The underlying study that sparked the debate was ambitious in scale and duration. Drawing on the Veterans Affairs Informatics and Computing Infrastructure, known as VINCI, the Houston team and international collaborators from Montpellier, Riyadh, Bronx, Beijing, and Istanbul followed sleep clinic patients for a quarter of a century, comparing mortality in people diagnosed with narcolepsy against other sleep clinic patients using propensity matching. Propensity matching is a statistical technique that attempts to simulate the balance of a randomized trial by pairing individuals with similar baseline characteristics, such as age, sex, comorbidities, and medication use, across comparison groups. The goal was to isolate whether narcolepsy itself, independent of other factors, is associated with elevated death rates.

In their letter, Sauvet and colleagues raised a fundamental concern about that design. Veterans, they argued, are not an ordinary patient population. Military service carries a distinctive burden of psychological trauma, and post-traumatic stress disorder is both more common in veterans and intimately entangled with sleep. A landmark review by Anne Germain in the American Journal of Psychiatry described sleep disturbances as the hallmark of PTSD, and subsequent research has documented how nightmares and insomnia in military service members with severe PTSD can be assessed through ecological, real-world approaches. If veterans with narcolepsy differ systematically from non-veterans with narcolepsy in ways that affect survival, then any mortality estimate drawn from a VA cohort may reflect the veteran experience as much as the sleep disorder itself.

The critique gains force from the broader cardiovascular literature. A meta-analytic review by Donald Edmondson and colleagues in the American Heart Journal found that PTSD is associated with elevated risk for coronary heart disease, one of the leading causes of death worldwide. Because narcolepsy has itself been linked in some studies to increased cardiovascular and cerebrovascular risk, the letter writers reasoned that a veteran-heavy sample could inflate apparent narcolepsy mortality if PTSD and its downstream consequences were unevenly distributed. In other words, the observed excess deaths might be attributable not to narcolepsy but to the trauma-related comorbidities concentrated in military populations, a classic case of confounding in observational epidemiology.

Confounding of this kind is one of the most persistent challenges in sleep medicine research. Unlike a randomized controlled trial, where chance alone determines who receives an exposure, observational cohorts inherit every preexisting difference between the groups they compare. Narcolepsy is a rare chronic disorder caused in most cases by the loss of hypocretin-producing neurons in the hypothalamus, producing irrepressible daytime sleepiness, cataplexy, and fragmented nocturnal sleep. Because large narcolepsy cohorts are hard to assemble, researchers often rely on health system databases, and those databases come embedded with the demographics of the institutions that created them. A VA database is overwhelmingly male and disproportionately shaped by military service; a national registry in France or China carries an entirely different profile. Results from one setting may not travel cleanly to another.

The Houston team’s reply, authored by Sharafkhaneh together with Yves Dauvilliers, Ahmed S. BaHammam, Mehnaz Azarian, Michael Thorpy, Fang Han, Gulcin Benbir Senel, Murat Aksu, Javad Razjouyan and colleagues, engages this critique directly. While the full technical exchange is available in the journal, the reply situates the original findings within the constraints of the VA data environment and addresses how veteran status relates to the mortality signals the team reported. The authors emphasize that their study was explicitly framed as a comparison within a sleep clinic population over 25 years, with propensity matching deployed to reduce the influence of measurable baseline differences. The reply also situates the work against a parallel evidence base, including a 2025 analysis published in JAMA Network Open by Hsu and colleagues that examined all-cause and cause-specific mortality among patients with narcolepsy in a non-veteran setting, allowing readers to weigh the VA findings against broader population data.

That convergence of evidence matters for patients and clinicians. Narcolepsy affects roughly one in two thousand people, and questions about its long-term prognosis have practical consequences: they influence how aggressively clinicians treat cardiovascular risk factors, how they counsel patients about lifestyle and occupational safety, and how health systems allocate resources for a disorder that is often diagnosed years after symptom onset. If the excess mortality observed in a VA cohort were driven primarily by veteran-specific factors such as PTSD, suicide risk, substance use, or service-related physical injuries, then generalizing those figures to the wider narcolepsy population would overstate the danger. Conversely, if narcolepsy independently accelerates mortality through mechanisms such as autonomic dysregulation, metabolic disturbance, or the cumulative toll of chronic sleep fragmentation, then the VA findings would represent a genuine warning signal rather than an artifact of sampling.

The exchange also illustrates how modern epidemiology polices itself. Letters and replies are a low-cost, high-value mechanism for stress-testing observational findings before they harden into clinical dogma. The Sauvet letter did not dispute that the Houston team had conducted a careful, propensity-matched analysis; it disputed what the analysis could legitimately claim about narcolepsy in general. The reply, in turn, had to walk a methodological tightrope, defending the study’s internal validity while conceding the limits of external generalizability that any single-institution cohort faces. This is the same tension that runs through much of big-data medicine, where electronic health records offer enormous sample sizes and long follow-up but cannot guarantee that the population studied resembles the population to whom results will be applied.

For the research community, the practical lessons are concrete. First, veteran status should be recorded and reported as an analytic variable in sleep medicine cohorts, not treated as invisible background. Second, sensitivity analyses that stratify or adjust for trauma-related comorbidities, particularly PTSD and its cardiovascular sequelae, should accompany mortality estimates drawn from military health systems. Third, replication in civilian cohorts, such as the JAMA Network Open analysis of all-cause and cause-specific narcolepsy mortality, remains the most reliable check on whether findings reflect the disease or the dataset. The Houston group’s acknowledgment of VINCI and the Michael E. DeBakey VA Medical Center, along with funding from Baylor College of Medicine, the Center for Innovations in Quality, Effectiveness and Safety, a National Heart, Lung, and Blood Institute K25 career development award, and a research grant from Jazz Pharmaceuticals, underscores the institutional ecosystem behind large-scale VA research, an ecosystem whose strengths and blind spots the current debate makes unusually visible.

What emerges from the exchange is not a demolition of the original study but a sharpening of its interpretation. The 25-year propensity-matched VA cohort remains one of the longest-running examinations of survival in narcolepsy patients within a defined health system, and its findings now come with a clearer warning label about the population from which they were drawn. For the estimated millions living with narcolepsy worldwide, the takeaway is measured: the debate over veteran status does not erase concerns about long-term health in narcolepsy, but it does remind researchers and clinicians alike that every mortality statistic is a portrait of a specific population, and that reading it correctly requires asking who, exactly, was in the frame.

Subject of Research: Confounding by veteran status in mortality analyses of narcolepsy patients in a long-term Veterans Affairs cohort study

Article Title: Reply to “Veteran status as an underappreciated confound in narcolepsy mortality”

Article References: Sharafkhaneh, A., Dauvilliers, Y., BaHammam, A. S., Azarian, M., Thorpy, M., Han, F., Senel, G. B., Aksu, M., & Razjouyan, J. (2026). Reply to “Veteran status as an underappreciated confound in narcolepsy mortality”. Journal of Clinical Sleep Medicine, 22(1), Article 111. https://doi.org/10.1007/s44470-026-00124-5

Image Credits: AI Generated

DOI: 10.1007/s44470-026-00124-5

Keywords: narcolepsy, mortality, veteran status, confounding, propensity matching, PTSD, sleep medicine, VA cohort, cardiovascular risk, epidemiology, sleep disorders, observational study