Multiple medication (polypharmacy) and chronic kidney disease in patients aged 60 and older: a pharmacoepidemiologic perspective

Abstract

Background:

Few studies have examined whether there was an independent association between multiple medication use and risk of chronic kidney disease (CKD), with adjustment for cardiometabolic factors. In the study, we aimed to examine this association using a nationally representative sample in CKD patients aged 60 and older.

Methods:

In the study, subjects aged ⩾60 years (n = 1306) who participated in the 2011–2012 National Health and Nutrition Examination Survey were analyzed cross-sectionally. CKD was defined using the CKD Epidemiology Collaboration (CKD-EPI) equation i.e. estimated glomerular filtration rate (eGFR) < 60 ml/min/1.73 m². Patients with multiple medications were classified as those having five or more prescription medications per day. All data analysis was performed using SAS 9.3 version.

Results:

The prevalence of CKD among age group ⩾80 years, age group 70–79 years and age group 60–69 years were 73.26%, 55.76% and 27.03% respectively (p < 0.001). About half of hypertension (HTN) and diabetic patients aged ⩾60 years had CKD. The prevalence of CKD in patients with cardiovascular disease (CVD) was 60.57%. The logistic regression model without adjustment reflects that those on multiple medications (⩾5 medications/day) had 1.53 (1.02–2.31) times as likely (53% increase) to have CKD compared with those on <5 medications/day. After adjustment for age, CVD, HTN and diabetes mellitus (DM), the odds of CKD for multiple medications appeared to have a protective effect, although it did not reach statistical significance. The adjusted odds ratio [95% confidence interval (CI)] was 0.89 (95% CI: 0.60–1.34); it showed an 11% decreased odds of CKD in patients who were taking multiple medications. The adjusted odds ratio for patients with CVD was 1.38 (95% CI: 0.97–1.98), HTN 1.13 (95% CI: 0.80–1.6), DM 1.78 (95% CI: 1.26–2.51) in age group 70–79 years 3.2 (95% CI: 2.1–4.87) and in age ⩾80 years 6.98 (95% CI: 4.02–12.11) compared with age group 60–69 years old, respectively.

Conclusion:

We did not find significant independent association between use of multiple medications and CKD. The switchover of odds for multiple medication suggested a confounding effect of covariates; further prospective studies are required to find the individualized effect of multiple medications on CKD.

Introduction

The aging of the world population is associated with a significant decrease in the prevalence of communicable diseases and an increase in the prevalence of noncommunicable diseases, partly because of the contribution of better healthcare practices. It is expected that by 2030, one out of five individuals in the US will be aged over than 65, i.e. approximately 70 million people [Stevens et al. 2010a]. Recent data based on the National Health and Nutrition Examination Survey NHANES) suggest that the prevalence of chronic kidney disease (CKD) in the elderly (>65 years) is approximately 38% [Stevens et al. 2010b]. Several population-based studies have suggested that patients with CKD have an increased risk of cardiovascular disease (CVD) morbidities and mortalities. Although CKD has a multifactorial etiology, many cases of CKD have no apparent cause. This has led to controversy about whether a moderate reduction in kidney function in the elderly without significant evidence of kidney damage should be consider a disease or not [Stevens et al. 2010b]. Despite of such controversies, reduced renal function and micro albuminuria in the elderly have adverse outcomes, as proved by several studies [Stevens et al. 2010b].

Recently, CKD has emerged as a serious public health problem especially among the elderly. CKD meets all four criteria to be defined as a public health burden. These criteria are: (i) disease has high prevalence and incidence, increased recently and will increase in near future; (ii) the disease distribution of CKD is unfairly, more common in the elderly; (iii) there is an evidence that upstream preventing strategies reduce the burden; and (iv) such preventive strategies have not been fully implemented [Schoolwerth et al. 2006]. CKD is a chronic progressive disease which progresses with the age. Cardiometabolic factors could make the progression of CKD faster.

To cope with these noncommunicable diseases in the elderly, use of multiple medications is a common medical practice and one which results in greater chances of adverse reactions and drug–drug interactions in the elderly. Today, physicians have a wide array of medication to choose from to treat diseases. Among these, many common agents used in everyday practices could lead to unwanted side effects if used inappropriately in the elderly [Patel, 2003]. The phenomenon of multiple medications is directly related to the number of chronic diagnoses a patient has, such as hypertension, diabetes, obesity and chronic obstructive pulmonary disease; as a result, the prevalence of multiple medication seems to have increased over the past decades [Strehblow et al. 2014]. Multiple medications might have a complicated interaction with foods and nutrients, as well, which could lead to a compromised nutritional status among the elderly [Heuberger and Caudell, 2011]. Aging itself would lead to decreased metabolism and excretion of medication as a result of a physiological decline in renal function. The complex metabolic and excretion pathways of different classes of medication and their interaction with other medication could lead to more adverse reactions and organ damage, if not reviewed carefully. CKD patients need multiple medications to prevent the progression of the disease. Furthermore, as renal function declines, more medications are added to control complications of CKD such as metabolic disorders, bone disorders, anemia, dyslipidemias and cardiovascular diseases [Mason, 2011].

The prevalence of hypertension, diabetes and obesity has increased in the US in the past few years, which also impacts on the prevalence of CKD [Ratliff et al. 2010]. On average, each elderly patient has more than three noncommunicable diseases [Stevens et al. 2010b]. Approximately two out of three elderly people have hypertension, one out of three elderly people is obese and one out of five elderly people has diabetes in the US [Stevens et al. 2010b]. Unfortunately, CKD is commonly underdiagnosed despite widespread use of estimated glomerular filtration rate (eGFR) reporting by laboratories. A special treatment consideration should be provided to patients with CKD and end stage renal disease (ESRD) regarding dosing of medication [Weir and Fink, 2014]. The primary prevention (early diagnosis and treatment) of CKD is of utmost important to reduce the risk of cardiovascular events, ESRD, morbidities and mortalities associated with CKD [James et al. 2010].

‘Multiple medication’ is poorly defined in the medical literature and lacks a standard definition. It often implies a negative situation involving unnecessary medication [Mason and Bakus, 2010]. The elderly usually have more than one noncommunicable disease such as congestive heart failure (CHF), diabetes, glaucoma, depression, hyperlipidemia, hypertension, insomnia, osteoarthritis and dyspepsia. To cope up with them, multiple medication use is a common phenomenon in the elderly. Drug–drug interactions and adverse drug reactions are also common in the elderly. Approximately two out of three elderly people use more than one drug daily [Heuberger and Caudell, 2011] and one-third of prescription drugs are used by the elderly [Patel, 2003]. CKD patients are medically complex to treat and have high risk of adverse reactions. Noncompliance to medication is also a great concern in CKD patients [Mason, 2011]. A recent research study suggested that CKD patients have higher prevalence of inappropriate medication prescriptions, mostly antihypertensives and antibiotics [Jones and Bhandari, 2013].

There are some known predictors for multiple medication, i.e. age, female gender, poor self-reported health, low educational status, dependency on instrumental activities of daily living, and medication disagreement between doctors and patients [Strehblow et al. 2014]. And there are a small number of comorbidities that predispose to multiple medication including diabetes, cardiovascular disease and respiratory disease [Strehblow et al. 2014]. A report on health system safety, entitled To Err is Human: Building a Safer Health System estimated that 45,000–98,000 people died each year due to medication error [Weir and Fink, 2014], a finding which reflects the fine line between safety and efficacy of multiple medication in the elderly. Physicians use their best judgement to identify that fine line to address patient’s health-related issues. Furthermore, a patient with CKD has more chance of drug interactions and adverse reactions due to change in pharmacodynamics and pharmacokinetics [Madero et al. 2008]. A recent population-based study in Taiwan has found an association between duration of multiple medication and acute renal failure (ARF) [Chang et al. 2012]. In the study, we aimed to examine the associations between multiple medication and CKD, with adjustment for multiple comorbidities.

Methods

Study population

We used data from NHANES 2011-2012 (see http://wwwn.cdc.gov/nchs/nhanes/search/nhanes11_12.aspx), which was conducted by the US Centers for Disease Control and Prevention (CDC). NHANES is an ongoing surveillance program designed to evaluate the health and nutritional status of adults and children in the US. NHANES data are used to assess the prevalence of major diseases and risk factors for diseases. The survey uses oversampled subpopulations such as adolescents, Hispanics and African Americans to increase the reliability and precision of estimates of health outcomes and parameters within these groups. A complete description of NHANES guidelines can be found online (see http://www.cdc.gov/nchs/nhanes.htm). NHANES uses a stratified, probability sampling design with oversampling of individuals thought to be at increased health risk. Weights are provided with the public use dataset so that estimates can be made to provide a nationally representative sample of the civilian, noninstitutionalized population of the US. Eligible persons ⩾16 years are interviewed directly and all persons that complete the interview are invited to participate in the medical examination component of NHANES. All subjects provided written informed consent and the protocol was approved by the National Center for Health Statistics (NCHS) Research Ethics Review Board.

The total sample size of the NHANES 2011-2012 cycle was 9756. As CKD is most commonly prevalent among subjects aged ⩾60 years, the present study focused on those aged ⩾60 years old and with available measures of biomarkers to calculate eGFR using the CKD Epidemiology Collaboration (CKD-EPI) definition. In this study, we considered age 60 or more years to stratify age groups and increase sample size. The sample size in the present analysis is 1791. We calculated 1530 subjects’ eGFR based on the CKD-EPI definition. We did not count 224 subjects with missing values of prescription medication. Thus, the final sample size was 1306.

Outcome

The CKD definition is as follows: eGFR <60 ml/min/1.73 m² and/ or kidney damage for 3 or more months. eGFR is a continuous variable. We divided it into two strata: persons with eGFR <60 ml/min/1.73 m² (CKD) and persons with eGFR ⩾60 ml/min/1.73 m² (no CKD). Based on CKD stage definitions developed by the CDC, eGFR has been further categorized in five strata: stage 1 eGFR ⩾90 ml/min/1.73 m²; stage 2 eGFR <90 to ⩾60 ml/min/1.73 m²; stage 3 eGFR <60 to ⩾30 ml/min/1.73 m²; stage 4 eGFR <30 to ⩾15 ml/min/1.73 m²; and stage 5 eGFR <15 ml/min/1.73 m².

CKD-EPI calculation

$$\begin{array}{l}\mathrm{GFR}=141\mathrm{X\; min}\left(\mathrm{Scr}/\kappa ,1\right)\alpha \mathrm{X}\\ \max \left(\mathrm{Scr}/\kappa ,1\right)-1.209\mathrm{X}0.993\mathrm{Age\; X}\\ 1.018\left[\mathrm{if\; female}\right]\mathrm{X}1.159\left[\mathrm{if\; black}\right]\end{array}$$

where Scr is serum creatinine (mg/dl), κ is 0.7 for females and 0.9 for males, α is −0.329 for females and −0.411 for males, min indicates the minimum of Scr/κ or 1, and max indicates the maximum of Scr/κ or 1 [Stevens et al. 2010b].

Exposure variable

There is no a standard definition for multiple medication. In this study, we used a cutoff of five or more medications used per day based on a literature review. We stratified the number of medications used into two categories: individuals who use ⩾5 medications per day and those who use <5 medications per day.

Covariates

It is well known that effect of uncontrolled hypertension and diabetes mellitus damages renal function especially in elderly. Cardiovascular diseases could have direct impact on renal end arteries. The other demographic variables that this paper would take into accounts are age, gender, and race/ethnicity. For this study, we used age ⩾ 60 years to see the effect of multiple medication and covariates on CKD. The inclusion of 60-65 years old helped to categorized age-groups and sample size.

Data analysis

All the data analysis was performed with SAS 9.3 version statistical analytic software. We included weight and cluster NHANES variables to stratify analysis.

Figure 1 shows the analysis conceptual model. It shows that the association between multiple medication and CKD might be confounded by multiple covariates such as age, sex, race and ethnicity, and multiple comorbidities. We applied the t-test, chi-square and unadjusted logistic regression models in univariate analyses and multivariate logistic regression models were used to control covariates. We did not take race into account in the logistic regression model due to the highly insignificant p value in the univariate model. We used a stepwise model and likelihood test selection process to include–exclude variables in the final model.

Figure 1.

Conceptual model demographic factors age, sex, race and ethnicity.

CKD, chronic kidney disease; CVD, cardiovascular disease; DM, diabetes mellitus; HTN, hypertension.

Results

Table 1 shows the frequency distributions across demographics and other covariates by CKD and non-CKD groups. Female gender and non-Hispanic Whites had a higher prevalence of CKD compared with the respective comparison groups. The prevalence of CKD among people with CVD was about 60% compared with those without CVD. One out of two diabetics and hypertensives had CKD. The prevalence of CKD in people receiving multiple medications is 56.72% with a statistically significant p value. Based on Table 1, as age increases the prevalence of CKD increases dramatically, with the age group ⩾80 years having the highest prevalence (73.26%).

Table 1.

Characteristics of person with age group ⩾60 years by CKD and no CKD.

Variables		CKD*		No CKD		p value$
		n	%	n	%
Gender
	Male	335	43.11	442	56.89	0.38
	Female	339	45.02	414	54.98
Race
	Non-Hispanic White	343	51.58	322	48.42
	Non-Hispanic Black	169	41.94	234	58.06
	Mexican American	31	29.81	73	70.19	0.030
	Other Hispanic	68	38.42	109	61.58
	Multiracial	63	34.81	118	65.19
High blood pressure
	Yes	482	50.74	468	49.26	<0.0001
	No	190	32.93	387	67.07
Diabetes mellitus
	Yes	203	54.13	172	45.87	0.034
	No	451	41.07	647	58.93
	Borderline	19	33.93	37	66.07
CVD
	Yes	212	60.57	138	39.43	<0.0001
	No	462	39.15	718	60.85
Multiple medication
	Yes	308	56.72	235	43.28	0.044
	No	315	41.28	448	58.72
Age group	(years)
	⩾60 to 69	216	27.03	583	72.97	<0.0001
	⩾70 to 79	247	55.76	196	44.24
	⩾80	211	73.26	77	26.74

^*CKD is defined as eGFR < 60 ml/min/1.73 m2.

^$p value is retrieved from proc survey frequency function and chi-square distribution.

CKD, chronic kidney disease; CVD, cardiovascular disease; eGFR, estimated globular filtration rate.

Table 2 shows the frequency distributions across demographics and other covariates by CKD stages 1–5 based on the CKD-EPI definition. Overall, there was a highest prevalence of CKD stage 3 (eGFR <60 to ⩾30 ml/min/1.73 m²) across all variables. Table 2 shows that non-Hispanics Whites had a higher prevalence of patients with CKD stage 3 (47.37%) compared with the other racial groups. Among the elderly with CVD, the prevalence of CKD stage 3 was 50.57%. Patients with CKD stage 3 or higher had a higher prevalence of multiple medication compared with those with CKD stage <3. Increased age was significantly associated with a higher prevalence of CKD. The age group ⩾80 years had a higher prevalence of CKD stage 3 (64.58%), stage 4 (7.64%) and stage 5 (1.03%) compared with the other age groups.

Table 2.

Characteristics of person with age group ⩾60 years by CKD stages.

Variables	Stage 1		Stage 2		Stage 3		Stage 4		Stage 5		p value
	n	%	n	%	n	%	n	%	n	%
Gender
Male	64	4.18	378	48.65	306	39.38	22	2.83	7	0.90	0.18
Female	67	8.90	347	46.08	303	40.24	27	3.59	9	1.20
Race
Non-Hispanic White	23	3.46	299	44.96	315	47.37	24	3.61	4	0.60
Non-Hispanic Black	60	14.89	174	43.18	146	36.23	15	3.72	8	1.99
Mexican American	15	14.42	58	55.77	27	25.96	2	1.92	2	1.92
Other Hispanic	11	6.21	98	55.37	59	33.33	8	4.52	1	0.56
Multiracial	22	12.15	96	53.04	62	34.25	0	0.00	1	0.55
High blood pressure
Yes	67	7.05	401	42.21	428	45.05	40	4.21	14	1.47	0.0001
No	64	11.09	323	55.98	179	31.02	9	1.56	2	0.35
Diabetes mellitus
Yes	33	8.80	139	37.07	176	46.93	19	5.07	8	2.13	0.002
No	94	8.56	553	50.36	415	37.80	29	2.64	7	0.64
Borderline	4	7.14	33	58.93	17	30.36	1	1.79	1	1.79
CVD
Yes	11	3.14	127	36.29	177	50.57	27	7.71	8	2.29	0.0001
No	120	10.17	598	50.68	432	36.61	22	1.86	8	0.68
Multiple medication
Yes	33	6.08	202	37.20	269	49.54	32	5.89	7	1.29	0.005
No	64	8.39	384	50.33	292	38.27	14	1.83	9	1.18
Age group (years)
⩾60 to 69	113	14.14	470	58.82	198	24.78	10	1.25	8	1.00	<0.0001
⩾70 to 79	16	3.61	180	40.63	225	50.79	17	3.84	5	1.13
⩾80	2	0.69	75	26.04	186	64.58	22	7.64	3	1.04

CKD stages are defined by eGFR: stage 1 eGFR ⩾90 ml/min/1.73 m²; stage 2 eGFR <90 to ⩾60 ml/min/1.73 m2; stage 3 eGFR <60 to⩾30 ml/min/1.73 m2; stage 4 eGFR <30 to ⩾15 ml/min/1.73 m²; and stage 5 eGFR <15 60 ml/min/1.73 m².

CKD, chronic kidney disease; CVD, cardiovascular disease; eGFR, estimated globular filtration rate.

Table 3 shows that in the unadjusted logistic regression model, multiple medication was 1.53 [95% confidence interval (CI), 1.02–2.31] times as likely (53% increased) to be associated with CKD (p = 0.041). The odds ratio of CKD for patients with CVD was 2.23 (95% CI, 1.5–3.3) times as likely (123% increased) compared with those without CVD (Figure 2). The corresponding odds ratios were 1.86 (95% CI, 1.42–2.43) in HFN patients and 2.07 (95% CI, 1.39–3.09) in patients with DM, respectively. After adjusting for CVD and HTN, the odds of CKD for multiple medication changed to 1.22 (95% CI, 0.78–1.93).

Table 3.

Odds ratio (OR) of chronic kidney disease with and without adjustment for covariates.

Model	Variables*	OR	95% CI	p value
Model 1 Univariate models
Multiple medication$		1.532	(1.018–2.307)	0.0408
CVD		2.229	(1.500–3.313)	<0.0001
HTN		1.856	(1.415–2.434)	<0.0001
DM		2.07	(1.386–3.091)	0.0186
Age group  60–69 years		1	reference
70–79 years		3.385	(2.287–5.010)	0.273
⩾80 years		7.133	(4.218–12.060)	<0.0001
Model 2 adjusted for CVD and HTN
Multiple medication$		1.223	(0.778–1.925)	0.3833
CVD		1.748	(1.185–2.579)	0.0048
HTN		1.384	(0.993–1.930)	0.0551
Model 3 adjusted for CVD, HTN, DM and age
Multiple medication$		0.89	(0.591–1.341)	0.5778
CVD		1.382	(0.967–1.977)	0.0759
HTN		1.13	(0.803–1.590)	0.4839
DM		1.778	(1.259–2.510)	0.0322
Age group  60–69 years		1	reference
70–79 years		3.202	(2.104–4.873)	0.3927
⩾80 years		6.981	(4.024–12.112)	<0.0001

^*CVD, DM and HTN based on medical questionnaire.

^$Defined as five or more medications per day.

CVD, cardiovascular disease; DM, diabetes mellitus; HTN, hypertension.

Figure 2.

Odds ratios with 95% confidence intervals of CKD by variables without adjustment.

CKD, chronic kidney disease; CVD, cardiovascular disease; DM, diabetes mellitus.

Model 3 (Figure 3) accounted for further adjustment for age and DM. The odds of CKD for multiple medication was 0.89 (95% CI, 0.60–1.34] times as likely as (11% decreased) compared with patients without multiple medication, though not statistically significant. Age and DM showed as strong qualitative confounders that reverse the strength of association between multiple medication and CKD. The odds ratio of CKD in subjects aged ⩾80 years was 6.98 (95% CI, 4.02–12.11) compared with those aged 60–69 years.

Figure 3.

Odds ratio with 95% confidence intervals of CKD by variables with adjustment.

CKD, chronic kidney disease; CVD, cardiovascular disease; DM, diabetes mellitus; HTN, hypertension.

Discussion

The prevalence of CKD in the elderly is approximately 40% and growing day by day. Medication in the human body is metabolized by the liver and the kidney. The two main routes of excretion of medication are the gastrointestinal (GI) tract and urinary tract. The measure of eGFR defines renal function and perfusion. The renal arteries are the end arteries that reflect highly sensitive vessels for any cardiovascular changes into the body. HTN, DM and CVD affect renal blood vessels and disrupt the renal function of the excretion of waste material. Certain medications have directly nephrotoxic effect such as nonsteroidal anti-inflammatory drugs (NSAIDs) which are available over the counter (OTC).

Multiple medication use is a common in the elderly due to multiple comorbidities such as HTN, DM and CVD. In our present study, the unadjusted odds ratio appeared to have a significant and positive association between multiple medication use and CKD. However, after adjustment for age, CVD, HTN and DM, this association became negative although it did not reach statistical significance. It suggests that the adjusted covariates play a pivotal role in the association between multiple medication and CKD.

The major limitation of this study is that its cross-sectional nature gives only a snapshot (prevalence). It helps us to generate a further evidence-based hypothesis and it is not necessary to interpret any cause–effect association. Another limitation is that the study does not include OTC medication usage in the elderly. NSAIDs are a common OTC medication class which has nephrotoxic effects. We did not able to adjust the analysis by the class of medications based on the limitations of data availability and sample size issue. These inclusions might change the point estimates and strength of association between the main predictor and outcome of this study. It is always hard to check interactions between variables such as interaction between age and prescription medication, etc., partly because of the current relative small sample size. It is hard to quantify the dose of prescription medication usage daily in the elderly, which is why the present analysis is based on the overall number of medications used.

The main advantage of this study is that NHANES represents the US national population. NHANES has the advantages of laboratory measurements with a formal standard approach, which is conducted using highly organized and professionalized methods. The calculation of eGFR using the CKD-EPI equation is possible for such dataset, giving more confidence in defining outcome or predictor variables. The medical questionnaire data used in NHANES helped to control for confounders such as HTN, DM and CVD in the study. The NHANES data helped to test the hypothesis in this cross-sectional study.

Conclusion

It is always hard to determine a specific association between exposures and outcomes due to the multifactorial nature of CKD. The findings of the present study suggest that there may be the possibility of a reverse causality or qualitative effect modification in CKD patients on multiple medication. Multiple medication is essentially used to control cardiometabolic conditions like HTN, DM and CVD among the elderly, and might serve as a protective role in the progression of CKD by controlling these cardiometabolic comorbidities. We did not find a significant independent association between multiple medication and CKD. The switchover of odds for multiple medication suggested confounding effect of covariates, requiring further prospective studies to identify the individualized effect of multiple medication on CKD.