Response to Korley et al.: Progesterone Treatment Does Not Decrease Serum Levels of Biomarkers of Glial and Neuronal Cell Injury in Moderate and Severe TBI Subjects: A Secondary Analysis of the Progesterone for Traumatic Brain Injury,Experimental Clinical Treatment (ProTECT) III Trial (DOI: 10.1089/neu.2020.7072)

To the Editor:

Korley and colleagues ¹ recently evaluated the relationship between progesterone treatment for traumatic brain injury (TBI) and serum levels of biomarkers of glial and neuronal cell death. They reported that no statistically significant differences in these biomarker levels between subjects randomized to progesterone treatment and those randomized to placebo were observed at any time-points examined (baseline, 24 and 48 h).

The authors state that their study was performed to understand the reasons for the negative clinical trial. In particular, they address the issue that “one of the proposed explanations for progesterone's failure to demonstrate improvement in neurological outcome is that the dose of progesterone was suboptimal and therefore inadequate to produce the desired biological effect of neuroprotection. However, this proposed explanation has not been formally investigated.” To the contrary, we and others have addressed the many problems surrounding the negative trial results at length. ^{2 –7} We have learned that, among the many factors influencing the outcome of TBI and its treatment, the heterogeneity of the injury, problematic neuroprotection end-points, suboptimal dosing, potentially critical pre-hospital factors affecting patient enrollment criteria (e.g., intubation and air vs. ground transport), and patient response to post-injury treatments, can confound the trial results.

Further, while there are a few negative results using progesterone, PubMed now lists about 500 papers showing that it is neuroprotective in 22 pre-clinical models of head injury. Recent meta-analyses of the various clinical trials using progesterone for the treatment of TBI also has shown that dosing, duration of treatment, and route of administration all appear to play a critical role in revealing progesterone efficacy. ^8,9

Pre-clinical reports studying progesterone and other agents have demonstrated a bell-shaped dose–response relationship on functional outcome measures in both TBI and stroke, revealing that too high a dose of progesterone produces a washout (hormetic) effect and loss of effectiveness—a finding with an important implication for selecting optimal doses for clinical use. ^{10 –14} These pre-clinical studies suggested that the 12 mg/kg/day human dose used in Progesterone for Traumatic Brain Injury, Experimental Clinical Treatment (ProTECT) II and both phase III trials may have been too high, possibly resulting in the same reduced or lost effectiveness seen in the animal studies at the higher 32 mg/kg dose. In addition, the patients received only 4 days of treatment, whereas 15 days is the human equivalent derived via allometric scaling that leverages pharmacokinetic data from nonclinical studies to predict human drug exposure to convert the rat mg/kg/day dose to the human dose (see Howard and colleagues ⁴ for more details). Based on our allometric evaluation, we suggested that the negative outcomes in phase III could be due to the patients having been given too short a course of treatment with the test drug, at doses that were 6-fold higher than those indicated by allometric scaling. ^3,4

We agree with Korley and colleagues that “the neuroprotective effect of a promising neuroprotective agent may be quantified by serial measurements of blood levels of proteins that are released into circulation following glial and neuronal cell death.” However, we do have concerns with the kinetics of the biomarkers glial fibrillary acidic protein (GFAP) and S100 calcium binding protein B (S100B) impacting the interpretation of the results. For instance, we know that ubiquitin carboxyl-terminal hydrolase 1 (UCH-L1) and αII-spectrin breakdown product of molecular weight 150 (SBDP150) peak at about 6-8 h after injury, and GFAP levels peak about 24-48 h after injury. Even small differences in injury-to-sample time may lead to marked changes in their levels during the first days after injury. ¹⁵ We suggest that measuring these biomarkers as primary (key) tests of efficacy at such a short time-point after the onset of injury, especially in patients with varying age, gender, and severity and locus of injury, could be too soon to evaluate the efficacy of the progesterone treatment and its actions on cerebral edema and modulation of inflammatory factors.

It is important to note that the rodent study ¹⁶ cited by Korley and colleagues in support of the theranostic value of GFAP and UCHL-1 cautions against their use for evaluating therapeutic effect. Although the Browning and colleagues study examined the effect of levetiracetam (LEV) on GFAP in a rodent TBI model and single time-point, the authors concluded that “We did not see a theranostic effect of LEV on GFAP in FPI despite benefit on cognitive outcome. The increase in GFAP at 24 hours in FPI, however, although statistically significant, was modest and did not provide a robust target for a therapeutic effect. Similarly, UCH-L1 was only significantly increased versus sham after injury in PBBI … and thus also did not provide a robust theranostic target.” The same group of researchers in the Operation Brain Trauma Therapy (OBTT) consortium have shown no significant theranostic efficacy for post-injury levels of UCH-L1 at 4 or 24 h in most of their studies. ^{17 –20} We suggest here that the markers themselves may be an indication of glial activity and blood–brain barrier penetration of immune factors into the brain, rather than markers of the salutary effects of treatment—which could take longer than 48 h after TBI, especially severe TBI, to appear. Despite the overwhelming number of pre-clinical studies showing progesterone's neuroprotective effects, no pre-clinical study to date has identified GFAP and UCHL-1 as suitable blood-based biomarkers for assessing progesterone efficacy.

Rather than selecting biomarkers that lack pre-clinical evidence for the efficacy of progesterone, we think it would make more sense to use blood/serum biomarkers shown to be useful readouts for progesterone efficacy in pre-clinical studies, such as those monitoring inflammatory responses, brain mitochondrial dysfunction, and oxidative stress, among others. For example, Rodney and colleagues ²¹ reviewed reports (published between 2006 and 2016) of pro- or anti-inflammatory biomarkers after a TBI in humans and found that increased levels of interleukin (IL)-6, IL-1, IL-8, IL-10 and tumor necrosis factor alpha (TNFα) were observed in various bio fluid/samples (cerebrospinal fluid, serum, and plasma) and were associated with worse outcomes. The findings were interpreted to indicate that these substances could serve as potential prognostic biomarkers of recovery from TBI. Many studies have shown that progesterone is a potential anti-inflammatory agent following acute stroke or TBI ^{22 –25} and there is pre-clinical evidence that progesterone modulates post-stroke systemic inflammation IL-1β, IL-6, and TNFα levels at different times following stroke. ²⁶

More recent studies have highlighted the central role of mitochondrial dysfunction in severe TBI. Serial monitoring of blood mitochondrial enzymes (Complex I [C1], Complex IV [C4] and pyruvate dehydrogenase complex) has been suggested to aid in prognostication ²⁷ and potentially guide mitochondrial-targeted therapies like progesterone. Recently, progesterone was shown to be beneficial in preventing the mitochondrial dysfunction that results in loss of hippocampal cells after a controlled cortical contusion. ^28,29 Although oxidative damage is not highly correlated with the severity of TBI, some markers of oxidative damage such as plasma isoprostane, an indicator of lipid peroxidation, do correlate with the Glasgow Coma Scale and appear to be modulated by progesterone as well. ^{30 –32} Caution should be observed in evaluating the efficacy of progesterone based on these blood/serum biomarkers because they are non–brain-specific, may impact injury severity (Glasgow Outcome Scale-Extended scores), and reflect disability and neuroprotection from multiple causes (e.g., polytrauma, inflammation).

Korley and colleagues also state that “at 24 hours, GFAP values were 22.8% higher than baseline in the progesterone group and 43.6% higher than baseline in the placebo group (i.e., less of an increase in the progesterone group). Similarly, at 48 h, GFAP values were 40.5% lower than baseline in the progesterone group and 27.5% lower than baseline in the placebo group (i.e., more of a decrease in the progesterone group). Although these differences may be scientifically relevant, the study was not powered to detect the interaction between treatment group allotment and time” [our italics]. Overall, we think that several factors, including the complexity of the brain injury, the well-documented insensitivity of behavioral and other functional outcome measures, suboptimal dosing, and other factors often inherent in clinical trial outcomes, can confound clinical trial results. In addition, shorter and restricted time-points sample evaluation for testing the efficacy of a drug after the onset of injury can lead to confounded post hoc biomarker analyses and serious difficulties in interpreting retrospective analyses of published results.

Concerning functional correlation of the biomarker data with behavioral outcome, the authors remark on the unreliability of the Glasgow Outcome Scale-Extended as the primary outcome of the ProTECT trial, an issue we and others have also addressed. ^3,5 As Agrafiotis and colleagues ³³ recently noted, “One of the greatest challenges in clinical trial design is dealing with the subjectivity and variability introduced by human raters when measuring clinical end-points.” These authors' main points continue to reflect what is likely the primary failure for so many clinical trials—the failure to select suitable end-points on which to generalize the effectiveness of a given treatment. A new neuroprotective agent may be very promising, but, in clinical trial, if both biomarkers and behavioral assessments are insensitive or ineffective, no amount of legitimate data manipulation and secondary analysis will result in rejection of the null hypothesis as currently employed.

We completely agree with Korley and colleagues' contention that “The concordance in the association between progesterone treatment and (1) neurologic outcome and (2) biomarker levels suggests that there is a low probability that a future trial of the same dose of progesterone in moderate/severe TBI will yield different results.” We also agree that performing another clinical trial using exactly the same design, parameters and protocols as employed in the ProTECT III trial would not be a good idea. Accordingly, their claim that “Since baseline biomarker levels are associated with TBI severity, we conclude that regardless of the initial severity of TBI, progesterone does not improve clinical outcomes in moderate/severe TBI” is unwarranted in the current context. Further, a report on the association of serum levels of S100B, GFAP, UCH-L1, and SBDP with outcome in ProTECT III, assuming the estimated biomarker effects to be similar in both progesterone and placebo groups, pooled the data from both treatment groups to provide greater power for analyzing the relationship between biomarkers and the primary outcome. ³⁴ We do not understand why, in this case, Korley and colleagues would have expected a different outcome given the confounding factors in the clinical trial that have been discussed here and in previous publications. What we have learned, at best, from this paper is that these biomarkers may be prognostic markers of TBI severity but not necessarily the best biomarkers of progesterone treatment efficacy. All of us in the basic translational science community would very much like to see future trials taking into consideration all of the variables that would increase the likelihood of greater success in finding a safe and effective treatment for TBI.