lymphoma, delayed thrombocytopenia occurred in 8 (10·4%), a median of 4 weeks after the last dose; delayed neutropenia was commoner (27·3%). Only previous chemotherapy and more than four doses of rituximab predicted delayed cytopenia.¹¹⁴ These patients had lymphoma and most had had chemotherapy, so the rates do not transfer to rituximab given for an autoimmune disease.

For tocilizumab, a randomised comparison with sarilumab in rheumatoid arthritis listed lower neutrophil and platelet counts among the laboratory changes, without a rate for thrombocytopenia.¹¹⁵ No source giving its frequency, timing or severity was found.

The JAK inhibitors show that a drug class does not predict the platelet effect. In the COMFORT-I trial, 155 patients with myelofibrosis were randomised to ruxolitinib and 154 to placebo. Anaemia and thrombocytopenia were the commonest adverse events on ruxolitinib, mostly early in treatment, but each led to stopping the drug in only one patient.¹¹⁶ In rheumatoid arthritis, the common adverse reactions listed for baricitinib include thrombocytosis, not thrombocytopenia.¹¹⁷ The myelofibrosis experience should not be applied to the rheumatological use of these drugs.

Analgesics and anti-inflammatory drugs

Analgesics matter because almost every patient takes them, often without a prescription, and they are easily left out of a drug history. In the Danish adverse reaction register from 1968 to 1991, NSAIDs were the class most often reported with thrombocytopenia;¹¹⁸ this reflects reporting, since the register has no count of users. The one case–control study found a relative risk of 2·6 for salicylates, with an excess risk of 0·4 cases per million users a week, the smallest of the significant associations.¹⁷ The laboratory evidence is stronger for three drugs. Ibuprofen is one of the 16 drugs with a definite laboratory diagnosis.³ Paracetamol (acetaminophen) and naproxen cause thrombocytopenia through antibodies that recognise a metabolite of the drug, not the drug itself, so a test with the parent drug is negative.^23,30

The key study is a laboratory series of five patients. Three became thrombocytopenic after naproxen and two after paracetamol, and no antibody could be found using the drugs themselves. In every patient an antibody that reacted with normal platelets was found when a known metabolite was added: naproxen glucuronide or acetaminophen sulfate. Four antibodies were specific for GPIIb/IIIa; one acetaminophen sulfate-dependent antibody reacted mainly with GPIb/IX/V.³⁰ The Wisconsin reference laboratory confirms that for paracetamol and naproxen, drug-dependent antibodies have been found only against metabolites and do not react with the unmodified drug, and that testing for these drugs, and for ibuprofen, required metabolites.²³ Both drugs are graded probable in the laboratory review, on the evidence of one laboratory.³

If paracetamol or naproxen is suspected, a negative test against the parent drug does not exclude it; the laboratory needs to be asked whether it tests with metabolites. This is an inference from the laboratory studies, which do not report how often a test with the parent drug misses these cases.

Ibuprofen is better established. It is one of the 16 drugs with a definite laboratory diagnosis, confirmed in two laboratories, one of which added an ibuprofen metabolite and the other urine collected after ingestion, which contains metabolites.³ Some patients have antibodies that react with unmodified ibuprofen, and others have antibodies that need a metabolite.²³ An expert review lists ibuprofen among the drugs most often implicated.³⁷ No source read for this review gives its typical onset or nadir.

Gastrointestinal drugs

Among gastrointestinal drugs, only ranitidine has a laboratory-confirmed drug-dependent antibody. The evidence for proton pump inhibitors is weak: a systematic search of the world literature to August 2024 found 18 cases. Mesalazine rests on a single case that recurred on rechallenge.

Ranitidine is the instructive drug in this group. A patient with acute, severe thrombocytopenia after ranitidine had an antibody that reacted with normal platelets only in the presence of the drug at therapeutic concentrations. It was specific for the GPIb/IX complex, and the epitope lay on GPIX at or near the site that is also the common target of quinine- and quinidine-induced antibodies, drugs unrelated to ranitidine in structure. It was the first drug-dependent antibody identified for an H2-receptor antagonist.¹¹⁹ This is a single case with detailed serology. Ranitidine is graded probable in the laboratory review and is among the 24 drugs with both clinical and antibody evidence.^3,23

For proton pump inhibitors, a systematic analysis of case reports, searching six databases including three Chinese ones to August 2024, found 16 publications describing 18 patients. Their median age was 62 years (range 23–98). The drugs were pantoprazole in 11, lansoprazole in 4, omeprazole in 2 and esomeprazole in 1. The count fell a median of 3 days after starting (range 2–7), and after withdrawal 11 patients recovered and 7 improved.¹²⁰ Eighteen cases against the very wide use of these drugs is weak evidence. It does not show that proton pump inhibitors are a common cause, and none has a laboratory-confirmed antibody.

Mesalazine rests on one case: a 25-year-old with ulcerative colitis developed marked thrombocytopenia on treatment and again on rechallenge. The authors noted that mesalazine, unlike sulfasalazine, had rarely been implicated in serious blood disorders.¹²¹ Rechallenge is the strongest clinical evidence short of an antibody, but it is still one patient. A UK analysis of spontaneous reports against 4·7 million sulfasalazine and 2·8 million mesalazine prescriptions (1991–1998) found no safety advantage of mesalazine over sulfasalazine across the disorders studied; pancreatitis and interstitial nephritis were significantly more common with mesalazine. It reported blood dyscrasias together and gave no rate for thrombocytopenia alone.¹²²

Neurology and psychiatry

The anticonvulsants cause thrombocytopenia in two different ways. Valproate lowers the count as the drug level rises. In the studies this was common and mostly mild. Carbamazepine and phenytoin cause the classic immune reaction, with drug-dependent antibodies, and as with any immune reaction the treatment is to stop the drug. Outside valproate the evidence is laboratory series and case reports. Among psychiatric drugs, mirtazapine has definite laboratory evidence and haloperidol probable evidence.^3,23

Valproate

Valproate is the best-studied drug in this section, and the evidence shows a dose–response. In a randomised, double-blind trial of divalproex monotherapy in 143 patients, one group was kept at a high level (80–150 µg/mL) and one at a low level (25–50 µg/mL). Thrombocytopenia was among the adverse events that were significantly more frequent in the high-level group.¹²³ A prospective analysis of the same trial programme covered 851 levels in 265 patients. In all, 17·7% had at least one count of 100 × 10⁹/L or below. The count fell as the level rose, and women were at higher risk. The chance of thrombocytopenia rose steeply at trough levels above 100 µg/mL in women and above 130 µg/mL in men. A lower baseline count was a further risk factor.¹⁰ In critically ill patients the free level cannot be predicted from the total level. A later study of 844 trough free levels in 264 patients found that a free level of 16·65 µg/mL best separated patients with and without a count of 100 × 10⁹/L or below.¹²⁴ These thresholds come from one research programme.

Children show the same pattern. In a retrospective case–control study, 21·6% of 167 children on valproate had a count below 200 × 10⁹/L, against 5·4% of 92 controls. The serum level was the strongest predictor; the dose was not. No child bled.¹²⁵ The threshold of 200 × 10⁹/L is unusually high, so this figure cannot be compared directly with the adult studies.

Valproate is also among the 24 drugs with both case reports and drug-dependent antibodies in the Reese analysis, although Arnold's method did not capture it.^3,23 The usual effect follows the level, but an immune reaction appears to occur in some patients. The sources read do not say how to tell the two apart in a single patient. Valproate was one of the two most often reported causes in the Danish adverse drug reaction register up to 1991.¹¹⁸

Carbamazepine, phenytoin and the newer anticonvulsants

Carbamazepine has the strongest immune evidence of any anticonvulsant. It is one of the 16 drugs with a definite laboratory diagnosis, with drug-dependent antibodies shown in two laboratories,³ and in the Reese analysis it has 10 or more case reports and antibodies in 10 or more patients.²³ An expert review lists it among the drugs most often implicated.³⁷ Phenytoin has probable laboratory evidence from one laboratory, and antibodies have been found in 10 or more patients.^3,23 No onset or nadir figures specific to either drug were found.

A small cross-sectional study of 137 patients on carbamazepine and 60 on oxcarbazepine found a count below 150 × 10⁹/L in 2·9% and 1·7%. The count fell with the daily dose but was not related to the serum level.¹²⁶ The study measured how many patients had a mild fall at one time, not the acute immune reaction. For oxcarbazepine, the rest of the evidence is case reports. For levetiracetam, the evidence read is a single case: the count recovered within a few days of stopping the drug, after valproate had already been stopped.¹²⁷

Psychiatric drugs

Mirtazapine is one of the 16 drugs with a definite laboratory diagnosis.³ The first well-documented case, in 2003, was a 66-year-old with severe thrombocytopenia and a drug-dependent antibody against GPIIb/IIIa.¹²⁸ Haloperidol has probable laboratory evidence and is among Reese's 24 drugs; no clinical detail was found for it.^3,23 Clozapine has a single case report: onset at 17 weeks, and a count that stayed low for 24 weeks despite a lower dose and recovered only when the drug was stopped.¹²⁹

For the selective serotonin and serotonin–noradrenaline reuptake inhibitors, the only cohort found is Japanese. It compared 4759 patients on an SSRI and 3440 on an SNRI with 2196 on paroxetine. The adjusted hazard ratios were 1·14 (95% CI 0·76–1·70) and 0·77 (0·48–1·21), neither significant. Paroxetine was itself thought to carry a risk, so the study had no unexposed control. The authors note that the sertraline package insert in Japan has been revised.¹³⁰

Oncology and haematology

A low count in a patient on cancer treatment usually reflects marrow suppression, but not always. Five mechanisms need to be kept apart: dose-dependent marrow suppression by chemotherapy; drug-dependent antibodies, of which oxaliplatin is the clearest example; drug-induced thrombotic microangiopathy with gemcitabine and mitomycin C; autoimmune thrombocytopenia triggered by immune checkpoint inhibitors; and a reversible effect on platelet release with bortezomib. The evidence here is stronger than elsewhere in this review, because thrombocytopenia is common enough to count in trials and large cohorts. Rituximab is discussed under rheumatology, and the thrombotic microangiopathies of gemcitabine and mitomycin C under drug-induced thrombotic microangiopathy.

Chemotherapy-induced thrombocytopenia

How often chemotherapy lowers the count depends mostly on how thrombocytopenia is defined. In a US claims cohort of 215 508 patients, episodes were identified from diagnosis, procedure and drug codes. Thrombocytopenia occurred in 9·7% (95% CI 9·6–9·8) over the course of treatment, and in 2·7–2·9% per cycle. By drug, the rate ranged from 6·1% with cyclophosphamide to 13·2% with carboplatin and 13·5% with gemcitabine, with oxaliplatin at 11·4%. About a third of episodes were managed in hospital. The authors note that coding captures the more severe cases.¹³¹ In an outpatient database of 47 159 patients, where any count below 150 × 10⁹/L counted, the prevalence during chemotherapy ranged from 21·9% with taxane-based regimens to 64·2% with gemcitabine-based regimens, and 11·1% were already below that level before treatment.¹³² The authors of the claims study point out that a threshold of 75 × 10⁹/L applied to those data gives 13·8%, so most of the gap is a matter of definition.¹³¹ A smaller study that counted only clinically significant falls (below 75 × 10⁹/L with bleeding, a dose delay or reduction, transfusion or stopping treatment) found 10·1%. The rate was highest with cisplatin and gemcitabine for bladder cancer (57%), on few patients.¹³³ Gemcitabine regimens carry the highest rates in all three studies.

Oxaliplatin

Oxaliplatin causes a true drug-dependent immune thrombocytopenia. It is one of the 16 drugs with a definite laboratory diagnosis, with antibodies against GPIIb/IIIa shown in two laboratories,³ and it is among Reese's 24 drugs.²³ The timing differs from the usual pattern of a first exposure. Reported cases appeared after many cycles: after the 18th and 20th cycles in two index patients, and at cycles 10, 17 and 28 in another series.^134,135 The fall can be sudden and profound. In one case the count reached zero 7·5 hours after the dose.¹³⁵ Among 30 patients with oxaliplatin-dependent antibodies, 7 had documented acute thrombocytopenia after 10 or more cycles, with a median nadir of 5 × 10⁹/L.¹³⁴

Two features make oxaliplatin harder to manage than most drugs. First, these patients often make antibodies to the other drugs given with it. Of the 30 patients with oxaliplatin-dependent antibodies, 14 (47%) had 34 further drug-dependent antibodies, to leucovorin, dexamethasone, ondansetron, palonosetron, irinotecan and diphenhydramine. Only 3 such antibodies were found in 23 patients without oxaliplatin antibodies. The two index patients had further falls when oxaliplatin was left out, and absorption studies showed separate antibodies rather than one cross-reacting antibody.¹³⁴ Stopping oxaliplatin alone may therefore not end the problem. Second, the antibodies can target other cells. A 79-year-old developed thrombocytopenia, haemolysis and neutropenia with the 15th and 17th cycles, but not with the 16th cycle, which left out oxaliplatin; oxaliplatin-dependent antibodies to platelets, red cells and neutrophils were found.¹³⁶ The evidence is case reports and one serological series.

Immune checkpoint inhibitors

Checkpoint inhibitors can trigger an autoimmune thrombocytopenia that resembles primary immune thrombocytopenia. The first large incidence estimate comes from a 2026 multicentre cohort of 86 467 adults starting a checkpoint inhibitor at 29 US hospitals. Immune thrombocytopenia occurred in 214 (0·25%). The median onset was 8 weeks after starting treatment (IQR 4–18) and the median nadir was 41 × 10⁹/L (IQR 17–64). About half received glucocorticoids, 18% immune globulin and 14% a thrombopoietin receptor agonist. Three-quarters recovered, at a median of 2·3 weeks. Of 76 patients rechallenged, 23 (30%) had a recurrence. The risk was higher with a lower baseline count, combination checkpoint therapy, stage IV disease and other immune-related adverse events. Severe thrombocytopenia carried an adjusted hazard ratio for death from any cause of 2·96 (95% CI 2·14–4·08).¹³⁷ This is a retrospective cohort, and the hazard ratio shows an association, not that the thrombocytopenia caused the deaths.

Earlier data could not give an incidence because they had no denominator. In the French pharmacovigilance database, immune thrombocytopenia was the commonest checkpoint-related cytopenia (51% of 75 events). Across all these cytopenias the median onset was 2 months, nearly half were grade 4 or higher, and 3 of 10 patients who restarted treatment relapsed.¹³⁸ A disproportionality analysis of the US adverse event reporting system found reporting signals for all five PD-1 and PD-L1 inhibitors studied; these are reporting signals, not risks.¹³⁹ The mechanism is uncertain, so management borrows from primary immune thrombocytopenia.¹⁴⁰

Bortezomib

Bortezomib lowers the count in a regular cycle. In 228 patients in two phase 2 trials, the mean count fell by about 60% during each cycle and recovered between cycles. In patients who responded, the count before each cycle rose over time. The fall did not depend on the baseline count, the paraprotein level or the marrow plasma cell burden.¹² In the phase 3 APEX trial, grade 3 and grade 4 thrombocytopenia occurred in 26% and 4% with bortezomib, against 5% and 1% with high-dose dexamethasone, but significant bleeding was similar (4% and 5%).¹⁴¹ These rates are specific to APEX. No rate was verified for the weekly and subcutaneous schedules in use now, so they should not be read as current rates.

The mechanism is not marrow destruction. In mice, a single dose lowered the count without harming megakaryocyte number, ploidy or shape, and the trial authors concluded that the drug has a reversible effect on megakaryocyte function.¹² Two independent laboratory studies showed that proteasome inhibition blocks proplatelet formation, the step in which megakaryocytes release platelets, through the RhoA pathway, and that blocking Rho kinase reverses the effect.^142,143

Lenalidomide, PARP inhibitors and ibrutinib

For lenalidomide, the figure verified here comes from myelodysplastic syndrome with deletion 5q. In that phase 2 trial of 148 patients, thrombocytopenia occurred in 44% and, with neutropenia, was the commonest reason to interrupt treatment or lower the dose.¹⁴⁴ No figure was verified for lenalidomide in myeloma, where the regimens and the patients differ.

Among PARP inhibitors, niraparib has the clearest data. In the phase 3 PRIMA trial of 733 patients, grade 3 or higher thrombocytopenia occurred in 28·7%.¹⁴⁵ In the earlier NOVA trial, 3·3% of patients stopped treatment because of thrombocytopenia, and 68·9% needed a dose reduction. A baseline body weight below 77 kg or a baseline count below 150 × 10⁹/L predicted grade 3 or higher thrombocytopenia, and the authors suggest a starting dose of 200 mg a day for these patients.¹⁴⁶ This is a retrospective analysis of trial data.

Ibrutinib is included in the table for a different reason. It causes a platelet function defect rather than a low count. In patients with non-Hodgkin lymphoma, both ibrutinib and acalabrutinib impaired aggregation to collagen, but only ibrutinib impaired thrombus formation under arterial flow.¹⁴⁷ It belongs in the differential of bleeding with a normal count, not of thrombocytopenia.

Other drugs and vaccines

This section collects drugs and vaccines that do not fit a single system. The measles–mumps–rubella (MMR) vaccine has the best data, with an absolute risk per dose from record linkage. Iodinated contrast has recently gained its first laboratory confirmation. Interferon has a cohort that shows how much the baseline count matters. For propylthiouracil, no evidence of isolated thrombocytopenia was verified.

Vaccines

The link between MMR vaccine and immune thrombocytopenia is causal. A UK study linked immunisation records to hospital admissions. The absolute risk within six weeks of immunisation was 1 in 22 300 doses, and two of every three cases in that six-week window were caused by the vaccine. Children who had had immune thrombocytopenia before MMR had no vaccine-associated recurrence.¹⁶ A US cohort of 1·8 million children and adolescents, with chart-confirmed cases, found no raised risk after any early-childhood vaccine other than MMR at 12–19 months. It did find raised risks for hepatitis A vaccine at 7–17 years and for varicella and Tdap at 11–17 years, but each rested on one or two vaccinated cases. Most cases were acute and mild, without long-term effects.¹⁵⁰ In the Danish register up to 1991, MMR vaccine was one of the two most often reported causes of thrombocytopenia.¹¹⁸

Vaccine-induced immune thrombotic thrombocytopenia (VITT) is a different disorder. Like heparin-induced thrombocytopenia, it is caused by antibodies to platelet factor 4 (PF4) that activate platelets, and the danger is thrombosis. It followed adenoviral-vector COVID-19 vaccines. The index series from Germany and Austria described 11 patients after ChAdOx1 nCoV-19, with onset 5–16 days after vaccination, cerebral venous thrombosis in 9, and no heparin exposure. The antibodies activated platelets in the presence of PF4 without heparin.¹⁵⁵ A Norwegian series of 5 patients, 7–10 days after a first dose, had high levels of anti-PF4 antibodies.¹⁵⁶ In the UK cohort of 220 definite or probable cases, presentation was 5–48 days after a first dose, and overall mortality was 22%. Cerebral venous sinus thrombosis and a lower platelet count predicted death.¹⁵⁷ A similar picture followed the Ad26.COV2.S vaccine in the United States.¹⁵⁸ The antibodies in VITT bind PF4 directly, without heparin, and a near-identical disorder can follow natural infection, especially with adenovirus. HIT and VITT antibodies need different assays.^56,159 VITT is not caused by a drug, but it clarified how the anti-PF4 disorders differ from one another.

Contrast media, interferon and antithyroid drugs

A 2025 case report described an older man with a third episode of sudden, severe, self-limiting thrombocytopenia after radiological examinations. Drug-dependent antibodies against GPIIb/IIIa were found in the presence of iohexol. The authors state that contrast-dependent platelet antibodies had not been detected before, and describe this as the first laboratory confirmation.¹⁵¹ It is a single case.

In 979 veterans with hepatitis C treated with pegylated interferon and ribavirin, a count below 50 × 10⁹/L occurred in 6·1%. The baseline count mattered most: the rate was 41·1% when the count before treatment was 50–100 × 10⁹/L, 11·7% at 100–150 × 10⁹/L and 0·55% above 150 × 10⁹/L. Most episodes occurred within the first 12 weeks, and the median nadir was 35 × 10⁹/L. Major bleeding was rare, with 5 events (0·51%) over 48 weeks.¹⁵² The study was a retrospective cohort.

For propylthiouracil, no study of isolated thrombocytopenia was verified, and it appears in neither the Reese nor the Arnold lists. The largest antithyroid cohort, of 50 385 patients with Graves' disease in Tokyo, found 55 with blood damage: 50 with agranulocytosis and 5 with pancytopenia. Of the 55, 4 had taken propylthiouracil and the rest methimazole. Anaemia or thrombocytopenia accompanied agranulocytosis in 7 of the 50.¹⁵³ It gives no rate for thrombocytopenia alone.

In a 1991 report, four men who injected cocaine developed thrombocytopenia with mucosal bleeding, related in time to use. One had two separate episodes. Three responded to intravenous immune globulin and steroids, and one needed a splenectomy. No antibody testing was reported.¹⁵⁴

Uncertainties and future research

Most drug-specific evidence is case reports. For most drugs, the link rests on published case reports judged by the George criteria. The fourth criterion, re-exposure, is rarely met because it is often unsafe. Counts of published reports reflect reporting, not risk, and cannot say how often a drug causes the reaction.^5,22 Population incidence estimates are few and old; the only case–control study giving relative risks for individual drugs was published in 1993.^1,17

A negative antibody test does not exclude the diagnosis. The test for drug-dependent antibodies is specific but insensitive, is not standardised, and is done in only a few reference laboratories.^3,6 It can be falsely negative when the antibody reacts with a metabolite rather than the parent drug, when the drug dissolves poorly in water, as sulfamethoxazole and trimethoprim do, or when the sample is taken too late.^6,30 In one patient whose count fell from 187 to 5 × 10⁹/L after a single dose of levofloxacin, no antibody was detected.⁴⁵ No study gives the proportion of patients with suspected drug-induced thrombocytopenia in whom an antibody is eventually found.

No treatment other than stopping the drug has been tested. No randomised trial of any treatment for immune drug-induced thrombocytopenia was found. Whether corticosteroids help is not established, and the use of platelet transfusion and immune globulin rests on case reports and expert opinion.^2,5,13

Bedside features that separate drug-induced thrombocytopenia from sepsis or disseminated intravascular coagulation were not found in the sources used for this review. A study of this question in critically ill patients would help.

Gaps for individual drugs. No onset or nadir figures were found for carbamazepine, one of the drugs with the strongest laboratory evidence. The bortezomib rates come from one trial schedule, and the lenalidomide rate from myelodysplastic syndrome, not myeloma. The mechanism of checkpoint-inhibitor thrombocytopenia is unknown. The full drug-by-drug results of the Reese analysis are in a supplement that could not be retrieved, so the absence of a drug from the tables here is not evidence that it is safe.²³

Few regional data. The only Saudi studies found were a single-centre cohort comparing linezolid with vancomycin and a report of two cases after piperacillin–tazobactam, both from Riyadh.^88,97 No population data were found from Saudi Arabia or the Gulf on the incidence of drug-induced thrombocytopenia or of heparin-induced thrombocytopenia, or on access to drug-dependent antibody testing in the region. Local incidence studies, and a regional route for referring samples for antibody testing, would both be useful.

Conclusion

Drug-induced thrombocytopenia is uncommon for any single drug, but in hospital it is easy to miss. The first clue is timing. A severe fall with bleeding 5–10 days after a new drug, or within hours of re-exposure or of a glycoprotein IIb/IIIa inhibitor, points to a drug-dependent antibody.⁴ A moderate fall with thrombosis on heparin points to heparin-induced thrombocytopenia, and a low 4Ts score rules it out.^7,8 Some patterns are not immune and have their own timing: valproate follows the drug level,¹⁰ bortezomib falls and recovers with each cycle,¹² and oxaliplatin and checkpoint inhibitors often appear only after weeks or many cycles.^134,137

Once a drug is suspected, stop it. When several drugs are candidates, stop all those started in the previous 2 weeks and substitute where needed.² The count usually begins to recover within 1–2 days.^4,5 Send a sample for drug-dependent antibody testing where it can be done, during the acute episode, and remember that a negative result does not exclude the diagnosis.⁶ Record the drug clearly in the patient's notes. A patient with a confirmed reaction should avoid the drug for life, and the reaction should be reported to the adverse event reporting system.^4,5

References

1. van den Bemt PM, Meyboom RH, Egberts AC. Drug-induced immune thrombocytopenia. Drug Saf 2004; 27: 1243–52. doi:10.2165/00002018-200427150-00007 PMID 15588119
2. Bakchoul T, Marini I. Drug-associated thrombocytopenia. Hematology Am Soc Hematol Educ Program 2018; 2018: 576–83. doi:10.1182/asheducation-2018.1.576 PMID 30504360
3. Arnold DM, Kukaswadia S, Nazi I, et al. A systematic evaluation of laboratory testing for drug-induced immune thrombocytopenia. J Thromb Haemost 2013; 11: 169–76. doi:10.1111/jth.12052 PMID 23121994
4. Arnold DM, Nazi I, Warkentin TE, et al. Approach to the diagnosis and management of drug-induced immune thrombocytopenia. Transfus Med Rev 2013; 27: 137–45. doi:10.1016/j.tmrv.2013.05.005 PMID 23845922
5. George JN, Aster RH. Drug-induced thrombocytopenia: pathogenesis, evaluation, and management. Hematology Am Soc Hematol Educ Program 2009; 2009: 153–8. doi:10.1182/asheducation-2009.1.153 PMID 20008194
6. Arnold DM, Curtis BR, Bakchoul T; Platelet Immunology Scientific Subcommittee of the ISTH. Recommendations for standardization of laboratory testing for drug-induced immune thrombocytopenia: communication from the SSC of the ISTH. J Thromb Haemost 2015; 13: 676–8. doi:10.1111/jth.12852 PMID 25604471
7. Cuker A, Gimotty PA, Crowther MA, Warkentin TE. Predictive value of the 4Ts scoring system for heparin-induced thrombocytopenia: a systematic review and meta-analysis. Blood 2012; 120: 4160–7. doi:10.1182/blood-2012-07-443051 PMID 22990018
8. Cuker A, Arepally GM, Chong BH, et al. American Society of Hematology 2018 guidelines for management of venous thromboembolism: heparin-induced thrombocytopenia. Blood Adv 2018; 2: 3360–92. doi:10.1182/bloodadvances.2018024489 PMID 30482768
9. Crass RL, Cojutti PG, Pai MP, Pea F. Reappraisal of linezolid dosing in renal impairment to improve safety. Antimicrob Agents Chemother 2019; 63: e00605-19. doi:10.1128/AAC.00605-19 PMID 31109977
10. Nasreddine W, Beydoun A. Valproate-induced thrombocytopenia: a prospective monotherapy study. Epilepsia 2008; 49: 438–45. doi:10.1111/j.1528-1167.2007.01429.x PMID 18031547
11. Kivity S, Zafrir Y, Loebstein R, Pauzner R, Mouallem M, Mayan H. Clinical characteristics and risk factors for low dose methotrexate toxicity: a cohort of 28 patients. Autoimmun Rev 2014; 13: 1109–13. doi:10.1016/j.autrev.2014.08.027 PMID 25172240
12. Lonial S, Waller EK, Richardson PG, et al. Risk factors and kinetics of thrombocytopenia associated with bortezomib for relapsed, refractory multiple myeloma. Blood 2005; 106: 3777–84. doi:10.1182/blood-2005-03-1173 PMID 16099887
13. Aster RH, Curtis BR, McFarland JG, Bougie DW. Drug-induced immune thrombocytopenia: pathogenesis, diagnosis, and management. J Thromb Haemost 2009; 7: 911–8. doi:10.1111/j.1538-7836.2009.03360.x PMID 19344362
14. Sane DC, Damaraju LV, Topol EJ, et al. Occurrence and clinical significance of pseudothrombocytopenia during abciximab therapy. J Am Coll Cardiol 2000; 36: 75–83. doi:10.1016/s0735-1097(00)00688-4 PMID 10898416
15. Reese JA, Bougie DW, Curtis BR, et al. Drug-induced thrombotic microangiopathy: experience of the Oklahoma Registry and the BloodCenter of Wisconsin. Am J Hematol 2015; 90: 406–10. doi:10.1002/ajh.23960 PMID 25639727
16. Miller E, Waight P, Farrington CP, Andrews N, Stowe J, Taylor B. Idiopathic thrombocytopenic purpura and MMR vaccine. Arch Dis Child 2001; 84: 227–9. doi:10.1136/adc.84.3.227 PMID 11207170
17. Kaufman DW, Kelly JP, Johannes CB, et al. Acute thrombocytopenic purpura in relation to the use of drugs. Blood 1993; 82: 2714–8. PMID 8219224
18. Garbe E, Andersohn F, Bronder E, et al. Drug-induced immune thrombocytopaenia: results from the Berlin Case-Control Surveillance Study. Eur J Clin Pharmacol 2012; 68: 821–32. doi:10.1007/s00228-011-1184-3 PMID 22187020
19. Andersohn F, Bronder E, Klimpel A, Garbe E. Proportion of drug-related serious rare blood dyscrasias: estimates from the Berlin Case-Control Surveillance Study. Am J Hematol 2004; 77: 316–8. doi:10.1002/ajh.20176 PMID 15495238
20. ten Berg MJ, Huisman A, Souverein PC, et al. Drug-induced thrombocytopenia: a population study. Drug Saf 2006; 29: 713–21. doi:10.2165/00002018-200629080-00008 PMID 16872245
21. Reese JA, Nguyen LP, Buchanan GR, et al. Drug-induced thrombocytopenia in children. Pediatr Blood Cancer 2013; 60: 1975–81. doi:10.1002/pbc.24682 PMID 24038783
22. George JN, Raskob GE, Shah SR, et al. Drug-induced thrombocytopenia: a systematic review of published case reports. Ann Intern Med 1998; 129: 886–90. doi:10.7326/0003-4819-129-11_part_1-199812010-00009 PMID 9867731
23. Reese JA, Li X, Hauben M, et al. Identifying drugs that cause acute thrombocytopenia: an analysis using 3 distinct methods. Blood 2010; 116: 2127–33. doi:10.1182/blood-2010-03-276691 PMID 20530792
24. Bougie DW, Wilker PR, Wuitschick ED, et al. Acute thrombocytopenia after treatment with tirofiban or eptifibatide is associated with antibodies specific for ligand-occupied GPIIb/IIIa. Blood 2002; 100: 2071–6. PMID 12200368
25. Bougie DW, Rasmussen M, Zhu J, Aster RH. Antibodies causing thrombocytopenia in patients treated with RGD-mimetic platelet inhibitors recognize ligand-specific conformers of αIIb/β3 integrin. Blood 2012; 119: 6317–25. doi:10.1182/blood-2012-01-406322 PMID 22490676
26. Curtis BR, Swyers J, Divgi A, McFarland JG, Aster RH. Thrombocytopenia after second exposure to abciximab is caused by antibodies that recognize abciximab-coated platelets. Blood 2002; 99: 2054–9. doi:10.1182/blood.v99.6.2054 PMID 11877279
27. Curtis BR, Divgi A, Garritty M, Aster RH. Delayed thrombocytopenia after treatment with abciximab: a distinct clinical entity associated with the immune response to the drug. J Thromb Haemost 2004; 2: 985–92. doi:10.1111/j.1538-7836.2004.00744.x PMID 15140135
28. Garner SF, Campbell K, Metcalfe P, et al. Glycoprotein V: the predominant target antigen in gold-induced autoimmune thrombocytopenia. Blood 2002; 100: 344–6. doi:10.1182/blood.v100.1.344 PMID 12070047
29. Bougie DW, Wilker PR, Aster RH. Patients with quinine-induced immune thrombocytopenia have both “drug-dependent” and “drug-specific” antibodies. Blood 2006; 108: 922–7. doi:10.1182/blood-2006-01-009803 PMID 16861345
30. Bougie D, Aster R. Immune thrombocytopenia resulting from sensitivity to metabolites of naproxen and acetaminophen. Blood 2001; 97: 3846–50. doi:10.1182/blood.v97.12.3846 PMID 11389025
31. Page EE, Little DJ, Vesely SK, George JN. Quinine-induced thrombotic microangiopathy: a report of 19 patients. Am J Kidney Dis 2017; 70: 686–95. doi:10.1053/j.ajkd.2017.05.023 PMID 28780041
32. Al-Nouri ZL, Reese JA, Terrell DR, Vesely SK, George JN. Drug-induced thrombotic microangiopathy: a systematic review of published reports. Blood 2015; 125: 616–8. doi:10.1182/blood-2014-11-611335 PMID 25414441
33. Kojouri K, Vesely SK, George JN. Quinine-associated thrombotic thrombocytopenic purpura-hemolytic uremic syndrome: frequency, clinical features, and long-term outcomes. Ann Intern Med 2001; 135: 1047–51. doi:10.7326/0003-4819-135-12-200112180-00008 PMID 11747383
34. Liles NW, Page EE, Liles AL, Vesely SK, Raskob GE, George JN. Diversity and severity of adverse reactions to quinine: a systematic review. Am J Hematol 2016; 91: 461–6. doi:10.1002/ajh.24314 PMID 26822544
35. Jacob S, Dunn BL, Qureshi ZP, et al. Ticlopidine-, clopidogrel-, and prasugrel-associated thrombotic thrombocytopenic purpura: a 20-year review from the Southern Network on Adverse Reactions (SONAR). Semin Thromb Hemost 2012; 38: 845–53. doi:10.1055/s-0032-1328894 PMID 23111862
36. Bennett CL, Kim B, Zakarija A, et al. Two mechanistic pathways for thienopyridine-associated thrombotic thrombocytopenic purpura: a report from the SERF-TTP Research Group and the RADAR Project. J Am Coll Cardiol 2007; 50: 1138–43. doi:10.1016/j.jacc.2007.04.093 PMID 17868804
37. Curtis BR. Drug-induced immune thrombocytopenia: incidence, clinical features, laboratory testing, and pathogenic mechanisms. Immunohematology 2014; 30: 55–65. PMID 25247620
38. Von Drygalski A, Curtis BR, Bougie DW, et al. Vancomycin-induced immune thrombocytopenia. N Engl J Med 2007; 356: 904–10. doi:10.1056/NEJMoa065066 PMID 17329697
39. Dery JP, Braden GA, Lincoff AM, et al. Final results of the ReoPro readministration registry. Am J Cardiol 2004; 93: 979–84. doi:10.1016/j.amjcard.2003.12.051 PMID 15081439
40. Jacquot C, Moayeri M, Kim B, Shugarts S, Lynch KL, Leavitt AD. Prolonged ceftriaxone-induced immune thrombocytopenia due to impaired drug clearance: a case report. Transfusion 2013; 53: 2715–21. doi:10.1111/trf.12138 PMID 23451879
41. Priziola JL, Smythe MA, Dager WE. Drug-induced thrombocytopenia in critically ill patients. Crit Care Med 2010; 38 (6 suppl): S145–54. doi:10.1097/CCM.0b013e3181de0b88 PMID 20502168
42. Hui P, Cook DJ, Lim W, Fraser GA, Arnold DM. The frequency and clinical significance of thrombocytopenia complicating critical illness: a systematic review. Chest 2011; 139: 271–8. doi:10.1378/chest.10-2243 PMID 21071526
43. Greinacher A, Selleng K. Thrombocytopenia in the intensive care unit patient. Hematology Am Soc Hematol Educ Program 2010; 2010: 135–43. doi:10.1182/asheducation-2010.1.135 PMID 21239783
44. Thiele T, Selleng K, Selleng S, Greinacher A, Bakchoul T. Thrombocytopenia in the intensive care unit—diagnostic approach and management. Semin Hematol 2013; 50: 239–50. doi:10.1053/j.seminhematol.2013.06.008 PMID 23953341
45. Shih AW, Lam AS, Warkentin TE. Levofloxacin-induced acute immune-mediated thrombocytopenia of rapid-onset. J Pharm Pract 2018; 31: 234–7. doi:10.1177/0897190017702306 PMID 28403679
46. Bennett CL, Weinberg PD, Rozenberg-Ben-Dror K, Yarnold PR, Kwaan HC, Green D. Thrombotic thrombocytopenic purpura associated with ticlopidine: a review of 60 cases. Ann Intern Med 1998; 128: 541–4. doi:10.7326/0003-4819-128-7-199804010-00004 PMID 9518398
47. Pishko AM, Frere C, Maier CL, Levy JH, Connors JM. Management of anticoagulation in patients with heparin-induced thrombocytopenia requiring cardiac surgery with cardiopulmonary bypass: guidance from the ISTH SSC on Perioperative and Critical Care Haemostasis and Thrombosis. J Thromb Haemost 2026; 24: 3369–91. doi:10.1016/j.jtha.2026.05.029 PMID 42248416
48. Martel N, Lee J, Wells PS. Risk for heparin-induced thrombocytopenia with unfractionated and low-molecular-weight heparin thromboprophylaxis: a meta-analysis. Blood 2005; 106: 2710–5. doi:10.1182/blood-2005-04-1546 PMID 15985543
49. Warkentin TE, Levine MN, Hirsh J, et al. Heparin-induced thrombocytopenia in patients treated with low-molecular-weight heparin or unfractionated heparin. N Engl J Med 1995; 332: 1330–5. doi:10.1056/NEJM199505183322003 PMID 7715641
50. PROTECT Investigators for the Canadian Critical Care Trials Group and the Australian and New Zealand Intensive Care Society Clinical Trials Group; Cook D, Meade M, Guyatt G, et al. Dalteparin versus unfractionated heparin in critically ill patients. N Engl J Med 2011; 364: 1305–14. doi:10.1056/NEJMoa1014475 PMID 21417952
51. Warkentin TE, Sheppard JA, Horsewood P, Simpson PJ, Moore JC, Kelton JG. Impact of the patient population on the risk for heparin-induced thrombocytopenia. Blood 2000; 96: 1703–8. PMID 10961867
52. Warkentin TE, Kelton JG. Temporal aspects of heparin-induced thrombocytopenia. N Engl J Med 2001; 344: 1286–92. doi:10.1056/NEJM200104263441704 PMID 11320387
53. Warkentin TE, Kelton JG. Delayed-onset heparin-induced thrombocytopenia and thrombosis. Ann Intern Med 2001; 135: 502–6. doi:10.7326/0003-4819-135-7-200110020-00009 PMID 11578153
54. Greinacher A, Selleng K, Warkentin TE. Autoimmune heparin-induced thrombocytopenia. J Thromb Haemost 2017; 15: 2099–114. doi:10.1111/jth.13813 PMID 28846826
55. Warkentin TE, Greinacher A. Spontaneous HIT syndrome: knee replacement, infection, and parallels with vaccine-induced immune thrombotic thrombocytopenia. Thromb Res 2021; 204: 40–51. doi:10.1016/j.thromres.2021.05.018 PMID 34144250
56. Greinacher A, Warkentin TE. Thrombotic anti-PF4 immune disorders: HIT, VITT, and beyond. Hematology Am Soc Hematol Educ Program 2023; 2023: 1–10. doi:10.1182/hematology.2023000503 PMID 38066843
57. Warkentin TE, Kelton JG. A 14-year study of heparin-induced thrombocytopenia. Am J Med 1996; 101: 502–7. doi:10.1016/s0002-9343(96)00258-6 PMID 8948273
58. Warkentin TE, Elavathil LJ, Hayward CP, Johnston MA, Russett JI, Kelton JG. The pathogenesis of venous limb gangrene associated with heparin-induced thrombocytopenia. Ann Intern Med 1997; 127: 804–12. doi:10.7326/0003-4819-127-9-199711010-00005 PMID 9382401
59. Lo GK, Juhl D, Warkentin TE, Sigouin CS, Eichler P, Greinacher A. Evaluation of pretest clinical score (4 T's) for the diagnosis of heparin-induced thrombocytopenia in two clinical settings. J Thromb Haemost 2006; 4: 759–65. doi:10.1111/j.1538-7836.2006.01787.x PMID 16634744
60. Nagler M, Bachmann LM, ten Cate H, ten Cate-Hoek A. Diagnostic value of immunoassays for heparin-induced thrombocytopenia: a systematic review and meta-analysis. Blood 2016; 127: 546–57. doi:10.1182/blood-2015-07-661215 PMID 26518436
61. Sheridan D, Carter C, Kelton JG. A diagnostic test for heparin-induced thrombocytopenia. Blood 1986; 67: 27–30. PMID 3940551
62. Greinacher A, Michels I, Kiefel V, Mueller-Eckhardt C. A rapid and sensitive test for diagnosing heparin-associated thrombocytopenia. Thromb Haemost 1991; 66: 734–6. PMID 1796420
63. Warkentin TE, Arnold DM, Nazi I, Kelton JG. The platelet serotonin-release assay. Am J Hematol 2015; 90: 564–72. doi:10.1002/ajh.24006 PMID 25775976
64. Warkentin TE, Nazy I, Sheppard JI, Smith JW, Kelton JG, Arnold DM. Serotonin-release assay-negative heparin-induced thrombocytopenia. Am J Hematol 2020; 95: 38–47. doi:10.1002/ajh.25660 PMID 31621093
65. Morgan RL, Ashoorion V, Cuker A, et al. Management of heparin-induced thrombocytopenia: systematic reviews and meta-analyses. Blood Adv 2020; 4: 5184–93. doi:10.1182/bloodadvances.2020002963 PMID 33095876
66. Goel R, Ness PM, Takemoto CM, Krishnamurti L, King KE, Tobian AA. Platelet transfusions in platelet consumptive disorders are associated with arterial thrombosis and in-hospital mortality. Blood 2015; 125: 1470–6. doi:10.1182/blood-2014-10-605493 PMID 25588677
67. Linkins LA, Dans AL, Moores LK, et al. Treatment and prevention of heparin-induced thrombocytopenia: Antithrombotic Therapy and Prevention of Thrombosis, 9th ed: American College of Chest Physicians Evidence-Based Clinical Practice Guidelines. Chest 2012; 141 (2 suppl): e495S–530S. doi:10.1378/chest.11-2303 PMID 22315270
68. Warkentin TE, Pai M, Linkins LA. Direct oral anticoagulants for treatment of HIT: update of Hamilton experience and literature review. Blood 2017; 130: 1104–13. doi:10.1182/blood-2017-04-778993 PMID 28646118
69. Schindewolf M, Steindl J, Beyer-Westendorf J, et al. Use of fondaparinux off-label or approved anticoagulants for management of heparin-induced thrombocytopenia. J Am Coll Cardiol 2017; 70: 2636–48. doi:10.1016/j.jacc.2017.09.1099 PMID 29169470
70. Nilius H, Kaufmann J, Cuker A, Nagler M. Comparative effectiveness and safety of anticoagulants for the treatment of heparin-induced thrombocytopenia. Am J Hematol 2021; 96: 805–15. doi:10.1002/ajh.26194 PMID 33857342
71. Galyfos G, Athanasiou A, Emmanuel A, et al. Direct oral anticoagulants as primary or secondary treatment for heparin-induced thrombocytopenia (HIT) and associated thromboembolism (HITT)—a meta-analysis. Eur J Haematol 2026; 117: 1035–43. doi:10.1111/ejh.70238 PMID 42306844
72. Chhatrala H, Ghetiya S, Chahin M, Zuberi L, Quan W. Amiodarone-induced immune thrombocytopenia: a rare hematologic side effect of a common cardiac drug. Cureus 2021; 13: e13671. doi:10.7759/cureus.13671 PMID 33824822
73. Okafor KC, Griffin C, Ngole PM. Hydrochlorothiazide-induced thrombocytopenic purpura. Drug Intell Clin Pharm 1986; 20: 60–1. doi:10.1177/106002808602000110 PMID 3943458
74. Steinhubl SR, Tan WA, Foody JM, Topol EJ. Incidence and clinical course of thrombotic thrombocytopenic purpura due to ticlopidine following coronary stenting. EPISTENT Investigators. Evaluation of Platelet IIb/IIIa Inhibitor for Stenting. JAMA 1999; 281: 806–10. doi:10.1001/jama.281.9.806 PMID 10071001
75. Bennett CL, Connors JM, Carwile JM, et al. Thrombotic thrombocytopenic purpura associated with clopidogrel. N Engl J Med 2000; 342: 1773–7. doi:10.1056/NEJM200006153422402 PMID 10852999
76. Kereiakes DJ, Berkowitz SD, Lincoff AM, et al. Clinical correlates and course of thrombocytopenia during percutaneous coronary intervention in the era of abciximab platelet glycoprotein IIb/IIIa blockade. Am Heart J 2000; 140: 74–80. doi:10.1067/mhj.2000.106615 PMID 10874266
77. McClure MW, Berkowitz SD, Sparapani R, et al. Clinical significance of thrombocytopenia during a non-ST-elevation acute coronary syndrome. The platelet glycoprotein IIb/IIIa in unstable angina: receptor suppression using integrilin therapy (PURSUIT) trial experience. Circulation 1999; 99: 2892–900. doi:10.1161/01.cir.99.22.2892 PMID 10359733
78. Rikken SAOF, van 't Hof AWJ, Ten Berg JM, Kereiakes DJ, Coller BS. Critical analysis of thrombocytopenia associated with glycoprotein IIb/IIIa inhibitors and potential role of zalunfiban, a novel small molecule glycoprotein inhibitor, in understanding the mechanism(s). J Am Heart Assoc 2023; 12: e031855. doi:10.1161/JAHA.123.031855 PMID 38063187
79. Williamson DR, Albert M, Heels-Ansdell D, et al. Thrombocytopenia in critically ill patients receiving thromboprophylaxis: frequency, risk factors, and outcomes. Chest 2013; 144: 1207–15. doi:10.1378/chest.13-0121 PMID 23788287
80. Williamson DR, Lesur O, Tétrault JP, Pilon D. Drug-induced thrombocytopenia in the critically ill: a case-control study. Ann Pharmacother 2014; 48: 697–704. doi:10.1177/1060028013519065 PMID 24662111
81. Mohammadi M, Jahangard-Rafsanjani Z, Sarayani A, Hadjibabaei M, Taghizadeh-Ghehi M. Vancomycin-induced thrombocytopenia: a narrative review. Drug Saf 2017; 40: 49–59. doi:10.1007/s40264-016-0469-y PMID 27848200
82. Zhu Y, Huang L, Zhang J, Liang L, Jin P. Incidence and risk factors of vancomycin-induced thrombocytopenia: a six-year real-world study. BMC Infect Dis 2025; 25: 7. doi:10.1186/s12879-024-10393-1 PMID 39748286
83. Gerson SL, Kaplan SL, Bruss JB, et al. Hematologic effects of linezolid: summary of clinical experience. Antimicrob Agents Chemother 2002; 46: 2723–6. doi:10.1128/AAC.46.8.2723-2726.2002 PMID 12121967
84. Bernstein WB, Trotta RF, Rector JT, Tjaden JA, Barile AJ. Mechanisms for linezolid-induced anemia and thrombocytopenia. Ann Pharmacother 2003; 37: 517–20. doi:10.1345/aph.1C361 PMID 12659607
85. Zhang D, Xu Y, Wang X, et al. Risk factors for thrombocytopenia in patients receiving linezolid therapy: a systematic review and meta-analysis. Eur J Clin Pharmacol 2023; 79: 1303–14. doi:10.1007/s00228-023-03542-z PMID 37578552
86. Choi GW, Lee JY, Chang MJ, et al. Risk factors for linezolid-induced thrombocytopenia in patients without haemato-oncologic diseases. Basic Clin Pharmacol Toxicol 2019; 124: 228–34. doi:10.1111/bcpt.13123 PMID 30171804
87. Oda K, Tsuruta T, Hanai Y, et al. Personalized dosing of linezolid to reduce the risk of thrombocytopenia: a systematic review and meta-analysis. Ther Drug Monit 2025; 47: 529–36. doi:10.1097/FTD.0000000000001300 PMID 39846763
88. Alzahrani M, Alrumaih I, Alhamad F, Abdel Warith A. Rapid onset severe thrombocytopenia following reexposure to piperacillin-tazobactam: report of two cases and review of the literature. Platelets 2018; 29: 628–31. doi:10.1080/09537104.2018.1468025 PMID 29863943
89. Slaught M, Rasmussen M, Bougie D, Aster R. Immune thrombocytopenia induced by beta-lactam antibiotics: cross-reactions of responsible antibodies with other beta-lactam drugs. Transfusion 2021; 61: 1600–8. doi:10.1111/trf.16295 PMID 33580978
90. Pereira J, Hidalgo P, Ocqueteau M, et al. Glycoprotein Ib/IX complex is the target in rifampicin-induced immune thrombocytopenia. Br J Haematol 2000; 110: 907–10. doi:10.1046/j.1365-2141.2000.02299.x PMID 11054081
91. Cheah CY, De Keulenaer B, Leahy MF. Fluoroquinolone-induced immune thrombocytopenia: a report and review. Intern Med J 2009; 39: 619–23. doi:10.1111/j.1445-5994.2009.01996.x PMID 19769684
92. Winter FD Jr. Immune thrombocytopenia associated with consumption of tonic water. Proc (Bayl Univ Med Cent) 2015; 28: 213–6. doi:10.1080/08998280.2015.11929233 PMID 25829658
93. Danziger-Isakov L, Mark Baillie G. Hematologic complications of anti-CMV therapy in solid organ transplant recipients. Clin Transplant 2009; 23: 295–304. doi:10.1111/j.1399-0012.2008.00942.x PMID 19191800
94. Paya C, Humar A, Dominguez E, et al. Efficacy and safety of valganciclovir vs. oral ganciclovir for prevention of cytomegalovirus disease in solid organ transplant recipients. Am J Transplant 2004; 4: 611–20. doi:10.1111/j.1600-6143.2004.00382.x PMID 15023154
95. Richman DD, Fischl MA, Grieco MH, et al. The toxicity of azidothymidine (AZT) in the treatment of patients with AIDS and AIDS-related complex. A double-blind, placebo-controlled trial. N Engl J Med 1987; 317: 192–7. doi:10.1056/NEJM198707233170402 PMID 3299090
96. Scaradavou A. HIV-related thrombocytopenia. Blood Rev 2002; 16: 73–6. doi:10.1054/blre.2001.0188 PMID 11914001
97. Al-Harbi D, Alturaiki A, Alshngeetee A, et al. Linezolid vs vancomycin in induced thrombocytopenia. Infect Dis Ther 2022; 11: 1649–60. doi:10.1007/s40121-022-00663-3 PMID 35727490
98. Harth M, Hickey JP, Coulter WK, Thompson JM, Disney TF. Gold-induced thrombocytopenia. J Rheumatol 1978; 5: 165–72. PMID 209184
99. Madhok R, Pullar T, Capell HA, Dawood F, Sturrock RD, Dick HM. Chrysotherapy and thrombocytopenia. Ann Rheum Dis 1985; 44: 589–91. doi:10.1136/ard.44.9.589 PMID 3929704
100. Taylor HG, Samanta A. Penicillamine in rheumatoid arthritis. A problem of toxicity. Drug Saf 1992; 7: 46–53. doi:10.2165/00002018-199207010-00006 PMID 1536698
101. Grove ML, Hassell AB, Hay EM, Shadforth MF. Adverse reactions to disease-modifying anti-rheumatic drugs in clinical practice. QJM 2001; 94: 309–19. doi:10.1093/qjmed/94.6.309 PMID 11391029
102. Gutierrez-Ureña S, Molina JF, García CO, Cuéllar ML, Espinoza LR. Pancytopenia secondary to methotrexate therapy in rheumatoid arthritis. Arthritis Rheum 1996; 39: 272–6. doi:10.1002/art.1780390214 PMID 8849378
103. Preet Singh Y, Aggarwal A, Misra R, Agarwal V. Low-dose methotrexate-induced pancytopenia. Clin Rheumatol 2007; 26: 84–7. doi:10.1007/s10067-006-0301-7 PMID 16636937
104. Kystaubayeva Z, Zhakupbekova M, Mukhamedjanova A, Nurpeissova R. Hematological side effects of methotrexate in rheumatoid arthritis: a systematic review. Curr Rheumatol Rev 2026; published online ahead of print. doi:10.2174/0115733971454973260424092936 PMID 42163751
105. McEwen J, Purcell PM, Hill RL, Calcino LJ, Riley CG. The incidence of pancytopenia in patients taking leflunomide alone or with methotrexate. Pharmacoepidemiol Drug Saf 2007; 16: 65–73. doi:10.1002/pds.1236 PMID 16634119
106. Chan J, Sanders DC, Du L, Pillans PI. Leflunomide-associated pancytopenia with or without methotrexate. Ann Pharmacother 2004; 38: 1206–11. doi:10.1345/aph.1E012 PMID 15187208
107. Narayan N, Rigby S, Carlucci F. Sulfasalazine induced immune thrombocytopenia in a patient with rheumatoid arthritis. Clin Rheumatol 2017; 36: 477–9. doi:10.1007/s10067-016-3420-9 PMID 27778097
108. Antón Vázquez V, Pascual L, Corominas H, Giménez Torrecilla I. Relapsed hydroxychloroquine induced thrombocytopenia in a systemic lupus erythematosus patient. Reumatol Clin 2017; 13: 294–6. doi:10.1016/j.reuma.2016.04.008 PMID 27263963
109. Relling MV, Schwab M, Whirl-Carrillo M, et al. Clinical Pharmacogenetics Implementation Consortium guideline for thiopurine dosing based on TPMT and NUDT15 genotypes: 2018 update. Clin Pharmacol Ther 2019; 105: 1095–105. doi:10.1002/cpt.1304 PMID 30447069
110. Walker GJ, Harrison JW, Heap GA, et al. Association of genetic variants in NUDT15 with thiopurine-induced myelosuppression in patients with inflammatory bowel disease. JAMA 2019; 321: 773–85. doi:10.1001/jama.2019.0709 PMID 30806694
111. Bessissow T, Renard M, Hoffman I, Vermeire S, Rutgeerts P, Van Assche G. Review article: non-malignant haematological complications of anti-tumour necrosis factor alpha therapy. Aliment Pharmacol Ther 2012; 36: 312–23. doi:10.1111/j.1365-2036.2012.05189.x PMID 22725726
112. Chen M, Holland MJ, Mir MR, et al. Frequency of thrombocytopenia in psoriasis patients treated with tumor necrosis factor-α inhibitors. J Drugs Dermatol 2011; 10: 280–4. PMID 21369645
113. Ram R, Bonstein L, Gafter-Gvili A, Ben-Bassat I, Shpilberg O, Raanani P. Rituximab-associated acute thrombocytopenia: an under-diagnosed phenomenon. Am J Hematol 2009; 84: 247–50. doi:10.1002/ajh.21372 PMID 19260124
114. Cattaneo C, Spedini P, Casari S, et al. Delayed-onset peripheral blood cytopenia after rituximab: frequency and risk factor assessment in a consecutive series of 77 treatments. Leuk Lymphoma 2006; 47: 1013–7. doi:10.1080/10428190500473113 PMID 16840190
115. Emery P, Rondon J, Parrino J, et al. Safety and tolerability of subcutaneous sarilumab and intravenous tocilizumab in patients with rheumatoid arthritis. Rheumatology (Oxford) 2019; 58: 849–58. doi:10.1093/rheumatology/key361 PMID 30590833
116. Verstovsek S, Mesa RA, Gotlib J, et al. A double-blind, placebo-controlled trial of ruxolitinib for myelofibrosis. N Engl J Med 2012; 366: 799–807. doi:10.1056/NEJMoa1110557 PMID 22375971
117. Al-Salama ZT, Scott LJ. Baricitinib: a review in rheumatoid arthritis. Drugs 2018; 78: 761–72. doi:10.1007/s40265-018-0908-4 PMID 29687421
118. Pedersen-Bjergaard U, Andersen M, Hansen PB. Thrombocytopenia induced by noncytotoxic drugs in Denmark 1968–91. J Intern Med 1996; 239: 509–15. doi:10.1046/j.1365-2796.1996.486822000.x PMID 8656144
119. Gentilini G, Curtis BR, Aster RH. An antibody from a patient with ranitidine-induced thrombocytopenia recognizes a site on glycoprotein IX that is a favored target for drug-induced antibodies. Blood 1998; 92: 2359–65. PMID 9746775
120. Yue X, Tian H. Proton pump inhibitors-induced thrombocytopenia: a systematic literature analysis of case reports. Open Med (Wars) 2025; 20: 20251297. doi:10.1515/med-2025-1297 PMID 41216201
121. Farrell RJ, Peppercorn MA, Fine SN, Michetti P. Mesalamine-associated thrombocytopenia. Am J Gastroenterol 1999; 94: 2304–6. doi:10.1111/j.1572-0241.1999.01324.x PMID 10445572
122. Ransford RA, Langman MJ. Sulphasalazine and mesalazine: serious adverse reactions re-evaluated on the basis of suspected adverse reaction reports to the Committee on Safety of Medicines. Gut 2002; 51: 536–9. doi:10.1136/gut.51.4.536 PMID 12235076
123. Beydoun A, Sackellares JC, Shu V. Safety and efficacy of divalproex sodium monotherapy in partial epilepsy: a double-blind, concentration-response design clinical trial. Depakote Monotherapy for Partial Seizures Study Group. Neurology 1997; 48: 182–8. doi:10.1212/wnl.48.1.182 PMID 9008516
124. Nasreddine W, Atweh SF, Beydoun AA, Dirani M, Nawfal O, Beydoun A. Predicting the occurrence of thrombocytopenia from free valproate levels: a prospective study. Seizure 2022; 94: 33–8. doi:10.1016/j.seizure.2021.11.018 PMID 34864249
125. Allarakhia IN, Garofalo EA, Komarynski MA, Robertson PL. Valproic acid and thrombocytopenia in children: a case-controlled retrospective study. Pediatr Neurol 1996; 14: 303–7. doi:10.1016/0887-8994(96)00052-5 PMID 8805173
126. Tutor-Crespo MJ, Hermida J, Tutor JC. Relation of blood platelet count during carbamazepine and oxcarbazepine treatment with daily dose, and serum concentrations of carbamazepine, carbamazepine-10,11-epoxide, and 10-hydroxycarbazepine. Biomed Pap Med Fac Univ Palacky Olomouc Czech Repub 2007; 151: 91–4. doi:10.5507/bp.2007.016 PMID 17690747
127. Meschede A, Runge U, Sabolek M. Thrombocytopenia during levetiracetam therapy. Epilepsy Res 2008; 80: 91–2. doi:10.1016/j.eplepsyres.2008.03.002 PMID 18424076
128. Liu X, Sahud MA. Glycoprotein IIb/IIIa complex is the target in mirtazapine-induced immune thrombocytopenia. Blood Cells Mol Dis 2003; 30: 241–5. doi:10.1016/s1079-9796(03)00037-8 PMID 12737939
129. Kate N, Grover S, Aggarwal M, Malhotra P, Sachdeva MS. Clozapine associated thrombocytopenia. J Pharmacol Pharmacother 2013; 4: 149–51. doi:10.4103/0976-500X.110913 PMID 23761715
130. Okada Y, Kajiyama K, Komamine M, et al. Risk of thrombocytopenia by SSRIs or SNRIs in patients with depression based on MID-NET: a cohort study in Japan. Clin Transl Sci 2025; 18: e70268. doi:10.1111/cts.70268 PMID 40474861
131. Weycker D, Hatfield M, Grossman A, et al. Risk and consequences of chemotherapy-induced thrombocytopenia in US clinical practice. BMC Cancer 2019; 19: 151. doi:10.1186/s12885-019-5354-5 PMID 30764783
132. Wu Y, Aravind S, Ranganathan G, Martin A, Nalysnyk L. Anemia and thrombocytopenia in patients undergoing chemotherapy for solid tumors: a descriptive study of a large outpatient oncology practice database, 2000–2007. Clin Ther 2009; 31 (pt 2): 2416–32. doi:10.1016/j.clinthera.2009.11.020 PMID 20110050
133. Hitron A, Steinke D, Sutphin S, Lawson A, Talbert J, Adams V. Incidence and risk factors of clinically significant chemotherapy-induced thrombocytopenia in patients with solid tumors. J Oncol Pharm Pract 2011; 17: 312–9. doi:10.1177/1078155210380293 PMID 20823048
134. Curtis BR, Hsu YS, Podoltsev N, et al. Patients treated with oxaliplatin are at risk for thrombocytopenia caused by multiple drug-dependent antibodies. Blood 2018; 131: 1486–9. doi:10.1182/blood-2017-10-812461 PMID 29439950
135. James E, Podoltsev N, Salehi E, Curtis BR, Saif MW. Oxaliplatin-induced immune thrombocytopenia: another cumulative dose-dependent side effect? Clin Colorectal Cancer 2009; 8: 220–4. doi:10.3816/CCC.2009.n.037 PMID 19822513
136. Taleghani BM, Meyer O, Fontana S, et al. Oxaliplatin-induced immune pancytopenia. Transfusion 2005; 45: 704–8. doi:10.1111/j.1537-2995.2005.04373.x PMID 15847658
137. Leaf RK, Mones JV, Shenoy T, et al. Immune thrombocytopenia in patients treated with immune checkpoint inhibitors. Blood 2026; 147: 1351–64. doi:10.1182/blood.2025031449 PMID 41359796
138. Martin M, Nguyen HM, Beuvon C, et al. Immune checkpoint inhibitor-related cytopenias: about 68 cases from the French pharmacovigilance database. Cancers (Basel) 2022; 14: 5030. doi:10.3390/cancers14205030 PMID 36291814
139. Moore DC, Elmes JB, Arnall JR, Strassel SA, Patel JN. PD-1/PD-L1 inhibitor-induced immune thrombocytopenia: a pharmacovigilance study and systematic review. Int Immunopharmacol 2024; 129: 111606. doi:10.1016/j.intimp.2024.111606 PMID 38359661
140. Kroll MH, Rojas-Hernandez C, Yee C. Hematologic complications of immune checkpoint inhibitors. Blood 2022; 139: 3594–604. doi:10.1182/blood.2020009016 PMID 34610113
141. Lonial S, Richardson PG, San Miguel J, et al. Characterisation of haematological profiles and low risk of thromboembolic events with bortezomib in patients with relapsed multiple myeloma. Br J Haematol 2008; 143: 222–9. doi:10.1111/j.1365-2141.2008.07321.x PMID 18713253
142. Murai K, Kowata S, Shimoyama T, et al. Bortezomib induces thrombocytopenia by the inhibition of proplatelet formation of megakaryocytes. Eur J Haematol 2014; 93: 290–6. doi:10.1111/ejh.12342 PMID 24750292
143. Shi DS, Smith MC, Campbell RA, et al. Proteasome function is required for platelet production. J Clin Invest 2014; 124: 3757–66. doi:10.1172/JCI75247 PMID 25061876
144. List A, Dewald G, Bennett J, et al. Lenalidomide in the myelodysplastic syndrome with chromosome 5q deletion. N Engl J Med 2006; 355: 1456–65. doi:10.1056/NEJMoa061292 PMID 17021321
145. González-Martín A, Pothuri B, Vergote I, et al. Niraparib in patients with newly diagnosed advanced ovarian cancer. N Engl J Med 2019; 381: 2391–402. doi:10.1056/NEJMoa1910962 PMID 31562799
146. Berek JS, Matulonis UA, Peen U, et al. Safety and dose modification for patients receiving niraparib. Ann Oncol 2018; 29: 1784–92. doi:10.1093/annonc/mdy181 PMID 29767688
147. Bye AP, Unsworth AJ, Desborough MJ, et al. Severe platelet dysfunction in NHL patients receiving ibrutinib is absent in patients receiving acalabrutinib. Blood Adv 2017; 1: 2610–23. doi:10.1182/bloodadvances.2017011999 PMID 29296914
148. Humphreys BD, Sharman JP, Henderson JM, et al. Gemcitabine-associated thrombotic microangiopathy. Cancer 2004; 100: 2664–70. doi:10.1002/cncr.20290 PMID 15197810
149. Cantrell JE Jr, Phillips TM, Schein PS. Carcinoma-associated hemolytic-uremic syndrome: a complication of mitomycin C chemotherapy. J Clin Oncol 1985; 3: 723–34. doi:10.1200/JCO.1985.3.5.723 PMID 3923162
150. O'Leary ST, Glanz JM, McClure DL, et al. The risk of immune thrombocytopenic purpura after vaccination in children and adolescents. Pediatrics 2012; 129: 248–55. doi:10.1542/peds.2011-1111 PMID 22232308
151. Lægreid IJ, Hald EM, Arsenovic MG, et al. Adding contrast to shed some light—a case of acute thrombocytopenia due to iohexol-dependent antibodies to glycoprotein IIb/IIIa. J Thromb Haemost 2026; 24: 298–302. doi:10.1016/j.jtha.2025.09.016 PMID 41076271
152. Hermos JA, Quach L, Gagnon DR, et al. Incident severe thrombocytopenia in veterans treated with pegylated interferon plus ribavirin for chronic hepatitis C infection. Pharmacoepidemiol Drug Saf 2014; 23: 480–8. doi:10.1002/pds.3585 PMID 24677630
153. Watanabe N, Narimatsu H, Noh JY, et al. Antithyroid drug-induced hematopoietic damage: a retrospective cohort study of agranulocytosis and pancytopenia involving 50,385 patients with Graves' disease. J Clin Endocrinol Metab 2012; 97: E49–53. doi:10.1210/jc.2011-2221 PMID 22049174
154. Burday MJ, Martin SE. Cocaine-associated thrombocytopenia. Am J Med 1991; 91: 656–60. doi:10.1016/0002-9343(91)90221-i PMID 1750437
155. Greinacher A, Thiele T, Warkentin TE, Weisser K, Kyrle PA, Eichinger S. Thrombotic thrombocytopenia after ChAdOx1 nCov-19 vaccination. N Engl J Med 2021; 384: 2092–101. doi:10.1056/NEJMoa2104840 PMID 33835769
156. Schultz NH, Sørvoll IH, Michelsen AE, et al. Thrombosis and thrombocytopenia after ChAdOx1 nCoV-19 vaccination. N Engl J Med 2021; 384: 2124–30. doi:10.1056/NEJMoa2104882 PMID 33835768
157. Pavord S, Scully M, Hunt BJ, et al. Clinical features of vaccine-induced immune thrombocytopenia and thrombosis. N Engl J Med 2021; 385: 1680–9. doi:10.1056/NEJMoa2109908 PMID 34379914
158. See I, Su JR, Lale A, et al. US case reports of cerebral venous sinus thrombosis with thrombocytopenia after Ad26.COV2.S vaccination, March 2 to April 21, 2021. JAMA 2021; 325: 2448–56. doi:10.1001/jama.2021.7517 PMID 33929487
159. Warkentin TE, Greinacher A. Platelet-activating anti-platelet factor 4 disorders. N Engl J Med 2026; 395: 478–91. doi:10.1056/NEJMra2502459 PMID 42441431
160. US Food and Drug Administration / Mullan Pharmaceutical Inc. Argatroban injection, for intravenous use: prescribing information, revised 16 June 2026. DailyMed
161. US Food and Drug Administration / Dr. Reddy's Laboratories Limited. Bivalirudin injection: prescribing information, revised 16 July 2019. DailyMed
162. US Food and Drug Administration / Mylan Institutional LLC. ARIXTRA (fondaparinux sodium) injection, solution: prescribing information, revised 10 December 2024. DailyMed
163. US Food and Drug Administration / Hospira, Inc. Vitamin K1 (phytonadione) injection, emulsion: prescribing information, revised October 2025. DailyMed
164. Bauersachs RM, Lindhoff-Last E, Klamroth R, et al. Danaparoid—consensus recommendations on its clinical use. Pharmaceuticals (Basel) 2024; 17: 1584. doi:10.3390/ph17121584 PMID 39770426