Septic shock is generally conceptualized as a state of pathological immune hyperactivity. Consequently, decades of work on immunomodulation in sepsis have focused on immunosuppressive medications (e.g. steroid, TNF-inhibitors, IL-1 inhibitors). Although most of these interventions haven’t worked, steroid offers some benefits and IL-1 receptor antagonism shows promise for a subset of patients.1
The natural history of septic shock begins with immune hyperactivity, but this is followed by a rebound period of immune suppression. This period of immune suppression might predispose to secondary, healthcare-associated infections. However, it’s difficult to sort out to what extent immunosuppression causes poor clinical outcomes, versus immunosuppression simply being a correlate of extreme illness.
Checkpoint inhibitors have recently emerged as a novel therapy to enhance immune activity against malignancies. These drugs remove normal regulatory steps in the immune response, thereby boosting the immune response (often likened to “cutting off the brakes” of the immune system). The use of checkpoint inhibitors in oncology has exploded within the past several years, with a rapidly expanding list of indications. In parallel, we are gradually realizing that immune enhancement opens a Pandora's box of autoimmune disorders (e.g. myocarditis, pneumonitis, hepatitis, colitis).
Immune checkpoints are involved in the immunosuppression which often follows septic shock. This has stimulated interest in the use of checkpoint inhibitors to reverse this immunosuppression. Some animal models do support a potential benefit of checkpoint inhibitors in septic shock (while others do not).2 Of course, there is also a potential risk that amplifying the immune response could worsen septic shock as shown below:
Stimulating the immune system in septic shock might not go exactly as planned.
Hotchkiss et al. 2019: Immune checkpoint inhibition in sepsis: A phase 1b randomized, placebo-controlled, single ascending dose study of antiprogrammed cell death-ligand 1 (BMS-936559)
This is a phase-1 trial designed to determine the safety of an antibody against antiprogrammed cell death-ligand 1 (BMS-936559) in sepsis.3 Patients were selected who had leukopenia (absolute leukocyte count <1,100/uL) in the context of sepsis plus organ dysfunction (defined as either vasopressor requirement, mechanical ventilation for >1 day, or acute kidney injury). Patients were randomized in a 4:1 fashion to receive BMS-936559 or placebo. Over the course of this trial, the dose of BMS-936559 was progressively elevated (as a phase-1 trial, the goal is to determine the safety and pharmacokinetics of escalating drug doses).
The primary endpoint was assessment of the safety and tolerability of BMS-936559 (including 90-day mortality). Secondary endpoints included pharmacokinetics and immune effects of BMS-936559.
24 patients were included in the study, with baseline demographics as shown below:
Adverse events are shown in the table below. There was a trend towards worse outcomes in patients receiving BMS-936559. None of these differences were statistically significant, due to the tiny number of patients in the control group.
Deaths and severe adverse events were all deemed to be unrelated to the use of BMS-936559, as shown below:
Pharmacokinetic studies demonstrated that the antibody remained bound for prolonged periods of time (often greater than a month). Pharmacodynamic studies found that the study drug increased levels of antigen presentation by monocytes, suggesting a state of heightened immune vigilance.
The authors reached the following conclusions:
Problems with this paper
(1) Exceptional levels of bias
Conflicts of interest are a perennial source of debate. The presence of industry funding and conflicts of interest are common. These certainly don’t disqualify a study from being accurate and useful. However, this study has signs of egregious conflict of interest:
- Five authors are employees of Bristol-Meyers Squibb.
- “Medical writing support was provided by Geraint Owens of Chameleon Communications, with funding from Bristol-Meyers Squibb.” Chameleon communications!
- This is a dose-escalation trial which involves exposing four patients to a certain drug dose, evaluating their response, and then exposing another four patients to a higher dose (if the drug continued to appear safe). The determination of whether the drug was safe was made by “the sponsor's Medical Monitor in consultation with the investigators.” This illustrates the deep level of involvement that pharma had in the conduct of this trial. It might have been preferable to use a data safety monitoring board independent of the sponsoring company.
The use of a professional medical writer funded by pharma is a limitation that is a bit difficult to understand. The paper is coauthored by fifteen people who all hold doctorate degrees. Many of the authors are prolific writers with literally hundreds of citations in Pubmed. Why couldn’t this esteemed team of coauthors manage to write a manuscript by themselves?
(2) Why BMS-936559?
There are numerous FDA-approved drugs available which block the PD-PD1 ligand system (e.g. Nivolumab, Pembrolizumab, Atezolizumab, Durvalumab, Avelumab). There is growing experience with these drugs, which may allow clinicians to select an appropriate agent at a reasonable dose.
If we are going to investigate the blockade of the PD-PD1 ligand system in septic shock, it would probably make sense to start with an established drug that we have familiarity with (say, Pembrolizumab). However, these authors chose to use an experimental agent (BMS-936559). Using an experimental agent for a novel indication creates an unnecessarily confusing situation.
The paper contains no explanation as to why BMS-636559 is being investigated for use in septic shock. Recently, Bristol-Meyers Squibb has been testing the use of BMS-636559 for treatment of HIV (NCT02576457) and sepsis (this study). This doesn’t make sense. If a new, awesome PD1-ligand inhibitor were developed, it would make the most sense to begin by testing its use in cancer (an established indication for PD1-ligand inhibitors) and comparing it against other PD1-ligand inhibitors.
(3) Premature study termination & unclear recruitment
The study was initially planned to have 35 patients. However, it was prematurely terminated because “Business Objectives Have Changed” (above image from ClinicalTrials.gov). The fact that the study was prematurely terminated appears to have been omitted from the manuscript.
According to the manuscript, 35 participants were enrolled but only 24 were included in the final analysis. The reasons that 11 participants were dropped from the study are a bit fuzzy. For example, six patients didn’t have leukopenia (which was an inclusion criteria). Why were these patients initially included, and then excluded?
There’s also a mismatch in the number of hospitals involved, if we compare the initial study plan with the final publication (above). Several sites were initially included, but subsequently disappeared. What happened at these centers which disappeared? Septic shock is a pretty common disease, so it’s difficult to believe that large centers should have much difficulty recruiting a few patients.
(4) Lack of meaningful data about patients’ baseline characteristics
Let’s look again at the table showing patient baseline characteristics. Do you notice anything missing? It’s impossible to determine from this table how sick the patients were. There is no recording of APACHE-II scores, lactate levels, the level of vasopressors patients were on, or how many patients were intubated. These are basic pieces of information generally reported in a sepsis trial.
Why is this information missing? Perhaps it's because the baseline data showed that these patients weren’t all that sick. This would make it look even worse that four patients treated with BMS-636559 died, if the patients weren’t very ill to begin with.
(5) Dubious attribution of adverse events
Adverse events were uniformly not attributed to BMS-636559. However, it’s unclear how this determination could be made. Checkpoint inhibitor toxicity is able to affect any organ of the body, with any timeframe (more on this here). Therefore, it’s almost impossible to definitively say whether an adverse event wasn’t related to BMS-636559. In reviewing the list of adverse events shown above, some conceivably might represent checkpoint inhibitor toxicity.
With a total of 24 patients, this study was only powered to detect obvious toxicity (e.g. anaphylactic reactions). The explanation for the sample size derivation is as follows:
According to this statistical plan, if an adverse event occurred with a rate of 10%, then there was a 34% that they would observe it once. Once! This assumes that the adverse event would be unique and obviously attributable to the drug (e.g. the patient received study drug, turned green, and their head exploded). However, most adverse events could easily blend in among the chaotic background of critically ill patients, thereby eluding detection.
This study should be compared to a similar phase-1 trial by Bristol-Meyers Squibb investigating BMS-936559 among patients with advanced cancer, published in NEJM in 2012.4 The NEJM trial included 207 patients, nearly ten times more patients than the current trial in septic shock. The sample size of the NEJM trial seems more appropriate. A large sample size is required to detect rare events, or to detect toxicities which are not obviously attributable to the drug.
Business objectives versus scientific objectives
This trial was prematurely terminated because “business objectives have changed” (direct quote from ClinicalTrials.gov). It's unclear exactly what happened, but those four words speak volumes. Maybe Bristol-Myers Squibb realized that the study wasn't going to reveal a favorable outcome so they terminated it early. Or perhaps they lost interest in the drug for some other reason? This obviously isn't the way clinical trials are supposed to work (e.g. the decision to stop the study should be made based on pre-determined stopping rules by an independent data safety monitoring board in a fully transparent fashion that is discussed in the manuscript).
This illustrates a schism between business objectives and scientific objective. In the past, some writers have suggested that industry and science have similar objectives so that they can work together in perfect harmony.5 Unfortunately, that's often not true.
We may never know exactly what transpired in this study. When comparing the study manuscript to the study plan in ClinicalTrials.gov, it's clear that things didn't go as planned. From a scientific standpoint, a transparent description of whatever happened would be most beneficial. Unfortunately, from a business standpoint that might not be a wise disclosure.
- Some patients with septic shock eventually develop a state of immunosuppression, which correlates with worse outcomes. This is a phase-1 trial to begin evaluating the role of an experimental checkpoint inhibitor (BMS-936559) to augment immune activity in septic shock.
- There were trends towards worse outcomes in patients who received study drug, but these didn't reach statistical significance due to low statistical power (table below).
- According to ClinicalTrials.gov, this study was terminated early by Bristol-Myers Squibb because “Business objectives have changed.” Study termination isn't mentioned in the manuscript.
- Adjudication of the cause of adverse events may have been biased by deep involvement of Bristol-Meyers Squibb in the execution and writing of the study.
- Currently the use of checkpoint inhibitors for immunocompetent patients with septic shock should be restricted to clinical trials. Ideally such trials should have an independent data safety monitoring board and adequate statistical power to detect uncommon adverse events.
- IBCC chapter & podcast on checkpoint inhibitors.
- Investigation Bias – When does industry roll the dice and fund a study?
- Do conflicts of interest matter?