Prescription Opioid Duration, Dose, and Increased Risk of Depression in 3 Large Patient Populations

The study by Scherrer et al. (1) posits a causal association between opioid prescribing of greater than one month duration and subsequent new onset of depression, occurring on average 3.4 years after the opioid prescription . While we agree that the association is supported by the analysis and could be consistent with a biological effect of opioids, we urge caution in the causal and clinical interpretation of this finding.

Unmeasured and residual confounding, acknowledged by the authors to be potential problems, may be more plausible explanations for the reported association. Adjustment for chronic disease diagnoses and pain scores are not likely to fully account for the severity of functional effects of chronic diseases, which could be associated with both opioid prescribing and depression. The authors acknowledge potential confounding due to possible associations between initially unreported depressed mood and opioid requests, and between that early depressed mood and later diagnosed depression. This also seems a plausible alternative explanation of the results. Further, the authors attempted to deal with potential detection bias by controlling for quartiles of the average number of visits per month. If within one or more quartiles those with longer opioid use have substantially more visits than those with shorter duration use, detection bias could still be a significant issue.

Finally, the incidence of depression in patients with longer opioid use does not seem unusual when compared with incidence in epidemiological studies of patients with many comorbidities. Incident rates in some diabetic populations (2) are 16.5% at 2 and 5 year follow up, and community-dwelling individuals 55 and over develop depression at rates of 11-19%, (varying by income). (3) Investigating incident depression for a propensity score matched group that had no opioid prescription during the same period would have been useful. If the rates of depression were similar for those with longer term opioid use and no opioid use, then short term use would be a marker for some factor that reduces the risk of depression.

Telling patients that there is an association between opioids and depression seems premature, even though the potential underlying neurological effects of long term opioid use are plausible at least within short time frames. Nevertheless, caution in opioid prescribing for chronic pain, as well as monitoring of depression for all adult patients is good clinical practice.

References
1. Scherrer JF, Salas J, Copeland LA, Stock EM, Ahmedani BK, Sullivan MD, Burroughs T, Scvhneider FD et al. Prescription opioid duration, dose and increased risk of depression in 3 large patient populations. Ann Fam Med. 2016; 14: 54-62.
2. deJong, P, Roy JF, Saz P, Marcos G, Lobo A, ZARADEMP Investigators. Prevalent and incident depression in community-dwelling elderly persons with diabetes mellitus: results from the ZARADEMP project. Diabetologia. 2006;49:2627-33.
3. Koster A, Bosma H, Kempen G, Penninx B, Beekman A, Deeg D, van Eijk J. Socioeconomic differences in incident depression in older adults: The role of psychosocial factors, physical health status, and behavioral factors. J Psychosom Res. 2006; 61:619-27.

Competing interests: None declared

Stefan G. Kertesz raises an important question that our research design is unable to test with certainty. Did patients have symptoms of depression even when they did not have a diagnosis in the medical record? Long term opioid use may worsen symptoms of depression and patients with depression are more likely than those without to be longer term users of opioids. Thus the pathway to new onset depression may be through worsening of depression symptoms during chronic opioid treatment. To overcome the limitations of our retrospective cohort design, we will need to collect data prospectively. We are planning to collect lifetime depression histories and then follow a cohort prospectively with repeated measures of depression and other symptoms of mood disorder and risk factors for mood disorder. We hope to be able to conclude if our observations represent a causal relationship or if other mechanisms explain the opioid-depression association.

Competing interests: None declared

This study is an extremely well-devised analysis of 3 databases. It should guide clinicians to discuss this issue as a possible risk that may apply to patients considering opioid therapy. The authors are careful in avoiding strong causal inferences. I wish to ask two questions that may strengthen or weaken the causal inference.

First, do the authors believe latent and undiagnosed depression was likely present in the analytic sample at inception, despite the good-faith intent of the authors to remove such individuals? If so, could I suggest a sensitivity analysis?

The standard for inclusion in this manuscript was at least one outpatient clinic visit of any kind in each of 2 years. The value of requiring prior utilization of care is that it enabled the researchers to attempt to exclude persons with pre-existing depression. A natural question is whether some persons with prior depression or subsyndromal depressive syndromes (such as dysthymia) might have crept into the sample, despite the best efforts of the research team.

If a person had longstanding depressive symptoms and appeared just once for primary care in each of two years, perhaps for care of hypertension or an upper respiratory tract infection, it seems possible that the latent depression might not have been detected in the course of a 15 minute visit by a primary care provider (if primary care providers were in fact responsible for the bulk of the prior service use in each of the 3 samples).

This speculation is somewhat buttressed by studies showing that primary care providers detect less than 50% of depression in primary care(1) and roughly 1/3 of mild depression in primary care(2).

A sensitivity analysis that would counter this concern might be as follows: repeat the analysis for a restricted sample of heavy service utilizers with many clinical encounters (perhaps above the median number?), where the probability of such depression being picked up in the pre-analysis time period would be much higher. If the association between opioid dose or duration and incident depression remains strong, then the stated concern can be partially allayed.

A separate concern is that at least in the VHA sample, 93.2% of incident depression episodes began after the end of opioid use, most being greater than 6 months later. The authors remind us that a delayed diagnosis of depression is so common that for some of these individuals the depression may well have begun far earlier. That said, can the authors find any signal (perhaps qualitatively) that depression beginning earlier in time might be tied to higher doses? Admittedly, this would be a speculative article.

This is an excellent paper, in any case, and extremely important.

Stefan G. Kertesz, MD, MSc, Associate Professor, University of Alabama at Birmingham School of Medicine. Birmingham VA Medical Center. Birmingham, AL

Opinions stated here are those of the author and do not represent positions or views of any agency of the United States federal government, including the US Department of Veterans Affairs.

References
1. Mitchell AJ, Vaze A, Rao S. Clinical diagnosis of depression in primary care: a meta-analysis. Lancet. 2009;374(9690):609-619.
2. Mitchell AJ, Rao S, Vaze A. Can general practitioners identify people with distress and mild depression? A meta-analysis of clinical accuracy. J Affect Disord. 2011;130(1-2):26-36.

Competing interests: None declared

Dr. David Smith expresses concern that our research on the association between chronic opioid use and depression is another threat to access to these important medications in pain control. Patients and providers should be aware of all pros and cons before initiating a medication and our research supports discussing depression. In addition to common practice of screening for depression at initiation, our study supports repeated screening for depression during the course of opioid therapy. Certainly not all patients will develop depression. However new onset depression certainly complicates pain management, not to mention the burden of the disease itself. It is reasonable to consider all risks and benefits and to suggest more frequent screening to detect depression and consider alternative treatments or taper. Chronic opioid use may benefit some but there is no strong evidence to support benefits of long term use for chronic pain. While we did not study overdose and abuse, I would argue that these consequences alone are sufficient to call for significant funding of research to find alternative pain medications. Risk of depression may contribute to this urgency which would be a welcomed, but not designed, outcome of our study.

Competing interests: None declared