Haemostasis & Thrombosis · Critical Appraisal

ATRA plus Eltrombopag in Refractory/Relapsed ITP

A 12-week course of all-retinoic acid added to eltrombopag raised the 18-month sustained response rate from 35% to 60% in glucocorticoid-resistant or relapsed immune thrombocytopenia. A promising second-line signal — with a baseline imbalance and a teratogenicity caveat that both deserve attention.

Bottom line first

In 96 adults with glucocorticoid-resistant or relapsed ITP, adding a 12-week course of all-retinoic acid (ATRA) to eltrombopag raised the 18-month sustained response rate from 35% to 60% (odds ratio 2.78; 95% CI 1.22–6.37; P=0.014). That is a 25 percentage point absolute gain, or roughly four patients treated for one extra sustained response.

Median response duration more than doubled, from 37 to 75 weeks, and no grade 3–4 adverse events or treatment-related deaths were reported.

Two things temper it. The trial was open-label with 96 patients at centres in a single country. And clinically important bleeding at baseline was present in 21% of the combination arm versus 10% of the monotherapy arm — an imbalance that, as discussed below, probably runs against the combination rather than in its favour. Not practice-changing on its own. Pending BSH review.

Download the full appraisal (PDF)


How to read this page

The trial is presented in the standard structure: PICO, headline numbers, critical appraisal, bottom line. A separate section covers ATRA safety, which matters more here than the efficacy signal for anyone considering off-protocol use. UK practice notes, teaching points and references follow.


The trial

Fu H et al. NEJM Evid 2026;5(8):EVIDoa2500333 · Multicentre randomised open-label trial (NCT05438875)

PICO

P · 96 adults with glucocorticoid-resistant or relapsed immune thrombocytopenia and platelets below 30×109/L
I · All-retinoic acid (ATRA) for 12 weeks plus eltrombopag
C · Eltrombopag monotherapy
O · 18-month sustained response — platelets ≥30×109/L without clinically important bleeding or rescue therapy

Headline results

OutcomeEltrombopag alone (n=48)ATRA + eltrombopag (n=48)Effect estimate
18-month sustained response17 (35%)29 (60%)OR 2.78 (95% CI 1.22–6.37); P=0.014
Complete response58%79%Difference 2 to 39 percentage points (95% CI)
Median response duration37 weeks75 weeksHR for relapse 0.45 (95% CI 0.23–0.86)
Grade 3–4 adverse eventsNone reportedNone reportedComparable between arms
Treatment-related deaths00
WHO grade 2–3 bleeding at baseline10%21%Imbalance favouring control at entry

Critical appraisal

Strengths. Randomised, multicentre, with a clinically meaningful primary endpoint assessed at 18 months rather than a short-term platelet count. The endpoint is composite but sensibly so — it requires a platelet response and the absence of clinically important bleeding and no rescue therapy, which is closer to what matters to patients than a platelet threshold alone. Response durability was reported, not just response rate, and the relapse hazard ratio of 0.45 (95% CI 0.23–0.86) supports the durability claim. ATRA is inexpensive and familiar to haematologists.

Limitations. Open-label, 96 patients, single country. With 48 per arm, the confidence interval around the odds ratio is wide (1.22–6.37) and a trial this size cannot exclude chance as a contributor.

On the baseline imbalance. WHO grade 2–3 bleeding was present in 21% of the combination arm versus 10% of the monotherapy arm at entry. This is worth thinking through rather than filing under "imbalance, therefore unreliable". Because the primary endpoint requires the absence of clinically important bleeding, an arm that starts with more bleeding faces a harder task, not an easier one. On that reading the observed advantage is likely conservative. The counter-argument is that baseline bleeding may mark a different disease phenotype altogether, in which case the two arms are not exchangeable and the direction of bias is not knowable. Either way it complicates clean attribution of the effect to ATRA, and it is the first thing a critical reader should raise.

Bottom line

A 12-week ATRA course added to eltrombopag produced a 25 percentage point absolute gain in 18-month sustained response in glucocorticoid-resistant or relapsed ITP, with no reported excess toxicity in this trial. The most interesting second-line ITP signal in some time, but a single open-label trial of 96 patients does not change second-line practice.

1B · Randomised Controlled Trial


ATRA safety — read before considering off-protocol use

Teratogenicity and other ATRA-specific risks

The trial reported no grade 3–4 adverse events and no treatment-related deaths. That is reassuring for the trial population, but it should not be read as "ATRA is a benign drug", and the following apply wherever it is used:

  • Teratogenicity. ATRA is strongly teratogenic. Pregnancy must be excluded before starting and effective contraception maintained throughout treatment and for the period specified in the product information. This matters particularly in ITP, which is common in women of childbearing age.
  • Differentiation syndrome. Well recognised with ATRA in acute promyelocytic leukaemia. It was not a feature of this trial population, but clinicians using ATRA outside APL should remain alert to it.
  • Raised intracranial pressure. Reported with ATRA, particularly in younger patients. Headache and visual disturbance warrant prompt assessment.
  • Hyperlipidaemia and hepatic derangement are recognised with retinoid therapy and warrant monitoring.

Consult the current summary of product characteristics and local policy before use. This page is educational and does not constitute a prescribing recommendation.


What this means for UK practice

Nothing here changes second-line ITP practice today. The BSH guideline on the investigation and management of adult ITP remains the reference point, and this trial has not yet been considered within it.

A reasonable position for now:

  1. Continue current second-line pathways. Thrombopoietin receptor agonists, rituximab and other established options remain as they were. This trial adds a possible adjunct, not a replacement.
  2. Note it for discussion. It is a good MDT and journal club paper precisely because the design question — how to read a baseline imbalance that runs against the treatment arm — is more interesting than the headline number.
  3. If asked about it by a patient or trainee, the honest summary is: promising, cheap, plausible mechanism, single open-label trial of 96 patients in one country, and a teratogenic drug. Not yet standard care.
  4. Watch for replication. A confirmatory trial in a different population, ideally blinded, would move this considerably.
  5. Pending BSH review. No local protocol should change on the basis of this trial alone.

Teaching points

  1. A composite endpoint is not automatically a weak endpoint. Here the composite — platelets ≥30×109/L, no clinically important bleeding, no rescue therapy — is arguably more clinically meaningful than any single component, because it captures the state a patient actually wants to be in.
  2. Work out the direction of a baseline imbalance before dismissing it. More baseline bleeding in the treatment arm makes a bleeding-free endpoint harder to reach. An imbalance that disadvantages the treatment arm makes a positive result more credible, not less — though only if the two groups are otherwise exchangeable.
  3. NNT comes from the absolute risk difference, not the sample size. A 25 percentage point absolute gain gives an NNT of about 4 whether the trial randomised 96 patients or 9,600. Sample size affects the precision of that estimate, not its value.
  4. Response duration is worth as much as response rate in ITP. A therapy that produces the same peak response but sustains it for 75 rather than 37 weeks changes how often the patient returns to clinic, and how often they need rescue.
  5. "No grade 3–4 events" describes the trial, not the drug. Known class effects of a drug — here teratogenicity — do not disappear because a particular trial did not record them.

Discussion questions

  1. Baseline WHO grade 2–3 bleeding was 21% in the combination arm versus 10% in the monotherapy arm. Which direction do you think this pushes the result, and what additional information from the paper would help you decide?
  2. The primary endpoint required the absence of clinically important bleeding and no rescue therapy and a platelet count above threshold. What would change in your interpretation if the trial had used platelet count alone?
  3. ATRA is inexpensive, familiar, and has a plausible immunomodulatory mechanism in ITP. What would you want to see before offering it to a woman of childbearing age with refractory ITP, and how would you frame that conversation?

References

Bibliographic details retrieved and verified against PubMed, July 2026.

  1. Fu H, An Z, Li M, Liu Y, Liu H, Feng R, et al. All-Retinoic Acid plus Eltrombopag for Refractory/Relapsed Immune Thrombocytopenia. NEJM Evid. 2026;5(8):EVIDoa2500333. PMID: 42517709 · DOI: 10.1056/EVIDoa2500333
  2. ClinicalTrials.gov. All-trans retinoic acid combined with eltrombopag in patients with corticosteroid-resistant/relapsed immune thrombocytopenia. Identifier NCT05438875.