FDA and EMA Accepted Endpoints and Trial Designs for Complement Inhibitors in Paroxysmal Nocturnal Hemoglobinuria

Endpoints and trial designs accepted by FDA and EMA for complement inhibitors in paroxysmal nocturnal hemoglobinuria (PNH)

Complement inhibitors for PNH have been approved on four distinct evidentiary models: (1) placebo-controlled superiority on hemoglobin stabilization and transfusion burden (eculizumab); (2) active-controlled non-inferiority against eculizumab on transfusion avoidance plus an LDH-based hemolysis endpoint (ravulizumab, crovalimab); (3) superiority on hemoglobin-based endpoints in patients with residual anemia on a C5 inhibitor, either as a switch or as add-on therapy (pegcetacoplan, iptacopan, danicopan); and (4) LDH-based comparative clinical studies for eculizumab biosimilars (Bkemv, Epysqli). Both agencies have accepted single-arm and supportive-care-controlled data in complement-inhibitor-naive patients when a randomized pivotal trial carried the program; the single-arm studies served as supportive, not standalone pivotal, evidence. The main points of friction have been non-inferiority margin derivation, open-label bias in transfusion decisions, and the interpretability of LDH and patient-reported fatigue.

Products and pivotal evidence at a glance

EU marketing authorization dates in the EPAR record range from Soliris in June 2007 to Piasky in August 2024 ^{5857565460525355}. In the US, Voydeya was approved on 2024-03-29 and PiaSky on 2024-06-20 ²⁵³²⁵⁵.

Model 1: placebo-controlled superiority (eculizumab)

Design. TRIUMPH randomized 87 transfusion-dependent adults to eculizumab (43) or placebo (44) for 26 weeks under double-blind conditions ⁴⁶¹⁴⁶². Eligibility required at least four transfusions in the preceding 12 months, a PNH clone of at least 10%, and platelets of at least 100,000/µL ²⁰¹. An observation period before randomization established each patient's hemoglobin transfusion "set-point" (≤9 g/dL if symptomatic, ≤7 g/dL if asymptomatic) ⁴⁶³²⁰¹.

Endpoints. FDA reviewers described co-primary superiority comparisons of (1) the proportion achieving hemoglobin stabilization, defined as remaining above the set-point with no RBC transfusion through 26 weeks, and (2) the number of packed RBC units transfused ⁴⁶³⁴⁶⁴. Secondary endpoints were transfusion avoidance, LDH area under the curve and FACIT-Fatigue ⁴⁶⁶. Stabilization occurred in 21/43 versus 0/44 patients and median units transfused were 0 versus 10 (P<0.001 for both) ⁴⁶⁷⁴⁶⁵. The EPAR describes the same co-primary structure ^720721722.

Reviewer qualifications. FDA did not treat fatigue and quality-of-life findings as a clean independent basis for benefit claims, because quality of life was not reassessed immediately after transfusion and FACIT-Fatigue and EORTC QLQ-C30 validation was questioned ⁴⁷⁰. Thrombosis analyses were confounded by anticoagulant use ⁴⁷¹. The 52-week open-label, single-arm SHEPHERD study was primarily a safety study whose LDH and fatigue data were supportive ^472473724725. In the EU, the TRIUMPH design initially limited the conclusion of clinical benefit to patients with a transfusion history. A later CHMP assessment extended benefit to patients with hemolysis and symptoms of high disease activity regardless of transfusion history, while cautioning that LDH alone was not an adequate measure of disease severity ^731732733734.

Model 2: active-controlled non-inferiority against eculizumab

Once eculizumab was standard of care, next-generation C5 inhibitors were evaluated against it rather than placebo.

Ravulizumab: different primary endpoints for naive and switch populations

• Study 301 (naive). The co-primary endpoints were transfusion avoidance through Day 183 (no transfusion and no meeting of protocol transfusion criteria) and LDH normalization (LDH ≤1 × ULN, Day 29 to Day 183). Both had to meet non-inferiority ³⁹⁸³⁹⁹. Results were 73.6% versus 66.1% for transfusion avoidance (difference 6.8 points; 95% CI −4.66 to 18.14) and 53.6% versus 49.4% for LDH normalization (odds ratio 1.19; 95% CI 0.80 to 1.77) ³⁰³.
• Study 302 (switch). Patients stable on eculizumab for at least six months were randomized to continue or switch. The primary endpoint was percent change in LDH to Day 183, with a 15-point margin ³⁸⁵⁴⁰⁸. FDA reviewers noted that transfusion avoidance was secondary and LDH normalization was not primary because these patients were already stable and largely transfusion-independent ⁴⁰⁹. The result was −0.82% versus +8.4% ³¹⁰.

Margin disputes. The sponsor's margins were −20 percentage points for transfusion avoidance (PNH Registry-based) and an odds ratio of 0.39 for LDH normalization (TRIUMPH-based) ^403404405. FDA had warned that 20 points exceeded the largest clinically acceptable difference and recommended no more than 10 points. The statistical reviewer stated that the margins had never been agreed with FDA and questioned the use of point estimates rather than confidence bounds of the historical effect. FDA nevertheless found Study 301 non-inferior under both the sponsor's and FDA's margins ^406395407. In Study 302, hemoglobin stabilization did not meet FDA's tighter −10-point margin ⁴¹⁰. CHMP also noted that a more conservative TRIUMPH-based derivation (using the lower 95% CI bound) would have yielded an 18.3-point transfusion margin, and that the LDH-normalization margin approach had not been discussed in scientific advice, but did not pursue either point given the results ^660661662663. CHMP accepted the 15-point Study 302 margin as preserving 89% of effect, while identifying reliance on registry data alone as a limitation ⁶⁶⁶⁶⁶⁷.

Design and endpoint shaping by FDA. FDA objected early that transfusion guidelines alone could bias transfusion avoidance, prompting addition of a prespecified hemoglobin criterion. FDA also objected to LDH percent change as a co-primary endpoint in untreated patients, considering normal or near-normal LDH more meaningful ⁴¹¹⁴¹². CHMP considered the open-label design reasonable because the intravenous schedules differed markedly and blinding would have required many sham infusions ⁶⁷⁰. Both studies failed their first superiority tests, so CHMP concluded similarity rather than superiority ⁶⁷¹.

Crovalimab: the same template with tightened margins

COMMODORE 2 randomized naive patients 2:1 to crovalimab or eculizumab for 24 weeks, with co-primary non-inferiority endpoints of transfusion avoidance through Week 25 and hemolysis control (central LDH ≤1.5 × ULN, Weeks 5 to 25) ^501502504509. Results were 65.7% versus 68.1% for transfusion avoidance (difference −2.8 points; 95% CI −15.7 to 11.1) and 79.3% versus 79.0% for hemolysis control (odds ratio 1.02; 95% CI 0.57 to 1.82) ⁴⁴⁴.

• FDA margins. The prespecified margins were −20 points and an odds ratio of 0.20, each intended to preserve half of the estimated eculizumab effect. FDA reviewers rederived them from the lower confidence bound of the TRIUMPH effect, producing −18 points and an odds ratio of 0.33. Both were met (lower bounds −15.7 points and 0.57) ^{505506507508510511512}.
• EMA margins. The −20% transfusion margin matched the ravulizumab precedent and was based on registry comparisons ⁶⁹³. CHMP considered the co-primary endpoints clinically meaningful and discussed and ruled out bias in the investigator-driven transfusion decision ^695696697698.
• Bias and missing data. FDA flagged potential transfusion-decision bias in an open-label trial but judged the protocol transfusion guidelines likely to minimize it. FDA found FACIT-Fatigue difficult to interpret and did not include it in labeling ⁵¹³⁵¹⁴. FDA also raised concerns about missing LDH data and the primary GEE analysis, addressed through sensitivity analyses ⁵¹⁵⁵¹⁶.
• External control and supportive studies. FDA rejected a proposed switch-study design using an external comparator as not adequate and well controlled ⁵²⁰. COMMODORE 1 (switch) stopped enrolment early and its efficacy findings remained exploratory; CHMP accepted the descriptive data as supportive despite uncertainty. COMMODORE 3 was a supportive single-arm study in Chinese patients ^699691521.

Model 3: hemoglobin-based superiority in residual anemia (proximal and add-on inhibitors)

Agents targeting C3, factor B and factor D were developed largely in patients with residual anemia on C5 inhibitors, where hemoglobin, rather than LDH, became the primary endpoint.

Pegcetacoplan (C3)

• APL2-302 (switch from eculizumab). In this open-label superiority trial, 41 patients received pegcetacoplan and 39 continued eculizumab for 16 weeks after a four-week run-in. Patients had Hb <10.5 g/dL despite stable eculizumab ³¹⁶¹⁷⁸. The primary endpoint was Hb change to Week 16: +2.37 versus −1.47 g/dL (difference 3.84 g/dL; 95% CI 2.33 to 5.34) ^261264332. Non-inferiority was shown for transfusion avoidance and reticulocyte count, but LDH change did not meet non-inferiority, which stopped hierarchical testing before FACIT-Fatigue ^33433588.
• CHMP view. CHMP judged the endpoints clinically relevant, with transfusion avoidance particularly important. It called the open-label design of a single pivotal trial a drawback but considered the measures taken to minimize bias sufficient ³³⁶³³⁷. CHMP also noted baseline imbalances and the exclusion of post-transfusion observations from the primary analysis ^338339340.
• APL2-308 / PRINCE (naive). Fifty-three patients were randomized 2:1 to pegcetacoplan or supportive care excluding complement inhibitors for 26 weeks, with crossover permitted on Hb decline of at least 2 g/dL ^8486600. The co-primary endpoints were Hb stabilization (no >1 g/dL fall without transfusion), achieved in 85.7% versus 0%, and LDH change, −1,470 U/L between groups ^601609610. CHMP accepted the population as representative but noted the open-label design and called supportive care a suboptimal comparator by EU standards ⁶⁰².

Iptacopan (factor B)

• APPLY-PNH (switch from anti-C5). This randomized, open-label superiority trial enrolled adults with residual anemia (Hb <10 g/dL) despite at least six months of eculizumab or ravulizumab ^538539634. Co-primary endpoints were a sustained Hb rise of ≥2 g/dL and sustained Hb ≥12 g/dL, both without transfusion. Hb was assessed at three of four visits on Days 126 to 168 ⁵⁴¹⁵⁴². Responses were 82.3% versus 0% and 67.7% versus 0% (p<0.0001) ¹⁰⁸⁵⁴³.
• APPOINT-PNH (naive). This single-arm, open-label study enrolled 40 patients with Hb <10 g/dL and LDH >1.5 × ULN ¹¹¹. The lower 95% CI bound for Hb response was compared with a 15% threshold derived from historical eculizumab data, not with a concurrent control ⁵⁴⁸⁶³⁹. FDA's statistical reviewer raised concerns about the model assumptions in the prespecified logistic regression for this sample size ⁵⁴⁹.
• Agency views. CHMP accepted the ≥2 g/dL rise as clinically meaningful and ≥12 g/dL as approximating normalization. It credited guideline-aligned transfusion criteria with reducing bias ⁶⁴². CHMP treated APPLY-PNH as the pivotal controlled evidence and APPOINT-PNH as supportive, and did not restrict the indication to anti-C5-experienced patients ⁶⁴³⁶¹⁶. FDA's statistical reviewer warned that FACIT-Fatigue could be biased by knowledge of treatment in the open-label trial ⁵⁴⁶.

Danicopan (factor D, add-on)

ALXN2040-PNH-301 was double-blind and placebo-controlled, with danicopan added to ongoing eculizumab or ravulizumab. Patients had clinically significant extravascular hemolysis, defined as Hb ≤9.5 g/dL and absolute reticulocyte count ≥120 × 10⁹/L ^587588757. The controlled period lasted 12 weeks, after which placebo patients switched to danicopan ⁵⁸⁸. At the prespecified interim analysis (n=63), Hb increased by 2.9 versus 0.5 g/dL (difference 2.4 g/dL) ⁵⁸⁹⁵⁹⁰. Key secondary endpoints (Hb rise ≥2 g/dL without transfusion, transfusion avoidance, FACIT-Fatigue, reticulocyte count) were tested in a fixed sequence and all favored danicopan ⁵⁹¹⁵⁹³.

CHMP accepted the add-on rationale (C5 blockade for intravascular hemolysis, danicopan for residual extravascular hemolysis), the interim analysis timing and alpha-spending, and the sequential testing ^756758759. It noted that the lack of a control arm after Week 12 hampered interpretation of long-term efficacy ⁷⁶¹⁷⁶². FDA reviewers noted the program was not designed to assess effects on thrombosis and contained no head-to-head comparisons with pegcetacoplan or iptacopan ⁵⁹⁴⁵⁹². The EU indication is restricted to add-on use in adults with residual hemolytic anemia ⁷⁶³. The US indication is likewise add-on to ravulizumab or eculizumab, for extravascular hemolysis in adults with PNH.

Model 4: eculizumab biosimilars (LDH as the comparative endpoint)

Both biosimilar programs used LDH as the sensitive comparative endpoint. The agencies differed on the analysis variable and the margin.

Two points matter for future biosimilar programs. FDA's 0.77 to 1.29 AUEC bound was linked to hemoglobin stabilization, judging a difference of no more than 10% clinically unimportant and preserving at least 80% of the Soliris effect ²⁴⁶. CHMP accepted non-inferiority rather than equivalence for Bekemv because superiority in maintaining efficacy was implausible in stabilized patients ²¹⁵.

Cross-cutting regulatory themes

• Transfusion avoidance needs protocol-defined triggers. Both agencies accepted transfusion avoidance as clinically meaningful in open-label trials only where prespecified transfusion criteria limited investigator discretion ^411513642698.
• LDH is accepted as a hemolysis measure, but framing matters. FDA preferred LDH normalization or threshold-based control over percent change in active disease ⁴¹². LDH percent change was acceptable for maintenance in switch populations ⁴⁰⁹⁶⁶⁸. CHMP cautioned against LDH as the sole severity criterion ⁷³⁴.
• Hemoglobin endpoints dominate where extravascular hemolysis drives residual anemia. Hb change, sustained Hb response and Hb stabilization served as primary endpoints for pegcetacoplan, iptacopan and danicopan ^261541589. LDH failed non-inferiority in APL2-302 without blocking approval ⁸⁸³³⁴.
• Non-inferiority margins are re-derived conservatively. FDA consistently recalculated margins from the lower confidence bound of TRIUMPH effects rather than point estimates or registry data ^395507512. CHMP raised similar points but did not pursue them where results were robust ⁶⁶¹⁶⁶³.
• PROs remain supportive in open-label settings. FACIT-Fatigue was described as potentially biased or difficult to interpret in open-label trials and was excluded from crovalimab labeling ^399413514546.
• Single-arm and external data are supportive, not pivotal. Single-arm studies (SHEPHERD, APPOINT-PNH, COMMODORE 3) supported randomized pivotal trials ^472643691. FDA rejected an external-comparator switch design ⁵²⁰.