Journal club critique: Dovramilast for ENL in Nepal (Shah et al., JMIR Res Protoc 2026;15:e88069)

1. One-minute summary for your opening slide

• Question: Is dovramilast (CC-11050, an oral PDE4 inhibitor) safe and tolerable in new or recurrent erythema nodosum leprosum (ENL), and does it show a preliminary efficacy signal?
• Design: Single-centre (Anandaban Hospital, Nepal), single-arm, open-label phase 2 pilot. The dose is 200 mg twice daily with fatty food.
  • Step 1: 10 men, 10 days of treatment, extended to 28 days if they responded.
  • Step 2: up to 40 men and women. Treatment is 12 weeks, with repeat 12-week cycles on relapse, up to 48 weeks, then 12 months of follow-up.
• No steroids or thalidomide are given during the study drug period.
• Primary outcome: the proportion with "complete clinical response" by day 10. Safety is graded with NCI CTCAE v5.0.
• Secondary outcomes: EESS (ENL International Study Group severity scale), Brief Pain Inventory, DN4 (neuropathic pain), recurrence, and exploratory plasma and skin biopsy markers.
• Analysis: descriptive only, with no hypothesis testing and no power calculation.
• Status: step 1 finished in 2020. Step 2 began in September 2024, and 15 participants were recruited by April 2026.

2. Strengths

1. Real unmet need. Prednisolone causes cumulative harm (diabetes, osteoporosis, cataract and glaucoma, infection). Thalidomide is teratogenic and often unavailable. Clofazimine has weak evidence and causes pigmentation. A steroid-sparing oral drug would matter, especially in low-resource settings.
2. Rational biology. PDE4 inhibition raises cAMP and lowers TNF-α, IL-12/23 and neutrophil activation. Those are the pathways implicated in ENL. The authors argue it is chemically distinct from thalidomide (Figure 2). Apremilast has only case reports and small pilots in ENL, for example Narang et al. 2021, a single-centre pilot published as a letter, so a prospective study of a related drug is reasonable.
3. Staged safety design. A small male-only lead-in cohort comes first. An independent DSMB reviewed it before step 2 opened. Women of childbearing potential were added only after that review. This is sensible for a new drug in a population with teratogenicity concerns.
4. Standardised case definition and instruments. ENL is defined using ENL International Study Group criteria and the EESS. Pain is measured with BPI and DN4. AEs are graded with CTCAE v5.0.
5. Long follow-up. Patients are followed for 12 months after stopping the drug. ENL is a relapsing disease, so durability is the clinically relevant question.
6. Translational add-ons. Plasma cytokines, acute-phase reactants, skin biopsy histology and a PK substudy could clarify mechanism in a poorly understood disease.
7. Good setting and governance. It is a national referral centre (60 to 70 ENL patients a year). There is a SPIRIT checklist, national ethics and drug authority approvals, trial insurance, REDCap data management, and GCP monitoring.
8. Honest framing. The authors say plainly that the study will not change practice and that it is "not yet evidence" for clinical use.

3. Weaknesses

A. Design: can it answer the efficacy question at all?

• No control arm, and no blinding. ENL episodes wax and wane, and many improve with supportive care. Without a comparator you can't separate drug effect from natural history or regression to the mean. The outcomes are subjective (pain, tenderness, nodule inflammation, clinician judgement), so open-label assessment inflates apparent benefit.
• Day 10 is a very short window for the primary outcome. The primary endpoint is the proportion with "complete clinical response by day 10", but the study also aims to show durable benefit and fewer recurrences. Prednisolone and thalidomide also act within days, so a day-10 response in a single arm tells you little about whether dovramilast is better, equal or worse.
• The primary research question is safety, but the primary outcome is efficacy. The abstract, aims and research question all lead with safety and tolerability. The formal primary outcome in Methods is clinical response. Safety is listed as an outcome without being a defined primary endpoint. There are no prespecified stopping rules, such as a threshold for serious AEs, in the text I read.
• The primary outcome is poorly defined. "Complete clinical response" is defined as "clinically meaningful improvement or resolution" of nodules, fever and nerve pain. That is not "complete". There is no EESS cut-off, no required fall in lesion count, and no rule for combining skin, fever and neuropathy components. The authors can redefine it after seeing the data, which is the outcome-switching risk. The cut-off should be fixed beforehand.
• A hypothesis the design can't test. The stated aim includes whether treatment "reduces recurrence without the need for prolonged corticosteroid or thalidomide use". But no steroids or thalidomide are given during the study drug period. There is also no comparator, so steroid sparing can't be measured. Cumulative steroid dose and time to rescue are not outcomes.
• Dose and schedule not justified in this population. 200 mg twice daily with fatty food is fixed, with no dose-finding in ENL. PK is measured only in a subset, so exposure can't be linked to response. Nothing is said about how food intake will be verified.

B. Selection and enrichment bias

• Responders-only extension in step 1. Participants who didn't respond by day 10 could be removed, and only responders continued to day 28. The 28-day safety data therefore come from patients selected for benefit, which is survivor bias. If step 1 safety is pooled with step 2, say how.
• ENL definition: BL or LL with a bacterial index above 3+ and 10 or more lesions. Bacterial index falls during MDT, so patients who relapse years after treatment may no longer meet the BI criterion, depending on whether the baseline or the current value is used. The text also mixes "≥3+" in the introduction with ">3+" in the case definition. The 10-lesion threshold excludes milder disease while the "moderate to severe" label is not defined. There is also an unspecified "physician discretion" rule for enrolling people with severe neuropathy.
• Exclusions limit generalisability. HIV, chronic hepatitis B and C, and active TB are excluded, but co-infection is common in endemic areas. Children and people over 65 are excluded. Pregnant and breastfeeding women are excluded (reasonable, but pregnancy is a recognised ENL trigger).
• Single centre, tertiary referral population. Patients seen here are probably sicker and more experienced with the health system than typical community cases.

C. Ethics and safety

• No steroid or thalidomide during the study drug period for moderate-to-severe ENL. ENL can cause neuritis and permanent nerve damage, and can involve eyes and testes. Is there a defined rescue rule, for example for new or worsening nerve function impairment? The paper describes removal at day 10 for poor response, but I did not find a clear rescue protocol. This is a major point for discussion.
• A 7-day washout from steroids or thalidomide is short. Thalidomide and prolonged steroids leave a clinical carryover effect, and abrupt steroid withdrawal may cause rebound. This confounds both the baseline and the day-10 assessments.
• Scheduled drug withdrawal every 12 weeks is a test of recurrence, but a relapse after stopping might be a rebound or the normal course of the disease. Stopping an effective drug on a schedule may harm the patient. This is justified as a "discontinuation test" without explanation.
• Safety monitoring of PDE4 inhibitors. Known class effects with apremilast include diarrhoea, nausea, headache, weight loss and depression. I did not see a specific mood or weight monitoring plan in the text. Check Figure 4 for the laboratory panel.
• Contraception: "abstinence where contraception is not culturally acceptable" is a weak safeguard for a drug with unknown reproductive toxicity. It also raises questions about autonomy for women in this setting.
• Rifampicin: daily rifampicin is excluded but monthly MDT rifampicin is allowed. The paper doesn't explain the interaction (presumably enzyme induction) or why monthly dosing is acceptable. Other concomitant drugs, such as clofazimine and dapsone, aren't discussed.

D. Sample size and statistics

• No power or precision calculation. The paper calls the study "observational" in the sample size section, which is incorrect for a drug trial, and calls the sample "inadequate" in the limitations. Choosing 40 on feasibility is acceptable for a pilot, but the paper should show what that sample can and can't tell you:
  • With 40 participants and a true 50% response, the 95% CI is roughly 35% to 65%. With step 1's 10 participants, it is roughly 24% to 76%.
  • For safety, the rule of three means that if no serious AE occurs in 40 people, you can only exclude a true rate above about 7.5% (3/40).
• No criteria for "go/no-go" to a phase 3 trial. Pilot trials should predefine what response rate, tolerability and feasibility metrics would justify a randomised trial.
• Missing data are not imputed. Reasonable for a pilot, but with relapsing disease and loss to follow-up, dropout could be informative.
• ITT is defined as everyone receiving at least one dose, which is closer to a modified ITT or safety population. Mixing step 1 (10 or 28 days, men only) and step 2 (12 to 48 weeks, men and women) for one primary analysis is unclear.

E. Measurement

• EESS is described as validated in the paper, but validation is mainly in ENLIST cohorts. It has no established minimal clinically important difference, so "clinically meaningful" improvement is undefined.
• BPI and DN4: were Nepali versions validated for this population? The paper doesn't say.
• Missing outcomes. There is no quality-of-life measure, no days hospitalised, no functional disability score, and no nerve function assessment (monofilament and voluntary muscle testing) as a defined outcome, even though preventing nerve damage is the central clinical goal. Time to relapse and time to first rescue are not specified.
• Definitions of remission and recurrence are not fully specified. Recurrence is "28 days or more" after treatment stops, and the chronic definition differs between the introduction (more than 6 months) and Methods (24 weeks or more).

F. Internal inconsistencies to point out (these show careful reading)

G. Transparency and reporting

• Timing. Step 1 was approved in 2017 and completed in 2020, but this protocol was published in 2026 with step 2 already recruiting. Check whether the registry entry predates enrolment and whether primary outcomes in the registry match the paper. The NCT number suggests registration around early 2019, but I haven't verified it. A protocol written after the trial started needs checking for outcome changes.
• Step 1 "results" without data. The paper says no substantive modifications were needed. It reports no AE counts, no response rate and no PK for the 10 men. That limits the reader's ability to judge the dose choice or the rationale for step 2.
• DSMB role. The DSMB reviewed safety, but the text also describes DSMB clinicians assessing efficacy at days 7 to 10 and deciding on continuation. Mixing monitoring and trial-conduct roles can affect independence. Several DSMB members are also thanked for "guidance" on study design.
• Funding and conflicts. The study was funded by Celgene, then Bristol Myers Squibb, and the drug rights moved to Amgen. Study drug supply goes through Medicines Development for Global Health. The paper states the funder "had no role in conduct or reporting", yet the acknowledgments thank Celgene staff for contributions to study design. Conflicts of interest are listed as "None declared", and one author's affiliation is a private consulting company. This is worth a polite question in discussion rather than an accusation.
• Data sharing: "available as supplemental materials with relevant publications" is vague.

4. How the study could be improved

1. Add a comparator. Options include a randomised controlled trial of dovramilast against standard care (prednisolone), or a steroid-taper design (dovramilast plus steroid versus placebo plus steroid) with cumulative steroid dose as the endpoint. A feasible pilot version is a 2-arm randomised open-label trial with blinded outcome assessment.
2. Blind the assessors, for example by photographing lesions and having an independent dermatologist score them, or by using a different clinician for the EESS.
3. Define the primary outcome precisely and make safety the primary endpoint. For example: "EESS reduced by at least X% or to below Y, with no new nerve function impairment, and no rescue steroid, at day 28." Predefine a MCID. Include day-28 and week-12 timepoints, not just day 10.
4. Predefine rescue criteria and stopping rules: rescue prednisolone for new neuritis or eye involvement, and stopping thresholds for CTCAE grade 3 or higher AEs. Record time to rescue as an outcome.
5. Add outcomes that matter to patients and to disability prevention: nerve function (monofilament, voluntary muscle testing), quality of life (such as EQ-5D), hospitalisations, cumulative steroid exposure, and days off work.
6. Run a proper pilot with progression criteria. Specify recruitment rate, retention, adherence (pill counts, fatty-food compliance) and a go/no-go threshold. Report CIs with the estimated sample needed for phase 3.
7. Report step 1 data fully (AEs, response, PK) so readers can judge the dose.
8. Use dose-response and PK/PD properly with sampling in all participants, or at least in enough of them to link exposure to response.
9. Broaden the population: multicentre or multi-country recruitment (Brazil, India, Ethiopia and so on), include HIV-controlled patients and people with chronic ENL, and plan for adolescents later.
10. Clarify and harmonise the protocol text: weights, durations, endpoints, BI cut-offs, chronic ENL definitions, and the follow-up schedule. Add a prespecified analysis for repeat-cycle patients (time-to-event or recurrent-event methods) rather than only descriptive summaries.
11. Strengthen the mechanistic arm by prespecifying biomarkers, sample timing and a hypothesis (for example TNF-α, IL-6, neutrophil markers), rather than "descriptive" analysis.
12. Improve independence and transparency: publish the protocol and registry changes with dates, separate DSMB safety roles from efficacy decisions, and share anonymised data.

5. Overall critical appraisal

6. Discussion questions to put to the group

1. Is it acceptable to withhold steroids and thalidomide in moderate-to-severe ENL for a single-arm trial? What rescue criteria would you require?
2. Why is the primary outcome efficacy at day 10 when the stated primary question is safety? What should "complete response" mean?
3. How much can an open-label, single-arm trial in a relapsing, fluctuating disease tell us? What would a randomised pilot look like in a centre with 60 to 70 cases a year?
4. Does the responders-only extension in step 1 bias the safety picture?
5. Is the male-first, women-later sequencing justified, and are the contraception requirements adequate?
6. What do the existing apremilast case reports and pilot add? A related PDE4 inhibitor with positive reports may increase publication bias in the evidence base.
7. What decision rules should link this pilot to a phase 3 trial?
8. How should the funding history (Celgene, Bristol Myers Squibb, Amgen) and the late publication of the protocol affect how we read it?

7. Literature context (quick check)

• The existing evidence on PDE4 inhibitors in ENL is case reports and a small pilot, all with apremilast: Narang 2021, pilot, PMID 34365679, Narang 2020, 2 cases, PMID 31218666, and further case letters from 2021 (PMID 32970828) and 2022 (PMID 34890096). These support the idea but are low-grade evidence.
• A 2026 semi-systematic review (PMID 41871154) addresses the natural history of ENL and its classification. That is directly relevant to your point about spontaneous fluctuation and inconsistent definitions. I only saw the title and metadata, so read it before citing details.
• My quick search of recent systematic reviews and meta-analyses found nothing that contradicts or updates this protocol. This was a limited check, not a full literature review.

1. Trials of apremilast in ENL

2. Is dovramilast different from apremilast?

• Same class and same origin. Both are oral PDE4 inhibitors. PDE4 inhibition raises cAMP and lowers TNF-α and other inflammatory cytokines. Both come from Celgene, and the protocol says the dovramilast rights passed to Amgen. I recall Amgen also acquired apremilast (Otezla) in 2019, but the paper does not say so. If that is right, the same company now owns both drugs.
• Different regulatory status. Apremilast is approved for psoriasis, psoriatic arthritis and Behçet oral ulcers. Dovramilast is investigational. The protocol cites phase 1 and phase 2 work in HIV, tuberculosis and SARS-CoV-2, where it was tested as an add-on to treat the host response.
• Different dose and food rule. Apremilast is usually 30 mg twice daily. The protocol uses dovramilast 200 mg twice daily "with fatty food". The dose gap suggests different potency or exposure, but I can't confirm that from the paper.
• Apremilast pharmacology. It has a half-life of about 6 to 9 hours. It is metabolised mainly by CYP3A4, and strong inducers such as rifampicin reduce its levels (review). The protocol's rifampicin exclusion looks like it follows the same logic, but the paper doesn't explain it.

3. Why not run an RCT of apremilast first?

1. Timing (documented in the paper). Step 1 was approved by the ethics board in August 2017 and ran until 2020. The first apremilast ENL case report I found is from 2020, so the apremilast case series did not exist when this trial was designed. The question is more relevant now, with the protocol published in 2026.
2. Sponsor and drug supply (documented). Celgene funded the study, then the rights moved to Amgen. Supply is organised through Medicines Development for Global Health. Trials follow whoever supplies the drug, and the paper's acknowledgments mention advocacy for continued use of CC-11050/dovramilast in ENL.
3. Access and cost (my inference, not stated). Brand-name apremilast is expensive for patients in low-income endemic countries. A non-profit-supplied investigational drug might become more accessible. Generic apremilast exists in some countries, so this argument is weaker than it was, and I haven't checked Nepal's market.
4. Regulatory logic. A new indication in a new population needs its own safety data. Apremilast's psoriasis safety record does not remove that need for apremilast either.
5. Possible rationale as a host-directed anti-mycobacterial adjunct. Dovramilast was previously studied alongside antituberculosis therapy. The authors may think it suits patients on MDT, but the protocol doesn't argue this.

• If apremilast is already licensed and has case-based signals, an RCT of apremilast (for example against prednisolone, or as a steroid-sparing add-on) could answer a practical question faster, with a known safety profile and probably cheaper access.
• A phase 2 safety pilot of an investigational drug, in which nothing shows it is better than apremilast, adds a step. A trial comparing both drugs, or a three-arm design, would be more informative.
• The protocol never explains why dovramilast was chosen over apremilast. It also says apremilast evidence is "scarce", which supports testing apremilast just as much as dovramilast.
• Some will see sponsor-driven drug selection as a source of bias. The conflicts-of-interest statement says "none declared", despite the funding history.

4. Suggested discussion questions

1. What advantage would dovramilast need over apremilast to justify a separate development path in ENL?
2. Would a randomised pilot of apremilast against prednisolone be a better use of resources?
3. Could the next step be a head-to-head or three-arm trial?
4. Should an investigational drug be studied first in a neglected disease when a marketed analogue has early signals? What about access and pricing after approval?
5. How should the funding history shape how we read the protocol?

1. Agents already used or tried in ENL

2. What dovramilast could offer, and what remains unproven

• No thalidomide-like neuropathy or teratogenicity. This is the main selling point. Patients with ENL often already have nerve damage, so a drug that doesn't add neuropathy matters. Dovramilast's reproductive toxicity in ENL is still unknown (the protocol says as much), so "not teratogenic" is a hope, not a finding.
• Oral and cheap to administer. It has no injections or infusions, and apremilast needs little routine lab monitoring in psoriasis (my background knowledge). Whether dovramilast needs the same monitoring is not stated.
• Less global immunosuppression than steroids, methotrexate or anti-TNF agents. Anti-TNF drugs carry TB reactivation risk, which matters in endemic settings.
• Possible fit with MDT. It was previously studied as an add-on in TB.

• Slower onset (my inference). PDE4 inhibitors take weeks to work in psoriasis. If the same holds in ENL, a day-10 primary endpoint with no steroid cover could underestimate benefit or leave patients with severe ENL untreated during the lag. Thalidomide and steroids act within days. I haven't seen onset data for dovramilast.
• Class side effects (GI upset, headache, depression, weight loss) are known for apremilast and presumably relevant here.
• Efficacy is not shown. It has no controlled data in ENL, and as the apremilast discussion showed, there is no evidence it beats its sister drug.

3. Is it actually steroid-sparing in this study?

4. A cautionary precedent: the methotrexate RCT

• 137 participants (68 methotrexate, 69 placebo).
• 62% needed additional prednisolone for a flare by 24 weeks, and 74.5% by 48 weeks.
• Adjusted hazard ratios were 0.98 (95% CI 0.62 to 1.54) at 24 weeks and 0.95 (0.62 to 1.43) at 48 weeks. So there was no steroid-sparing effect.
• Quality of life improved a lot in both arms, with no difference between them.

1. A drug that looks reasonable on mechanism and in case series can fail in a controlled trial.
2. Improvement in a single-arm study such as the dovramilast one could be natural history or placebo effect, since both arms improved substantially in MaPs.
3. The background flare rate is high (about 60% at 24 weeks even on prednisolone), so recurrence data without a control arm will be hard to interpret.

5. Other recent pointers

• A 2025 JEADV letter on novel drugs for recalcitrant ENL (PMID 40035479) may be worth a look. I only saw its title, so I can't summarise it.
• The 2026 semi-systematic review of ENL natural history (PMID 41871154) is relevant to the problem of classification and spontaneous fluctuation.
• Nothing I found updates or contradicts the protocol, but MaPs is the most relevant recent controlled evidence.

6. One-line answers for the discussion

• How is dovramilast different from existing drugs? It is an oral PDE4 inhibitor, a different mechanism from steroids, thalidomide, methotrexate and TNF blockers, with a theoretically safer profile but no ENL efficacy data.
• How is it different from apremilast? No demonstrated difference (see my previous answer).
• Is it a steroid-sparing agent? Not as tested. It is evaluated as monotherapy without a control.
• What would convince you? A randomised, double-blind, add-on trial like MaPs, with time to first flare needing extra steroids and cumulative steroid dose as endpoints.

What the statement says

• 2008 to 2012 warnings. Healthcare professionals were told of rare but serious liver injury and severe blistering skin reactions. The EMA also stated that, because of hepatic, cardiac (QT prolongation, especially in women and the elderly) and intestinal concerns, moxifloxacin should only be used when other antibiotics can't be used (review of moxifloxacin safety).
• 2018 to 2019 class review. All fluoroquinolones were restricted because of rare but disabling, long-lasting and possibly irreversible musculoskeletal and nervous system effects, such as tendon damage and peripheral neuropathy (EMA statement).

Putting it together

How to critique it in journal club

• A regulatory warning is not trial evidence. Ask whether the authors give PK/PD data or a published dose-finding study showing the new dose still works.
• EMA signals reflect rare events across large, mixed populations, often with long courses. Short courses in selected patients may carry lower risk, so the dose cut may cost efficacy for little safety gain.
• Under-dosing a fluoroquinolone risks treatment failure and resistance. That loses an important drug class for later regimens, which matters in TB and leprosy.
• Timing matters. Was the change prespecified, or a mid-trial amendment after the safety news? Changing doses during a trial complicates how you pool participants who received different doses.
• Monitoring. What was done to monitor QT, liver enzymes, tendon symptoms and neuropathy? Neuropathy is a particular concern if participants already have nerve damage.