Antidepressants and survival in glioma: Unraveling a complex puzzle

This month in Neuro-Oncology Practice, Edström and colleagues¹ generate new insights to the relationship between antidepressant prescription and survival among patients with malignant glioma. In a retrospective study, the authors used Swedish national population-based registries to study 2 broad questions: (1) the prevalence of antidepressant prescriptions in 1231 glioma patients, compared to 6400 age- and sex-matched controls without glioma; and (2) associations between antidepressant prescription and survival in glioma patients. A subgroup analysis compared survival outcomes among glioma patients prescribed selective serotonin reuptake inhibitor (SSRI) versus non-SSRI antidepressants.

Edström et al. found that the prevalence of antidepressant prescription was significantly higher in patients with glioma (27%) compared to matched controls without glioma (16%). This finding is consistent with earlier data suggesting that depression is more frequent after a glioma diagnosis than in the general population.² There are 2 broad hypotheses as to why depression might be more common after glioma. One is that the psychosocial stress of getting brain cancer is a strong potential trigger. The other broad explanatory hypothesis is that glioma and its consequences may structurally or functionally disrupt neural pathways implicated in mood regulation. As is often the case in psychiatry the truth of the matter may lie somewhere in the middle, with different mechanisms at play in different individuals. Nevertheless, the finding of a higher antidepressant prescription frequency in glioma patients is what one might predict.

The more surprising result was that a new antidepressant prescription after surgery for glioma was associated with significantly poorer survival. In the subgroup analysis, the negative outcome was independent of antidepressant class. A negative effect on survival is surprising because several prior lines of evidence had encouraged the hypothesis that—if anything—antidepressants might have a beneficial effect on survival.

For example, in a retrospective chart review of glioblastoma (GBM) patients, Paul Brown and colleagues reported better 2-year survival statistics for those prescribed an SSRI.³ The association remained on the threshold of statistical significance when corrected for age, recursive partitioning analysis (RPA) class, and extent of resection. A more recent study of GBM patients drawn from insurance records found that patients prescribed Fluoxetine had significantly longer median overall survival when controlling for age, sex, and 6-month baseline pre-GBM comorbidities among other factors.⁴ These retrospective studies support a body of preclinical data suggesting that antidepressants kill glioma cells in culture and in preclinical animal models⁵ and a systematic review identifying depression as an independent predictor of mortality in glioma.⁶ Together these lines of evidence generate the hypothesis that adjunctive treatment with antidepressants might have beneficial effects on glioma patient survival.

It is worth noting, however, that other retrospective chart reviews of glioma patients found no evidence of an association between SSRI prescription and overall survival, in either direction.⁷ Others still reported a negative association between antidepressants and survival in GBM patients starting antidepressants during chemotherapy.⁸ The question therefore remains open and, to date, has generally been explored only by retrospective studies.

The result obtained by Edström and colleagues has several possible explanations. Given the large sample size, and discrepant results among similar studies, it may yet be the case that antidepressants are harmful in patients with glioma. Potentially serious side effects of antidepressants include hyponatremia, bleeding disorders, or prolongation of the corrected Q-T (QTc) interval on the electrocardiogram.⁹ These side effects are however generally rare or of only limited impact in usual clinical practice, and antidepressants are not an established cause of increased mortality in other settings.

The study design allows other possible explanations. Data on functional status could not be included in the model. Because poor functional status is a strong independent predictor of depression in newly diagnosed glioma patients,¹⁰ the result could have reflected differential functional status between the patient groups. It would be enlightening for future studies to control especially for those severely functionally impaired patients who are the most likely to have concomitant depressive symptoms. Other variables which could not be derived from the parent database, but which would aid interpretation, included antidepressant dose, medication adherence, glioma MGMT methylation status, and primary cause of death.

Another uncertainty is how many antidepressant-treated patients in the study had clinical depression. One might assume that it was most of them, but making this diagnosis in practice is often difficult. It can involve nuanced clinical judgments about whether some depressive symptoms (such as low mood) are too severe in the context, and whether others (such as fatigue) are not fully explained by the tumor or its treatment. These ambiguities introduce the possibility for divergent diagnostic thresholds and practices between different doctors. In addition, antidepressants are a first-line treatment for other disorders including various anxiety-spectrum diagnoses and chronic pain. Therefore, in a study like this there is usually room for discussion about the nature of the population to which it might generalize.

From a neuropsychiatric perspective, the biopsychosocial complexity of glioma, and the requirement for conceptual clarity at the intersection with depression, makes it hard to draw final conclusions from observational studies alone. The solution is probably to test specific theories in randomized controlled trials, where diagnoses and prescriptions can be standardized and the many confounding variables can be adequately controlled. Given that antidepressants are extremely widely used in clinical practice, the results of Edström et al. contribute to an intriguing case for equipoise on whether they are indeed beneficial in glioma, be it for overall survival, quality of life, or both.