Primum non nocere: Safety in clinical trials for IPF

The number of clinical trials investigating novel therapies for IPF has increased substantially in the last several years due to new insights into disease pathobiology.1 In this issue of Respirology, Malouf and colleagues report the results of one of these trials, a 36-month randomized controlled trial of everolimus, an mTOR inhibitor.2 Unexpectedly, everolimus therapy led to worse outcomes: it was an independent predictor of the presence of disease progression (hazard ratio 2.37, P < 0.01), was associated with a significantly shorter time to disease progression (median 180 vs 450 days, P < 0.01), and was associated with a trend towards shorter transplant-free survival (P = 0.11). In addition, 68% of subjects randomized to everolimus had stopped the study drug by 12 months compared with only 12% of subjects randomized to placebo. The finding that a potential therapy such as everolimus is harmful to patients is a regrettable but perfectly legitimate outcome in a clinical trial. Through informed consent, research subjects understand that they assume some risk by participating in trials of novel therapies that have potential for harm. Such a result should, however, give us pause to review the ways in which we attempt to minimize risk to clinical trial participants. Among other things, this involves having a strong biological rationale for therapy based on solid preclinical data, an assessment of safety and tolerability in the population of interest, and active oversight of the clinical trial by an independent data and safety monitoring board (DSMB). The biologic rationale for everolimus was relatively strong. As stated by the investigators, everolimus is known to inhibit proliferation of lymphocytes, smooth muscle cells and fibroblasts, and it has been shown to attenuate bleomycin-induced pulmonary fibrosis in rats.2 Although the underlying pathobiology of IPF remains unknown, these findings suggest that everolimus could have an effect on fibroproliferation in the lung, thereby slowing the rate of progression of disease. Everolimus has been used previously in solid organ transplant recipients and has been shown to be safe and effective.3,4 However, to our knowledge it had never been previously tested in patients with IPF. Strategies to assess risk in a novel study population before the initiation of a large clinical trial include phase I and II studies to determine safety, tolerability and optimal dosing. Whether the results of such early phase trials of everolimus in patients with IPF would have demonstrated the problems with tolerability and worse outcome seen in the current study is impossible to determine. Regardless, it is instructive to note how different a profile everolimus appears to have in patients with IPF. While institutional review boards and study investigators serve an important role in minimizing risk during human subjects research, safety in clinical trials is primarily overseen by the DSMB. The DSMB minimizes risk to study participants in several ways: it reviews the study protocol before finalization and can require changes to enhance subject safety; it reviews data on adverse events during the course of the trial and may suggest protocol modifications if there are safety concerns; and it conducts interim analyses of the data (preferably unblinded) to determine if a trial should be stopped early for benefit, futility or harm.5A priori stopping boundaries for these three stopping scenarios are useful as they give objective guidelines to the DSMB.6 Stopping a trial for benefit is not infrequent.7 In IPF, an early study of pirfenidone was stopped due to an apparent reduction in the risk of acute exacerbation, a secondary outcome.8 However, there are potential drawbacks to stopping a trial for benefit, including less confidence in the benefit (e.g. due to reduced statistical significance), inadequate documentation of side effects and the inability to comment on the impact of the intervention on important secondary outcomes and subgroups.9 The pirfenidone example above illustrates these risks well, as subsequent larger studies failed to demonstrate any benefit with respect to risk of acute exacerbation. Stopping a trial for futility is also not uncommon. Futility may be met for reasons other than no detectable difference in the primary outcome measure (i.e. lack of efficacy). These include slow enrolment, high non-compliance, poor data quality and high dropout.10 In IPF, the INSPIRE (interferon gamma-1b) and ARTEMIS-IPF (ambrisentan) trials were both stopped for futility—in both cases for lack of efficacy.11,12 In the current study of everolimus, the high and discordant rate of non-compliance at 12 months surely gave the DSMB cause to consider stopping the trial for futility. Handling negative trends during the conduct of a clinical trial is particularly challenging, and must be done cautiously.6 Perhaps because of these difficult issues, the number of trials stopped for harm is small. Stopping a trial for harm is a difficult decision that goes beyond mere statistics. Stopping a trial for harm may condemn an ultimately useful therapy on the one hand, or provide insufficient evidence to convince investigators to abandon a truly harmful intervention on the other. These two possibilities must be considered in the context of the knowledge that continuing the trial may be exposing participants to additional harm. In addition, DSMBs must consider several factors that may contribute to potentially inaccurate trends observed at interim analyses including baseline differences between groups, potential biases in outcome ascertainment, the impact of missing data, differential co-intervention, non-compliance, the internal consistency of the findings, and ultimately, the impact of stopping the trial early on medical practice and public health. All of these issues likely explain the DMSB's decision to continue the everolimus trial despite negative trends in important outcomes. In the search for an effective treatment for IPF, investigators must remain mindful of a therapy's potential for harm. While a strong biological rationale, early-phase assessment of safety and tolerability, and active oversight by the DSMB provide important patient protections, risk to study subjects cannot be completely avoided. Our responsibility as clinical researchers is to constantly balance risk with benefit, realizing that this is an inherently subjective enterprise. Careful attention to this balance from investigators, study sponsors and DSMB members is essential.