“Regression to the truth”: lessons learned from negative IPF trials

Abstract

Idiopathic pulmonary fibrosis (IPF) is a chronic lung disease with limited treatment options. Despite the approval of pirfenidone and nintedanib that slow disease progression, IPF remains a disease with poor survival. Promising therapeutic candidates were tested as potential treatments for IPF and while some drugs were successful in phase II clinical trials, their successful transition to positive phase III was unfortunately disappointing. This highlights the “regression to the truth” concept in drug development, whereby positive phase II trial results may simply be a statistical anomaly rather than the result of true efficacy.

We examine three pivotal trials of novel IPF therapies, zinpentraxin alfa, ziritaxestat and pamrevlumab, that failed in late-stage clinical development. These failures underscore common pitfalls in IPF drug development, including inadequate phase II sample sizes, reliance on surrogate endpoints like forced vital capacity, and challenges integrating background antifibrotic therapies.

Moving forward, innovative approaches like adaptive trial designs, Bayesian statistics and composite endpoints could improve trial robustness. Moreover, platform trials may accelerate drug development by testing multiple therapies simultaneously. Negative trials are not failures but opportunities for learning. By recognising and addressing these challenges, while also embracing novel trial methodologies, we can enhance drug development and improve IPF outcomes.

Shareable abstract

What did we learn from negative phase III trials which followed promising positive phase II trials in IPF? How can we improve the drug development success rate and eventually patient outcomes? https://bit.ly/3CzlY9o

Commentary on:

• Richeldi L, et al. Zinpentraxin Alfa for idiopathic pulmonary fibrosis: the randomized phase III STARSCAPE trial. Am J Respir Crit Care Med 2024; 209: 1132–1140.

• Maher TM, et al. Ziritaxestat, a novel autotaxin inhibitor, and lung function in idiopathic pulmonary fibrosis: the ISABELA 1 and 2 randomized clinical trials. JAMA 2023; 329: 1567–1578.

• Raghu G, et al. Pamrevlumab for idiopathic pulmonary fibrosis: the ZEPHYRUS-1 randomized clinical trial. JAMA 2024; 332: 380–389.

Introduction

Idiopathic pulmonary fibrosis (IPF) is a chronic lung disease characterised by progressive destruction of the lung parenchyma which leads to respiratory failure. It has been a decade since two antifibrotic agents, namely nintedanib and pirfenidone, were approved for IPF management. While individual randomised clinical trials for these drugs did not demonstrate a mortality benefit, pooled analyses indicated a reduction in all-cause mortality and a delay in the time to first acute exacerbation with antifibrotic treatment [1]. Despite these advancements, IPF still remains a devastating disease with a poor prognosis, and with a median survival of approximately 3–5 years.

Several new therapeutic agents for IPF were investigated in recent years, but the results were disappointing. Unfortunately, even promising phase II trial results often led to negative larger phase III trials. Phase II clinical trials are critical for estimating the efficacy of a drug while simultaneously gathering safety data. In the context of drug development, these trials are essential for determining whether a treatment shows enough promise to progress to larger, more definitive phase III trials.

However, it is not uncommon for the effect sizes observed in phase II trials to differ from those in phase III. This phenomenon is not limited to interstitial lung diseases (ILDs), but has been observed across various medical fields. In fact, research suggests that only about 30% of phase III trials yield positive results after a successful phase II trial [2]. One key mitigation strategy identified in other therapeutic areas is that increasing the number of patients enrolled in phase II trials improves the chances of a positive outcome in phase III, as larger sample sizes tend to provide more reliable efficacy data.

It could be argued that researchers and biopharmaceutical companies place undue confidence in the findings of small phase II trials, where certain positive outcomes may be more a product of statistical error than a true indication of efficacy. Here we will introduce the “Regression to the truth” concept which seems particularly relevant in the context of IPF drug development [3, 4]. Regression to the truth is rooted in statistical principles and refers to the tendency for extreme results – either positive or negative – observed in an initial trial to move closer to the true effect when replicated or tested again. This phenomenon is particularly relevant in clinical trials, where early-phase studies with small sample sizes may show exaggerated results due to random chance, selection bias or other confounding factors [4].

When these studies are repeated with a larger sample size and more rigorous design, the results often “regress” or move closer to the true underlying effect, which is typically more moderate. In the case of IPF, a large number of phase II trials were conducted. This makes it statistically plausible that some positive outcomes observed in phase II trials may occur purely by chance rather than reflecting true therapeutic efficacy. The regression to the truth concept underscores the importance of carefully considering sample sizes and study design to reduce the likelihood of false-positive findings. By recognising these trends and applying lessons learned from other disease areas, we can improve the robustness of clinical trials for ILDs. Here, we review and discuss the results of recent IPF trials investigating three proposed novel therapeutic agents: zinpentraxin alfa, ziritaxestat and pamrevlumab (table 1).

TABLE 1.

Summary of phase II and phase III studies of three novel therapeutic agents for idiopathic pulmonary fibrosis

Molecule	Mechanism	Route of administration	Phase II RCT			Phase III RCT
			Study name	Patients, n	Results	Study name	Patients, n	Results
Zinpentraxin alfa	Recombinant human pentraxin-2	i.v. infusion	Phase II (2018) [10]	117	Positive (FVC, 6MWD)	Phase III, STARSCAPE (2024) [12]	664	Negative (terminated early due to futility)
Ziritaxestat	Autotaxin inhibitor	Oral	Phase II, FLORA (2018) [13]	23	Positive (plasma LPA, nonsignificant positive signal for FVC)	Phase III, ISABELA 1 and ISABELA 2 (2023) [18]	525 and 781	Negative (terminated early due to safety concerns)
Pamrevlumab	Monoclonal CTGF antibody	i.v. infusion	Phase II, PRAISE (2020) [22]	103	Positive (FVC)	Phase III, ZEPHYRUS-1 (2024) [24]	356	Negative

RCT: randomised controlled trial; FVC: forced vital capacity; 6MWD: 6-min walking distance; LPA: lysophosphatidic acid; CTGF: connective tissue growth factor.

Zinpentraxin alfa

Pentraxins are proteins produced by the liver in response to infection or inflammation. Pentraxins play a critical role in innate immunity through their function to bind with pathogens and cellular debris. They recognise and bind with Fc receptors and trigger opsonisation and macrophage activation [5]. Pentraxin-1 is the well-known acute phase C-reactive protein (CRP) [6]. Pentraxin-2, also known as serum amyloid P (SAP), is a constituent of human amyloid deposits [7].

Several studies investigated the role of SAP in wound healing and identified its antifibrotic roles. More specifically, SAP was shown to inhibit fibrocyte differentiation and modulate pulmonary macrophage differentiation and fibrotic response [8]. Interestingly, SAP levels are reduced in IPF patients and associated with disease severity [9]. This preclinical evidence led to the development of zinpentraxin alfa, a recombinant human pentraxin-2 protein.

Zinpentraxin was evaluated in a phase II clinical trial which offered optimism for the future [10]. In total, 117 IPF patients were randomised in a 2:1 ratio; 78 patients received the study drug while 39 received placebo. The patients were followed up for 28 weeks. At week 28, the per cent predicted change in forced vital capacity (FVC) was −4.8% in the placebo group and −2.5% in the treatment arm (mean difference 2.3, p=0.001). The change in 6-min walking distance (6MWD) was −31.8 m for the placebo group and −0.5 m for the zinpentraxin group (mean difference 31.3 m, p< 0.001).

The open-label extension of the trial assessed the safety and sustained efficacy [11]. No significant safety concerns were raised over the period of 52 weeks. There was a noted change in the slope of FVC decline in the placebo arm in the cross-over period (−8.7% change per year for weeks 0−28 to −0.9% per year for weeks 28−52; p<0.0001).

In sharp contrast with the phase II clinical trial, the phase III STARSCAPE trial showed no added benefit of zinpentraxin alfa compared with placebo on FVC decline and the key secondary outcomes (6MWD, patient-reported outcomes (PROs) and disease progression) [12]. Altogether, 664 patients were randomised, of which 333 received the study drug. The study was terminated early due to futility. The primary endpoint was absolute change in FVC from baseline to week 52, but the study was terminated due to futility when only 16% of the recruited patients had completed the trial duration. A Bayesian linear extrapolation was applied using data from the 40% of participants who reached week 28, predicting outcomes to week 52. The results revealed no significant difference in FVC change from baseline to week 52 between the placebo and zinpentraxin alfa arms (−214.89 mL versus −235.72 mL, p=0.54), and there was no observed benefit in secondary endpoints.

This raises the question: how did we get here, and what did we learn from this failed IPF trial? The authors attempted to address this by reanalysing the phase II data. A post hoc analysis revealed that the positive results in the phase II trial were primarily driven by two outliers in the placebo group who experienced a stark decline in FVC of >2000 mL per year.

Ziritaxestat

Ziritaxestat is a novel, selective autotaxin inhibitor [13]. Autotaxin is the primary enzyme producing lysophosphatidic acid (LPA), which is considered an important profibrotic mediator in pulmonary fibrosis [14, 15]. The LPA signalling pathway mediates wound-healing responses in lung tissue both through lung fibroblast migration and vascular leak [15]. IPF was associated with elevated LPA concentrations in bronchoalveolar lavage fluid and increased autotaxin concentrations in serum [15, 16]. In animal studies, ziritaxestat reduced LPA C18:2 plasma concentrations in rats and the severity of fibrosis in mice with bleomycin-induced pulmonary fibrosis, which led to investigating it as a potential novel IPF treatment [13].

FLORA was the phase IIa multicentre randomised placebo-controlled trial investigating the safety, tolerability, pharmacokinetics and pharmacodynamics of orally administered ziritaxestat in IPF patients [13]. The study population consisted of 23 stable IPF patients without recent or current antifibrotic treatment, 17 of whom were randomised to receive ziritaxestat for 12 weeks and six of whom received placebo. Patients were followed up to 2 weeks after the end of treatment. Ziritaxestat was well tolerated: the adverse events were mostly mild to moderate in severity and similar in the treatment and placebo groups. Plasma LPA C18:2 concentrations were lower in the ziritaxestat group during treatment and returned to baseline after treatment cessation. Furthermore, ziritaxestat showed positive signs of clinical efficacy, as mean FVC was stable in the intervention group but slightly decreased in the placebo group, although the difference between groups was not statistically significant (p=0.06 at week 14).

The promising results of the phase IIa trial led to conducting the phase III trials, ISABELA 1 and 2 [17, 18]. These identically designed phase III randomised controlled trials compared placebo and two different doses of ziritaxestat (200 mg and 600 mg once daily) in patients with mild-to-moderate IPF with standard of care background therapy, allowing for background antifibrotic therapy. The primary outcome was annual rate of FVC decline at 52 weeks. Key secondary outcomes were disease progression, time to first respiratory-related hospitalisation, and change in St. George's Respiratory Questionnaire (SGRQ).

Enrolment began in November 2018 and the aim was to recruit 750 subjects to each trial. The trials were terminated early in early February 2021, as a planned interim analysis showed lack of efficacy and raised safety concerns. At study termination, ISABELA 1 included 525 patients and ISABELA 2 included 781 patients.

Neither the primary nor the key secondary outcomes were met. There was no observed benefit in the rate of FVC decline despite the promising results of the phase IIa trial. However, the FVC decline was somewhat unexpected in both groups: notably, FVC seemed to decline faster in patients treated with pirfenidone in both the ziritaxestat and placebo groups. In placebo group patients treated with antifibrotics, FVC declined faster than expected based on the original trials of these agents, whereas those without antifibrotic therapy showed a slower FVC decline than expected.

All-cause mortality was higher in both ziritaxestat groups in the ISABELA 2 trial and in the 600 mg dose group in the ISABELA 1 trial. Other secondary outcomes, including time to respiratory-related mortality, respiratory-related hospitalisation and time to first acute IPF exacerbation, were also worse in the ziritaxestat groups. Adverse effects seemed to be more pronounced in patients with background antifibrotic therapy [19].

Pamrevlumab

Pamrevlumab, originally developed as an anticancer agent, is a fully human monoclonal antibody that specifically targets connective tissue growth factor (CTGF), a pivotal regulator within fibrotic signalling cascades. CTGF is synthesised by diverse cell types – including fibroblasts, myofibroblasts and endothelial cells – and interacts with key regulatory molecules such as transforming growth factor-β, vascular endothelial growth factor and integrins. Through these molecular interactions, CTGF orchestrates a spectrum of cellular processes, encompassing extracellular matrix organisation, cell motility and adhesion. These activities collectively underpin maladaptive tissue repair, driving both fibrogenesis and tumourigenesis. Inhibiting CTGF with pamrevlumab disrupts these profibrotic pathways, thereby attenuating aberrant wound healing and fibrotic tissue deposition within the lungs [20].

Preclinical studies in murine model of radiation-induced pulmonary fibrosis, demonstrated pamrevlumab's potential therapeutic efficacy by reversing established lung remodelling and restoring lung function, reinforcing its potential as a therapeutic agent [21].

Pamrevlumab's safety and efficacy were further validated in the phase II PRAISE trial, which was a randomised, double-blind, placebo-controlled trial involving patients with IPF [22]. Conducted from 2013 to 2017 across 39 medical centres in seven countries, the trial enrolled 103 patients who were randomised to receive either pamrevlumab (30 mg·kg⁻¹) or placebo via intravenous infusion every 3 weeks, totalling 16 infusions over a 48-week period. Of the participants, 78 (76%) completed the study, with disease progression constituting the primary reason for discontinuation (13%), alongside patient withdrawal, adverse events and mortality. The primary endpoint, assessing the decline in predicted FVC, was notably lower in the pamrevlumab group compared with the placebo group. From baseline to week 48, the mean FVC decline in the pamrevlumab cohort was −2.9%, compared with −7.2% in the placebo cohort, yielding a significant between-group difference of 4.3% (95% CI 0.4–8.3; p=0.033). This corresponded to a 60% reduction in FVC decline among pamrevlumab-treated patients. Quantitative high-resolution computed tomography scores for lung fibrosis were also significantly lower in the pamrevlumab group at week 24, a difference maintained through to week 48. Although nonsignificant, an improvement in SGRQ scores was observed, favouring pamrevlumab over placebo.

Overall, the findings indicated a marked 60–70% reduction in FVC decline among patients receiving pamrevlumab, surpassing the 50% reduction observed in the pirfenidone and nintedanib trials [23]. These outcomes demonstrated robustness across both relative and absolute measures of FVC decline. Furthermore, significant improvements were recorded in other pulmonary function parameters, imaging outcomes and symptomatology measures. Notably, pamrevlumab exhibited a safety profile comparable to placebo, which, along with its tolerability, was associated with enhanced quality-of-life scores.

However, the phase II results were not replicated in the subsequent phase III ZEPHYRUS-1 trial, a double-blind, randomised controlled trial conducted across 117 sites in nine countries [24]. This trial enrolled 356 IPF patients, of whom 181 received pamrevlumab. Following a 6-week screening phase, participants underwent a 48-week treatment period and were randomised in a 1:1 ratio to receive either 30 mg·kg⁻¹ of pamrevlumab or placebo intravenously every 3 weeks. The primary endpoint was the absolute change in FVC from baseline, evaluated every 6–12 weeks through to week 48. Secondary endpoints included time to disease progression (defined as a decline of ≥10% in predicted FVC or death); time to clinical composite events (IPF exacerbation, respiratory hospitalisation or death); quantitative lung fibrosis volume changes (evaluated via machine learning) from baseline to week 48; time to first IPF acute exacerbation; time to all-cause mortality; and time to first respiratory hospitalisation.

Between July 2019 and April 2022, 612 IPF patients were screened, with 356 meeting randomisation criteria (181 assigned to pamrevlumab, 175 to placebo). FVC decline was comparable in both groups, with a least-squares mean difference favouring pamrevlumab by 70 mL (−260 mL versus −330 mL, respectively). No significant between-group differences were identified for any secondary efficacy outcomes. Given the phase II results, pamrevlumab was anticipated to reduce FVC decline, a prevalent marker of IPF progression, relative to placebo.

Discussion

What did we learn?

Reviewing the collective results of these trials highlights several critical insights for future clinical trial design:

• 1) Sample size and randomisation: One of the main issues in all three trials was the small sample size in the initial phase II studies. Especially in the zinpentraxin alfa study, where the 2:1 randomisation ratio did not allow for accurate effect size calculations. This raises the question whether small and quick phase II studies are sufficient to assess efficiency. Future phase II trials may need to either be longer or use a single-arm phase II design which would assess clinical stability or target engagement as an endpoint. This was suggested for rare diseases and was previously employed in IPF trial design [25]. The use of 3-month FVC change could offer an alternative to increase the sample size of patients receiving active medication while assessing drug safety and efficacy [26].

• 2) Handling outliers in statistical analysis: One point highlighted from the zinpentraxin alfa data is statistical handling of outliers. While excluding outliers may not be ideal, as they could represent biological phenomena, sensitivity analyses with and without outliers should routinely be performed to ensure the robustness of the results.

• 3) Background antifibrotic therapy and the definition of standard of care: The approach to background antifibrotic therapy across trial phases affected the results, as demonstrated in the ziritaxestat and pamrevlumab studies [27]. While the phase II trials excluded antifibrotics, phase III trials included patients with and without these therapies. In the ziritaxestat phase III trial, antifibrotic therapy seemed to have a paradoxically negative effect on FVC decline in placebo group patients. This paradoxical FVC decline raises potential biases in the placebo group and underscores the need to assess real-world representativeness [28]. Patients who respond poorly to antifibrotics may be more inclined to enrol in trials, thus introducing bias.

How can we address the challenge?

Some of the challenges of IPF drug development could be addressed through innovative trial designs, particularly as previous trials were hindered by issues like inadequate sample sizes, regression to the mean and endpoint limitations. Among the most promising approaches is employing adaptive, Bayesian design, which allows trials to dynamically respond to emerging data. Adaptive design provides flexibility by enabling interim modifications to sample sizes, treatment arms or randomisation ratios based on accumulating evidence. This responsiveness is critical particularly in IPF, where the nature of disease progression is unpredictable and requires flexibility. Moreover, adaptive designs not only improve the efficiency of clinical trials but also enhance their ethical integrity by minimising patient exposure to potentially ineffective treatments. These features are increasingly recognised as essential to address the complexities of rare and heterogenous diseases such as IPF.

Bayesian statistics underpin many adaptive trials by continuously updating probabilistic estimates of treatment efficacy, allowing for early stopping or increasing the sample size depending on efficacy signal or futility. This approach was used in the STARSCAPE trial for zinpentraxin, where Bayesian predictive power helped determine early futility based on interim data. Such methodological advancements represent a paradigm shift in clinical research, underscoring the need for statistical innovations tailored to the nuanced demands of specific disease domains, as exemplified by the case of IPF.

Platform and Basket trial designs are also gaining traction [29]. Platform trials are randomised adaptive trials that allow for simultaneous testing of multiple treatments under a single protocol, while increasing the likelihood of the patients receiving an active medication. Basket trials assess the same treatment in multiple diseases or patient subgroups and can thus reveal broader efficacy patterns and identify responder profiles within the heterogeneous broader ILD population. These trial designs exemplify the shift towards more flexible and efficient trial frameworks that align with the principles of precision medicine, aiming to optimise therapeutic strategies based on diverse patient characteristics and disease pathways.

More importantly, we have yet to identify the optimal endpoint for IPF trials. While FVC decline is currently the US Food and Drug Administration (FDA) standard and offers insights into disease progression, it has limitations. Small changes in FVC may not be clinically significant and may not accurately reflect patient symptoms or outcomes. It is time to start employing more clinically meaningful endpoints in accordance with the “Feels, Functions, Survives” principles [30]. Composite measures that integrate FVC with secondary endpoints like 6MWD, PROs and other biomarkers might align better with the FDA's “feels, functions, survives” criteria. PROs assess symptoms and functional impact, while biomarkers provide insights into biological pathways related to IPF progression. Despite methodological challenges, composite endpoints can enhance statistical efficiency and reduce sample size by capturing multiple aspects of disease. The regulatory support for these integrative endpoints highlights a shift towards trials that align closely with patient needs.

Conclusion

In conclusion, despite the disappointing outcomes of three once-promising agents, these trials added valuable lessons. A negative trial is not necessarily a failure but a stepping stone for the future: for example, the PANTHER trial was a milestone, showing that the combination of prednisone, azathioprine and N-acetylcysteine increased risk of death and hospitalisation [31]. The ability to successfully conduct large, randomised trials in a rare disease like IPF is itself a notable achievement.

Moving forward, innovative trial designs, such as adaptive and platform trials, offer opportunities to improve efficiency and increase the likelihood of success. Refining inclusion and exclusion criteria in clinical trials can enhance patient stratification, ensuring a more homogeneous study population and thereby improving the reliability of outcomes in evaluating the efficacy of new therapeutic agents. Furthermore, the integration of composite endpoints that reflect meaningful clinical outcomes can provide a more comprehensive evaluation of patient-centred benefits. By acknowledging past pitfalls and embracing innovation, we can enhance drug development and ultimately, improve the lives of patients living with this devastating disease.