Mortality Hazard Associated with Prescription Hypnotics

Abstract

Background: The American Cancer Society’s Cancer Prevention Study II was a large survey designed primarily to examine cancer risks such as cigarette smoking. From the same survey and methods, data on usage of “prescription sleeping pills” in 1982 were examined.

Methods: Standardized mortality ratios were computed. Because sleeping pill use could be a proxy for other risk factors, Cox proportional hazards models were computed to control for possible confounding factors as extensively as the data permitted.

Results: Men and women who reported taking prescription sleeping pills 30+ times in the past month had standardized mortality ratios of 3.18 and 2.82, respectively; controlling for 10-year age groups (p < 0.001). The standardized mortality ratios for usage 1–29 times/month were 1.8 and 1.48, respectively (p < 0.001). In proportional hazards models that controlled for 30 other risk factors and comorbidities simultaneously, the excess mortality risk associated with usage 30+ times per month remained significant, but hazard ratios were reduced to 1.35 for men and 1.22 for women.

Conclusions: Use of hypnotics was associated with excess mortality. This methodology could not determine if hypnotic compounds caused the risks associated with their use, nor could the risks of individual compounds be determined. Since millions of Americans are currently taking hypnotics, long-term controlled trials are urgently needed to further guide both patients and physicians.

Introduction

The first Cancer Prevention Study (CPSI) inquired about “sleeping pills” as well as cigarette smoking in questionnaires distributed from 1959–60 by the American Cancer Society Garfinkel 1985. Evidence from this study that cigarette use was a risk factor for cancer and heart disease has been widely discussed. Six-year follow-up also indicated that participants reporting that they “often” consumed “sleeping pills” had about 50% higher mortality than comparable participants reporting that they “never” used sleeping pills Kripke et al 1979. The mortality risk associated with “sleeping pills” in CPSI data seemed robust after control for age, gender, reported hours of sleep, reported insomnia, and prior histories of heart disease, high blood pressure, stroke, or diabetes. Moreover, deaths associated with “often” using “sleeping pills” appeared to be concentrated at night Kripke and Garfinkel 1984. Because of the limited computational capacity available at the time, no further exploration was possible of whether sleeping pill use might be only a proxy or marker for other primary risk factors and illnesses.

At the time of CPSI, barbiturates were the predominant prescription hypnotics. By the time that CPSI evidence of mortality associated with barbiturate hypnotics was published, barbiturates had been largely replaced by benzodiazepines that were believed to be much safer Wysowski and Daum 1991; Solomon et al 1979; NIOM 1979. Additional limitations of CPSI were the imprecise definitions of usage “often” and of “sleeping pills” in the questionnaire, as no attempt had been made to distinguish prescription “sleeping pills” from over-the-counter preparations or tranquilizers. When CPSII was planned, updating the research on cancer risk factors, revised questions were prepared to more specifically reexamine the frequency of using “prescription sleeping pills” as well as “Valium” and “Librium,” in an attempt to distinguish prescription drugs sold as hypnotics (mainly benzodiazepines in 1982) from the most common minor tranquilizers and from over-the-counter hypnotics. Moreover, modern computer techniques made it possible to adjust more extensively for numerous risk factors. We now report analyses of CPSII to explore whether improved questionnaire phrasing and additional control for potential confounding factors eliminated the statistical association of hypnotic usage with reduced survival.