Abstract
We appreciate the thoughtful editorial commentary by Honigman et al. (2019) pertaining to our recent article describing the association between suicide risk and altitude in the VA Healthcare System (Sabic et al., 2019). There is much in their commentary with which we agree. We agree, in particular, that many factors can, and do, contribute to suicide other than altitude, and that many of those factors could covary with altitude.
There are, however, two main aspects in which we disagree. First, we dispute the implication that the existence of multiple sociodemographic factors that contribute to suicide risk and that also covary with altitude shows that altitude does not itself contribute to suicide risk. Rather, we think it is possible that altitude has an independent association with suicide rates, although we agree that this issue needs more research. Second, we think the claim that there is no evidence to date of physiological connections between altitude and suicide risk is too strong—there is, instead, preliminary evidence of such an association that also merits further study.
Regarding the first point of disagreement, although many sociodemographic risk factors for suicide could covary with altitude and even explain some of the observed associations between altitude and suicide, there are two reasons to think that the association is genuine, rather than a matter of the ecological fallacy. First, the association has been replicated in numerous regions outside of the United States, including in Chile, Spain, Switzerland, and other countries (Kious et al., 2018). Although we grant the possibility that the relevant altitude-covarying sociodemographic factors of these disparate regions unanimously mirror those of the United States, it seems an unlikely series of coincidences.
Second, the association between altitude and U.S. suicide rates has been replicated using different modeling strategies (Ha and Tu, 2018). We would note, too, that the only reportedly negative study of the relationship between altitude and U.S. suicide rates (Betz et al., 2011) did not show that altitude per se is not significantly associated with suicide risk. Rather, the study confirmed that suicide rates in high-altitude counties were higher than in low-altitude counties (17.7/100,000 vs. 11.9/100,000). The study's authors concluded that “the observed differences in suicide rates at different altitudes might be explained by an ecological fallacy…; that is, there may be no basis for the inference that hypoxia or high altitude operates on an individual level to increase the risk of suicide” (p. 569). But the basis for this inference was their finding that high-altitude suicide decedents differed significantly from low-altitude suicide decedents with respect to a variety of sociodemographic factors, such as race, rurality, current depressed mood, alcohol abuse, and history of suicide attempts. There are two problems with concluding from this that the association between altitude and suicide risk is merely apparent. First, the study design they used precluded assessing whether differences in those other sociodemographic factors fully accounted for the variance in suicide rates between different altitudes; it remains possible that altitude independently predicts suicide risk. Second, many of the identified differences between the high- and low-altitude groups—such as the proportions of persons with a history of depression, alcohol abuse, or suicide attempts—might themselves be consequences of physiologic changes wrought by altitude that ultimately increase the risk of suicide; this is precisely what we hypothesize is the case.
Ultimately, though, we strongly agree with Honigman et al. about the limitations of population-level studies such as ours, and concede that there is, despite the consistency of findings to date, still some risk of succumbing to the ecological fallacy in research of this type. We also, therefore, agree that there is considerable need for person-level studies that assess the contribution of altitude to suicide risk and other mental health problems. Although additional work is needed, recently two studies have provided further preliminary evidence that altitude contributes to suicide risk and depressive symptoms. Riblet et al. (2019) found, in a nation-wide sample of 9,620,944 U.S. military veterans in the Veterans Affairs Health Care System (VA), among whom there were 22,403 suicide deaths during the follow-up period, that there was a statistically significant association between increasing altitude and suicide risk, with an odds ratio of 1.22 for every 1000 m increase in altitude from sea level, despite controlling for age, gender, race, ethnicity, rurality, and VA service use. They also found that other conditions associated with hypoxia, namely chronic obstructive pulmonary disease and current smoking, were significantly associated with increased suicide risk in adjusted models, although the study did not assess hypoxia itself, and did not control for other factors that might affect suicide risk such as psychiatric diagnoses, gun ownership, or substance use.
Similarly, we have recently published an analysis of data from the Intern Health Study—a longitudinal study of psychiatric symptoms and related characteristics among medical interns—in which we found that residing at high altitude (in this case, >900 m) was associated with increased PHQ-9 depression scores and increased suicidal ideation (Kious et al., 2019a). Because the study followed subjects as they moved from one location in medical school to another location for residency, we were able to assess the effect of moving from low altitude to high altitude on psychiatric symptoms over a year of follow-up. We found that moving from low altitude to high altitude was associated with increased PHQ-9 total score, increased GAD-7 total score, and increased suicidal ideation, although this study was also limited by the relatively poor sensitivity to change of the PHQ-9 and GAD-7, and it is noteworthy that increases in these scores associated with living at high elevation did not carry most subjects into the clinically significant range (e.g., a PHQ-9 score of around 10 or more, as evidence of major depressive disorder).
Regarding the second point of disagreement, Honigman et al. state that “one of the accepted criteria for inferring ‘causality’ in epidemiological studies is the existence of a plausible biological mechanism for an observed association.” Although we have merely hypothesized a link between altitude and suicide to this point, and although there is much evidence that cerebral oxygen delivery is maintained across a broad range of altitudes, there are at least two mechanisms that we think could mediate such a link: (1) decreased blood oxygen content leading to altered neurotransmitter synthesis and (2) decreased brain energy stores related to inefficiencies in oxidative phosphorylation. Honigman et al. (2019) reject such possible links, since they claim that “physiological changes indicating hypoxic responsiveness only occur at higher altitudes.”
We think, however, that chronic lower altitude hypoxia may occur. Crapo et al. (1999) demonstrated that in lifelong nonsmokers with no history of lung disease, the partial pressure of oxygen in arterial blood (PaO2) is decreased by ∼20 mmHg in Salt Lake City, Utah, compared with near sea level in Los Angeles, California, or New Haven, Connecticut. In other words, in humans, even a slight increase in altitude might produce significant and sustained reductions in PaO2, although it remains true that compensatory increases in hematocrit, respiratory alkalosis, cardiac output, and cerebral blood flow could correct for the impact of a slightly lower PaO2 and partially or fully maintain cerebral blood flow and oxygen delivery to the brain. Although human data regarding the impact of reductions in PaO2 on standardized measures of cognition and mood are lacking, these data appear to challenge the notion that meaningful hypoxia does not occur until higher elevations. Reduced PaO2 may be significant because central serotonin synthesis is dependent on arterial oxygen tension (Davis et al., 1973), although in this study serotonin synthesis was measurably reduced only at PaO2 values corresponding to ∼3000 m or more.
Likewise, there is evidence that major depressive disorder and other psychiatric conditions, which are clearly linked to suicide, are themselves associated with reduced or altered cerebral energy stores (Kious et al., 2019b). Slight reductions in brain oxygen delivery could, we think, reduce cerebral adenosine triphosphate (ATP) synthesis and alter subsequent production of phosphocreatine, leading to relative inefficiencies in brain energy storage. Indeed, a recent magnetic resonance spectroscopy study showed that persons residing chronically in Salt Lake City (∼1400 m) had lower cerebral ratios of nucleotide triphosphates (including ATP) to total phosphorus and increased ratios of phosphocreatine to total phosphorus compared with controls residing at sea level (Hwang et al., 2019).