Article Serotonin-Norepinephrine Reuptake Inhibitors and the Risk of AKI: A Cohort Study of Eight Administrative Databases and Meta-Analysis Christel Renoux, Lisa M. Lix, Vale´rie Patenaude, Lauren C. Bresee, J. Michael Paterson, Jean-Philippe Lafrance, Hala Tamim, Salaheddin M. Mahmud, Mhd. Wasem Alsabbagh, Brenda Hemmelgarn, Colin R. Dormuth, Pierre Ernst, and the Canadian Network of Observational Drug Effect Studies (CNODES) Investigators Due to the number of contributing authors, Abstract the affiliations are Background and objectives A safety signal regarding cases of AKI after exposure to serotonin-norepinephrine provided in the reuptake inhibitors (SNRIs) was identified by Health Canada. Therefore, this study assessed whether the use of Supplemental SNRIs increases the risk of AKI compared with selective serotonin reuptake inhibitors (SSRIs) and examined the Material. risk associated with each individual SNRI. Correspondence: Dr. Pierre Ernst, Center Design, setting, participants, & measurements Multiple retrospective population-based cohort studies were for Clinical conducted within eight administrative databases from Canada, the United States, and the United Kingdom Epidemiology, Lady between January 1997 and March 2010. Within each cohort, a nested case-control analysis was performed to Davis Research estimate incidence rate ratios (RRs) of AKI associated with SNRIs compared with SSRIs using conditional logistic Institute, Jewish regression, with adjustment for high-dimensional propensity scores. The overall effect across sites was estimated General Hospital, 3755 Cote Ste- using meta-analytic methods. Catherine, Montreal, QC H3T 1E2, Canada. Results There were 38,974 cases of AKI matched to 384,034 controls. Current use of SNRIs was not associated Email: pierre.ernst@ with a higher risk of AKI compared with SSRIs (fixed-effect RR, 0.97; 95% confidence interval [95% CI], 0.94 to mcgill.ca 1.01). Current use of venlafaxine and desvenlafaxine considered together was not associated with a higher risk of AKI (RR, 0.96; 95% CI, 0.92 to 1.00). For current use of duloxetine, there was significant heterogeneity among site- specific estimates such that a random-effects meta-analysis was performed showing a 16% higher risk, although this risk was not statistically significant (RR, 1.16; 95% CI, 0.96 to 1.40). This result is compatible with residual confounding, because there was a substantial imbalance in the prevalence of diabetes between users of duloxetine and users of others SNRIs or SSRIs. After further adjustment by including diabetes as a covariate in the model along with propensity scores, the fixed-effect RR was 1.02 (95% CI, 0.95 to 1.10). Conclusions There is no evidence that use of SNRIs is associated with a higher risk of hospitalization for AKI compared with SSRIs. Clin J Am Soc Nephrol 10: 1716–1722, 2015. doi: 10.2215/CJN.11271114 Introduction diabetic neuropathy, fibromyalgia, chronic low back Antidepressants represent one of the most pre- pain, and osteoarthritis of the knee. Desvenlafaxine, the scribed classes of drugs in developed countries active metabolite of venlafaxine, is indicated for the treat- and a steadily increasing use has been noted, with ment of major depressive disorder only. an estimated prevalence around 10% in the general The reasons for the success of SSRIs and SNRIs, which – population (1 4). The introduction of selective sero- have now superseded older antidepressants as the first- — tonin reuptake inhibitors (SSRIs) as well as the line treatment of depression, include their better toler- newer agents, the serotonin-norepinephrine reup- ability and safety profiles (5). However, a recent safety take inhibitors (SNRIs)—contributed to this in- signal was identified by Health Canada regarding crease. SNRIs are a class of antidepressants that includes duloxetine, venlafaxine, and desvenlafax- SNRIs. A series of 16 spontaneous reports of AKI in fi ine. Aside from its indication for major depressive association with duloxetine were identi ed in the des- disorder, venlafaxine is also indicated in most coun- ignated Medical Event Surveillance of the Canada Vig- tries for the treatment of generalized anxiety disor- ilance database. Moreover, there were 157 reports of AKI der, social anxiety disorder, and panic disorder, associated with duloxetine in the World Health Organi- whereas duloxetine is also indicated for generalized zation (WHO) Adverse Reaction Database. There were anxiety disorder, neuropathic pain associated with also postmarketing reports of AKI with the two other 1716 Copyright © 2015 by the American Society of Nephrology www.cjasn.org Vol 10 October, 2015 Clin J Am Soc Nephrol 10: 1716–1722, October, 2015 SNRIs and Risk of AKI, Renoux et al. 1717 SNRIs, venlafaxine and desvenlafaxine. The biologic mecha- the study period, whichever occurred first. Each individ- nism linking exposure to SNRIs and AKI is unknown, and ual was allowed to enter the cohort only once. there is limited evidence in the cases and in the scientificlit- erature. However, this safety issue should be investigated in Case Definition more depth, because of its severity and the large numbers of The primary end point was defined, using a validated individuals exposed to these drugs in the general population. algorithm (7–9), as hospitalization with a diagnosis of AKI The primary objective of our study was therefore to within 730 days of initiation of treatment with an SNRI or assess whether the use of SNRIs increased the risk of AKI SSRI (diagnosis code for AKI in any of the listed diagnoses; compared with the use of SSRIs. The secondary objective International Classification of Diseases, Ninth Revision, was to examine the risk of AKI associated with each Clinical Modification codes 584, 584.5, 584.6, 584.7, individual SNRI drug. SSRIs were chosen as the compar- 584.8, or 584.9; International Classification of Diseases, ator to help the clinically relevant decision of which Tenth Revision, codes N17, N17.0, N17.1, N17.2, N17.8, antidepressant medication to use as opposed to the ques- or N17.9). tion of whether to prescribe an antidepressant. Control Selection Materials and Methods For each case, up to 10 controls were randomly selected among the cohort members in the risk sets defined by the Study Design case, after matching on age (61 year), sex, and date of We conducted a retrospective population-based cohort 6 study within each of eight databases comprising the cohort entry ( 1 year). The matching on calendar time Canadian Network of Observational Drug Effect Studies permitted control for trends over time in the use of anti- depressant drugs and changes in incidence of the outcome. (CNODES) (6), using a common analytical protocol. fi Within each cohort, a nested case-control analysis was per- Because controls were selected from the risk set de ned by formed to estimate incidence rate ratios (RRs) of AKI as- each case, all controls were necessarily alive and event free sociated with SNRIs compared with SSRIs, and the overall when matched to their corresponding case. The index date effect was then estimated using meta-analytic methods. for each case was the date of hospital admission for the Research ethics approval was obtained from the relevant outcome event, and the index date for the controls was institutions in each site. chosen in order for the controls to have the same duration of follow-up as their matched cases. Data Sources Definition of Exposure The eight CNODES databases used in this study in- The clinically relevant exposure period considered for cluded the health administration databases of the prov- data analysis was a maximum of 2 years preceding the inces of Alberta, Manitoba, Ontario, Nova Scotia, Quebec, index date, and exposure to SNRIs and SSRIs was classified and Saskatchewan, as well as the UK Clinical Practice according to prescriptions issued during this time period Research Datalink (CPRD) and the US MarketScan data- (Figure 1). Current exposure to SNRIs and SSRIs was de- base, corresponding to a total source population of .43 fined as a prescription dispensed in the 60 days before the million patients. index date. A 60-day observation period was chosen based on the assumption of a 30-day duration of prescription with a 30-day grace period. Past recent use was defined Cohort Definition as a prescription dispensed 60–180 days before the index The cohort comprised all patients in each database aged date. Finally, past use was defined as a prescription issued $12 years (or $65 years in Alberta, Nova Scotia, and On- .180 days, but not .730 days, before the index date. Pa- tario) with a first prescription of an SNRI (duloxetine, ven- tients currently exposed to both an SNRI and an SSRI (i.e., lafaxine, or desvenlafaxine) or an SSRI (citalopram, switching between classes within the 60 days before the escitalopram, fluoxetine, fluvoxamine, paroxetine, or ser- index date) were classified according to the last prescrip- traline) between January 1, 1997 (or 1 year after the begin- tion issued during the 60-day current exposure period. ning of data availability) and March 31, 2010. The Among the SNRI and SSRI current exposure categories, indication for an antidepressant was not considered for we also defined categories of switchers and nonswitchers. cohort entry. We defined a new-users cohort by selecting A SNRI switcher was a current SNRI user with an SSRI only patients newly treated with an SNRI or SSRI (i.e., prescription at any time during follow-up including the 60 without any SSRIs or SNRIs dispensed in the previous days before the index date. An SSRI switcher was a current 365 days). Patients were excluded if they had (1) ,1 SSRI user with an SNRI prescription at any time during year of information in the database before the date of co- follow-up including the 60 days before the index date. hort entry, (2) a prescription or dispensation of an SNRI or SSRI in the year preceding cohort entry, or (3) a history of kidney transplantation, dialysis, or AKI in the year pre- Covariates ceding cohort entry.