Introduction
Chronic diseases, acute events, and the need for symptom management increase with age, which is reflected in the concurrent intake of multiple medications [
1]. This phenomenon is commonly referred to as polypharmacy and is especially frequent in older adults [
2]. However, there is no official consensus on the definition of polypharmacy [
3]. The most common definition is the use of at least five medications, often without further specification as to prescription requirement (i.e. prescription or over-the-counter (OTC)) or pattern of intake (i.e. regular or on-demand). Polypharmacy itself is associated with adverse events such as falls, hospital admissions, and mortality [
2].
The prevalence of frailty also increases with age [
4]. Frailty is described as a biological syndrome with accelerated decline in physiological reserves and resilience to stressors, also resulting in increased risk of similar adverse outcomes such as risk of falls, hospitalization, disability in activities of daily living, need of nursing home, or mortality [
5,
6]. Studies analyzing cross-sectional data mostly found an association between polypharmacy and frailty but the temporal relationship between both remains unclear [
7]. Longitudinal studies on polypharmacy and incident frailty have been inconclusive [
8‐
13] and sparse [
14]. A study from Germany showed that in older adults with a mean age of 70 years, polypharmacy was associated with a 1.5 odds for incident frailty [
9].
Both the incidence of frailty as well as polypharmacy are associated with another globally prevalent health burden affecting older adults, namely chronic kidney disease (CKD) [
15‐
17]. Prevalence of CKD varies from one third to two thirds in the adult population over the age of 75 years in Europe [
18]. Older individuals with CKD often have a high prevalence of comorbidities and therefore polypharmacy is inevitable [
19,
20]. The complex treatment regimens in individuals with CKD for example increase the potential for adverse drug-drug interactions and consequently, adverse side effects [
15,
21].
In order to disentangle the unclear relationship between polypharmacy, frailty, and CKD, we hypothesize that in very old adults (1) different definitions of polypharmacy alter the effect estimation on incident frailty and (2) that CKD modifies the effect of polypharmacy on incident frailty.
Discussion
In this cohort of older adults with a mean age of 82.9 years, about 40% were taking at least five prescribed drugs regularly. Among those with regular prescription polypharmacy and hyperpolypharmacy, the adjusted OR of incident frailty within the next two years were 1.96 (95% CI 1.20–3.19) and 2.87 (95% CI 0.85–9.69) compared to individuals without polypharmacy respectively. Using extended polypharmacy definitions, the association with incident frailty was attenuated for polypharmacy. The effect of polypharmacy on incident frailty was modified by CKD on an additive scale (RERI 1.56; 95% CI 0.01, 3.12).
The polypharmacy prevalence of 40% found in our study is comparable to the prevalence in the age group ≥ 75 years in many European countries as assessed by the Survey of Health, Ageing, and Retirement in Europe (SHARE) [
37]. The most prevalent medication categories at the therapeutic subgroup level are similar to those reported for a community-based cohort study in the United States and across multiple European countries, likely because they reflect the major chronic diseases in older adults [
13,
38]. This probably explains why the top ten medication groups and the five most prescribed substances within the groups did not differ between polypharmacy categories although the prevalence of each substance varied across categories.
A meta-analysis on global incidence of frailty among community-dwelling older adults found that among non-frail individuals who survived a median (IQR) of 3.0 (1.0-11.7) years, 13.6% became frail [
39]. In our study, 16.2% (105 of 649 participants who survived the follow-up period) became frail within a median follow-up of 2.1 years. Since it has been shown that frailty incidence increases with age [
9], the higher cumulative incidence in our study could be explained by the older age of our study participants.
Longitudinal studies that investigated the association of polypharmacy and incident frailty in older adults were inconclusive. Some found that polypharmacy increased the risk of incident frailty [
8,
9,
12] while others demonstrated no association [
10,
11,
13]. One possible explanation may be different definitions of polypharmacy. Our results show that older adults with polypharmacy have almost double the odds for incident frailty compared to individuals without polypharmacy. The estimated effect decreases when the definition of polypharmacy is expanded to include on-demand drugs, OTC drugs, vitamins and supplements and the effect becomes more prominent in the hyperpolypharmacy group as the definition expands. Our results using the active substance polypharmacy definition are comparable to another study that used that same definition but with younger individuals (mean age 70 years) and three years of follow-up [
9]. Another study with a two-year follow-up of older men (mean age of 77 years) used our primary polypharmacy definition (regular prescription) and found for individuals with hyperpolypharmacy a 2.5 odds of incident frailty [
8] which is comparable to our findings. Studies that did not find an association between polypharmacy and incident frailty used polypharmacy definitions including a lower cut-off (three medications and not five) [
11], assessed medication on a linear scale [
10], or used an all polypharmacy definition and a longer follow-up [
12,
13]. Shmuel et al. additionally included a regular and on-demand polypharmacy definition and also found a non-significant OR of 1.4 (95% CI 0.9-2.0) of incident frailty for individuals with polypharmacy comparable to our study (OR: 1.46; 95% CI 0.89–2.40) [
13]. Thus, when investigating the risk of polypharmacy on incident frailty in older adults, it is important to note that the use of different definitions of polypharmacy may result in different effect estimates of the impact of polypharmacy on incident frailty.
Since the association of polypharmacy and incident frailty remained significant after adjusting for comorbidities, other possible pathways independent of comorbidities may exist in which polypharmacy could lead to frailty. (1) Overall, polypharmacy increases the risk of taking potentially inappropriate medications, adverse drug events, and low adherence, all of which are also associated with frailty [
40]. (2) Polypharmacy and specific drug classes (e.g., acetylcholinesterase inhibitors and HMG-CoA reductase inhibitors) can lead to weight loss, malnutrition, and sarcopenia through alterations in taste, intestinal absorption and metabolism, or elimination of vitamins and minerals, which in turn reflect important components of the fraily phenotype [
5,
41]. (3) Another important aspect of the frailty phenotype is physical activity. It has been shown that an increasing number of medications and polypharmacy is associated with decreased physical activity [
42]. This pathway operates probably via specific medications such as statins which are known to be associated with myalgias that could lead to less physical activity [
42]. (4) It has also been shown that polypharmacy is associated with slowness, a third component of the frailty phenotype [
43].
As CKD is common in old age, it is frequently accompanied by the intake of several drugs and has also been associated with incident frailty [
16,
44]. Therefore, we also investigated CKD as a modifier of the effect of polypharmacy on incident frailty. Our results provide strong indications that the estimated effect of polypharmacy on incident frailty is modified by CKD on the additive scale. It has been argued that analyzing the biological interaction on an additive rather than a multiplicative scale is the appropriate approach in public health [
34]. This implies that the combined effect of polypharmacy and CKD on incident frailty is larger than the sum of the individual effects [
45]. A possible explanation for the effect modification could be that several potential pathways in which polypharmacy contributes to frailty are similar to those in which CKD can lead to frailty. For example, as described before, polypharmacy can lead to malnutrition [
41]. Malnutrition is common in individuals with CKD and may also further decrease kidney function and lead to frailty worsening [
46]. Another possible explanation could be that in older adults the drug metabolism and clearance may change especially in individuals with reduced kidney function and the risk of adverse drug reactions is higher in older individuals with CKD [
47].
This emphasizes that the consequences of polypharmacy especially in old age are multifaceted. One part of the problem is certainly that guideline treatment decisions are often based on results from clinical trials where older adults, especially with multimorbidity are excluded [
48]. Thus, the grounds on which guideline-adherent treatment decisions were made did not include the population in which they are then applied [
49]. Furthermore, guidelines are often focused on a single disease [
48]. When multimorbidity is treated in older adults this in turn contributes to polypharmacy. Although polypharmacy is recognized as a risk factor for adverse events, it is very prevalent in older adults [
50]. Another aspect that should be considered in older adults are the trade-offs between future risk reductions and the potential current risks for adverse events due to polypharmacy [
51]. Both the general practitioner and the older patient have to balance risk and benefit from deprescribing carefully [
52,
53], particularly since studies showing improvement in clinical outcomes are scarce [
2]. A positive development in terms of clinical outcomes are the new drugs such as sodium-glucose cotransporters and steroidal mineralocorticoid receptor antagonists for reducing albuminuria. These could also have a preventive impact, e.g. lowering the risk of frailty.
Our study has several strengths. We have a very old study population with a longitudinal design that is phenotyped in much detail which enabled us to investigate incident frailty. Due to the complimentary linkage of primary (BIS) and secondary (AOK claims) data, it was possible to have available information on an extensive range of different valid health indicators. In addition, a comprehensive assessment of all medications, including packaging and medication schedules, was conducted. This allowed the inclusion of not only prescription but also OTC drugs. This detailed medication phenotyping of participants added to the strength of the study and enabled different polypharmacy definitions to be compared. Furthermore, we applied the methodological approach proposed by Knol and VanderWeele to analyze and demonstrate effect modification not only on the multiplicative but also on the additive scale to assess its public health importance [
33].
The study has some limitations. The assessment of polypharmacy was based on self-reported medication use which is prone to recall bias, supported by medication plans and packages when available. However, we could not determine the number of pills, the dosage, or the extent of medication adherence. We also did not consider whether medications were appropriate for the severity and progression of disease, although this may be important in assessing frailty risk [
8]. We did, however, consider the number of morbidities using a comorbidity score (CCI) which includes weights for comorbidity severity but does not take individual disease progression into account. Adjustment for comorbidities also served to address confounding by indication. Although we adjusted for multiple confounders we cannot exclude residual confounding. On the other hand, the research question addressed in this study could not have been answered in a randomized controlled trial, as it is neither ethically acceptable to assign an individual to polypharmacy if it is not indicated nor is it even possible to initiate CKD in individuals and consequently randomly assign them. A further possible limitation could be that we only assessed CKD at baseline; however, our CKD definition was consistent with the KDIGO aspect on chronicity [
29] in at least 97% of the participants, thus technically justifying the use of the term CKD. Another aspect is loss to follow-up during the observation period that could have led to selection bias. The loss to follow-up seems to be independent of the exposure since it is evenly distributed across the exposure categories. Frailty incidence is cumulative as we analyzed frailty over two time points. Since frailty is a dynamic process, we cannot exclude that we missed transitions e.g. from frailty to non-frailty / prefrailty.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.