Polypharmacy and mortality: new insights from a large cohort of older adults by detection of effect modification by multi-morbidity and comprehensive correction of confounding by indication

The objective was to investigate whether the association of polypharmacy with non-cancer mortality is independent from comorbidity and is not a result of confounding by indication. Analyses were conducted in 2687 participants of a German, population-based cohort of older adults with data collection 2008-2010. Polypharmacy was defined as >= 5 drugs and hyperpolypharmacy as >= 10 drugs. Drugs without relevant propensity of causing adverse drug reactions or drug-drug interactions were not counted. Confounding by indication was addressed by model adjustment for a propensity score for polypharmacy. The median age of study participants was 70 years, 10.7% had multi-morbidity, and 47.4% took five drugs or more (8.6% took >= 10 drugs). During 4.4 years of follow-up, 87 participants died of a cause other than cancer. Statistically significant, more than twofold increased non-cancer mortality was observed for subjects with polypharmacy or hyperpolypharmacy in a model adjusted for age, sex, education, lifestyle variables, and comorbidity, but associations lost statistical significance after additional adjustment for a propensity score for polypharmacy. However, a significant interaction of hyperpolypharmacy and multi-morbidity was detected (p = 0.019). The hazard ratio for the association of hyperpolypharmacy with non-cancer mortality was 1.42 (95%CI 0.57; 3.57) in subjects without multi-morbidity and 0.51 (95%CI 0.11; 2.27) in subjects with multi-morbidity. Polypharmacy was not independently associated with non-cancer mortality. This study highlights the importance to adjust for confounding by indication in studies on polypharmacy by a propensity score. The detected interaction suggests that hyperpolypharmacy can be indicated in subjects with multi-morbidity and may only be harmful in subjects without multi-morbidity.