NOTE: Dr. Kripke is responding to a study by Elisabetta Patorno, Robert J Glynn, Raisa Levin, Moa P Lee, and Krista F Huybrechts entitled Benzodiazepines and risk of all cause mortality in adults: cohort study. To date, Dr. Kripke’s questions have not been addressed.
Patorno and colleagues must be admired for heroic use of propensity scores to control potential confounders in their massive comparison of cohorts identified by benzodiazepine prescription initiation versus no initiation in a small time window. The results give the appearance that benzodiazepines had only small associations with increased mortality that might have been due to uncontrolled confounding factors. But is there a serious flaw?
“Intention to treat” designs are admirable in randomized placebo-controlled clinical trials wherein most of the treated group is fully treated and all of the placebo group gets only placebo. This current report gave us no information concerning how much benzodiazepine (e.g., mean defined daily dosage) was prescribed for the benzodiazepine “initiators” versus the “non-initiator” cohort during the intervals from study initiation to end of follow-up. The report “disregarded treatment variations occurring during the follow-up for both cohorts.” One wonders if many of the benzodiazepine-initiators actually received very little benzodiazepine during the follow-up intervals, whereas many of the non-initiators might have received substantial benzodiazepine doses during the follow-up.
Would the authors please supply the benzodiazepine consumption data for the contrast groups during each follow-up interval? May we have dose-response mortality data, please?
Until we know the benzodiazepine consumption data, this report should not be compared with Kripke et al. and Weich et al. (their references 15 and 16). In those two studies, dose-response effects were observed for the benzodiazepine user cohorts, but zero benzodiazepine was prescribed for the non-user cohorts throughout the follow-up intervals.
There is a second important difference between our report and the current one. We studied hypnotics, not benzodiazepines in general. Moreover, 62% of our hypnotic cohort received short-acting benzodiazepine agonists, mainly zolpidem, that were NOT among the drugs included by Patorno et al. In an earlier report, with American Cancer Society colleagues, we observed a distinct difference in mortality between “prescription hypnotics” and diazepam or chlordiazepoxide.2 Of all the drugs studied, eszopiclone appeared to have the highest mortality risks. I suspect that some of the mortality risk of hypnotics comes from taking rapidly-absorbed drugs at bedtime,1 whereas long-acting benzodiazepine tranquilizers taken for anxiety or mood disorders during the day may produce tolerance for nocturnal respiratory suppression. The difference in hazard ratios between short-acting and long-acting benzodiazepines observed by Patorno et al. may support this.
1. Kripke DF. Hypnotic drug risks of mortality, infection, depression, and cancer: but lack of benefit [version 2]. F1000Research. 2017;5: 918 ( doi: http://dx.doi.org/10.12688/f1000research.8729.2 ).
2. Kripke DF, Garfinkel L, Wingard DL, Klauber MR, Marler MR. Mortality associated with sleep duration and insomnia. Arch Gen Psychiatry. 2002;59(2): 131-6.
Competing interests: Dr. Kripke is a long-time critic of hypnotic risks, for example, at his non-profit web site www.DarkSideOfSleepingPills.com
