Article Text
PDF
Abstract
Background Non-steroidal anti-inflammatory drugs (NSAIDs) are associated with an increased risk of heart failure. NSAIDs inhibit the synthesis of renal prostaglandin, which results in a higher total blood volume, cardiac output and preload. The association between recent start of NSAIDs in elderly people and echocardiographic parameters was investigated.
Methods In the Rotterdam Study, a population-based cohort study, the effect of NSAIDs on left ventricular end-systolic dimension, left ventricular end-diastolic dimension, fractional shortening and left ventricular systolic function was studied in all participants for whom an echocardiogram was available (n=5307). NSAID use was categorised as current NSAID use on the date of echocardiography, past use and never used before echocardiography during the study period. Current use was divided into short-term NSAID use (≤14 days) and long-term NSAID use (>14 days). Associations between drug exposure and echocardiographic measurements were assessed using linear and logistic regression analyses.
Results Current NSAID use for <14 days was associated with a significantly higher left ventricular end-systolic dimension (+1.74 mm, 95% CI 0.20 to 3.28), left ventricular end-diastolic dimension (+3.69 mm, 95% CI 1.08 to 6.31) and significantly lower fractional shortening (−6.03%, 95% CI −9.81% to −2.26%) compared with non-users. Current NSAID use for >14 days was associated with a higher left end-diastolic dimension (+1.96 mm, 95% CI 0.82 to 3.11) but there was no change in the other echocardiographic parameters.
Conclusion This study is the first to investigate the association between NSAIDs and echocardiographic parameters and suggests that there is a transient effect of short-term use of NSAIDs on the left ventricular dimension and function of the heart.
- Non-steroidal anti-inflammatory agents
- echocardiography
- elderly
- heart failure
- population-based cohort study
- echocardiography (three-dimensional)
- old age
- nsaids
- epidemiology
Statistics from Altmetric.com
- Non-steroidal anti-inflammatory agents
- echocardiography
- elderly
- heart failure
- population-based cohort study
- echocardiography (three-dimensional)
- old age
- nsaids
- epidemiology
Introduction
The incidence and prevalence of heart failure is increasing in the elderly population.1 Heart failure has a high morbidity and mortality. Despite improvements in the treatment of cardiovascular disease and myocardial infarction over the past 50 years, the 5-year survival rates are 25% for men and 38% for women.2–4 Besides known direct and indirect risk factors for heart failure such as hypertension, diabetes mellitus, cigarette smoking, high body mass index, male gender, low education, physical inactivity, valvular heart disease and coronary heart disease,5 there are several drugs which are associated with an increased risk of heart failure. Non-steroidal anti-inflammatory drugs (NSAIDs) are known for their sometimes adverse influence on cardiovascular homeostasis,6 and have been associated with an increased risk of induction and exacerbation of heart failure.7 8 Furthermore, the literature shows that concomitant use of NSAIDs and diuretics is associated with a twofold higher risk of hospitalisation for heart failure.9 NSAIDs inhibit the synthesis of renal prostaglandin10 which results in a lower renal blood flow and subsequently in increased renal sodium reabsorption and water retention.11 In addition, inhibition of renal prostaglandins stimulates the vasopressin effect in collecting tubules which results in water retention.10 12 The additional water retention results in a higher total blood volume and thus a higher cardiac output and preload. The short-term effect of NSAIDs on cardiac function in the elderly population is not known.
The objective of this study was to investigate whether short-term NSAID use is associated with impaired echocardiographic parameters in a population of community dwelling elderly people.
Methods
Setting
Data were obtained from the Rotterdam Study, a large prospective population-based cohort study which has been described in detail elsewhere.13–15 In brief, all inhabitants of Ommoord, a suburb of Rotterdam, The Netherlands, aged ≥55 years were invited to participate, of whom 7983 (78%) agreed. All participants were examined in detail at baseline during the period 1991–3 and have been followed since then. They were interviewed at home by trained interviewers and during two subsequent visits at the research centre, and underwent laboratory assessments, clinical examinations and imaging procedures. Follow-up examinations took place every 3–4 years. In 2000–1 the cohort was extended with 3011 participants aged ≥55 years from the same suburb. The study population consisted of all participants in the Rotterdam Study for whom an echocardiogram was available (n=5307).
Exposure to NSAIDs
Computerised pharmacy records were available for all participants in the Rotterdam Study at 1 January 1991. All prescriptions dispensed to participants in the Rotterdam Study by the seven automated pharmacies in the study area are continuously collected in a database. Drugs were coded according to the Anatomical Therapeutic Chemical Classification (ATC code).
Detailed information was available for the filing date, product name, dosage, number of filled tablets or capsules and the regimen. Subjects with a prescription for NSAIDs were identified. For each participant of the study population, exposure to NSAIDs (ATC code M01A) according to registration in the pharmacy was assessed on the day of their echocardiography.
The method of analysing drug exposure has been published elsewhere.16 In line with this, the use of NSAIDs was divided into three mutually exclusive categories: current NSAID use on the day of echocardiography, past use and no use during the study period. Current users were divided into participants with NSAID use during echocardiography who had started <14 days before echocardiography and participants who were current users of NSAIDs on the date of the echocardiography and who had used the drug for >14 days on that date. Past users did not use NSAIDs on the index date but had used them in the past during the study period. Participants who did not use NSAIDs during the study period were considered as the reference group. The dose was expressed in Defined Daily Doses (DDD), defined by the World Health Organization as the recommended daily dosage for an adult for the main indication.17
Echocardiographic outcome parameters
Left ventricular end-diastolic dimension (LVED), left ventricular end-systolic dimension (LVES), left ventricular systolic function (LVSF) and fractional shortening (FRAC) were analysed. LVED and LVES were measured in the parasternal long axis view using M-mode with two-dimensional guidance. FRAC at the endocardium was calculated as (LVED – LVES)/LVED × 100%. Global LVSF was qualitatively assessed, without quantitative measurement, from the two-dimensional echocardiogram and classified as normal, fair, moderate or poor.18 LVSF is a binary variable of which the normal group was coded as 0 and the fair, moderate and poor group as 1.
Covariables
The following covariables were included in our model: gender, age at the time of the echocardiography, and whether heart failure was present or absent before the date of the echocardiography. We considered potential confounding by use of antihypertensive drugs, low ceiling diuretics, high ceiling diuretics, β blockers, calcium antagonists and ACE inhibitors on the day of echocardiography. Furthermore, body mass index, past use of NSAIDs and calculated modification of diet in renal disease (MDRD, renal function calculated using serum creatinine and age at echocardiography)19 were considered potential confounders. The two Rotterdam Study cohorts were analysed together with adjustment for cohort membership to take cohort effects into account.
Data analysis
Differences in baseline characteristics between current NSAID users, past users and non-users were tested using a χ2 test for dichotomous variables and a t test for continuous variables. Linear regression analysis was used to analyse the association between current NSAID use on the date of echocardiography and echocardiographic parameters. To analyse the effect on LVSF, binary logistic regression was used. Dose-effect relationships were analysed within the exposed group only with the lowest dose as a reference. Current NSAID use was divided into four categories: ≤1 DDD for up to 14 days before the date of echocardiography, >1 DDD for up to 14 days before the date of echocardiography, ≤1 DDD for >14 days before the date of echocardiography and >1 DDD for >14 days before the date of echocardiography. Alternatively, we assessed by goodness of fit using F tests whether a dose-effect relationship fitted better than a fixed dose model. We adjusted for covariables which changed the effect estimate by more than 10%.20 Interaction of NSAIDs and renal function was tested by adding an interaction term to the model. SPSS Version 15.0 was used and p values were considered significant if <0.05.
Results
Of the 5307 participants, 2249 (42%) were male. Two hundred and sixty-one participants (5%) were current users of NSAIDs on the date of echocardiography, 92 of whom had started NSAID use <14 days before the date of echocardiography and 169 had been using NSAIDs for >14 days at the echocardiography date. High ceiling diuretics were used in 281 participants (5%). In both groups of individuals who were currently using NSAIDs at the date of echocardiography, 8% had a history of heart failure. For 4209 participants it was possible to calculate the MDRD. The general characteristics of the participants on the date of echocardiography are shown in table 1.
Characteristics of the Rotterdam Study population (on date of echocardiography)
Compared with those who were not current NSAID users, current NSAID users who started <14 days before their echocardiography showed a significantly increased LVED and LVES (1.56 mm (95% CI 0.08 to 3.12) and 3.59 mm (95% CI 0.96 to 6.22), respectively). After adjustment for cohort and present heart failure before the date of echocardiography, the effect of short-term NSAID use increased to 1.74 mm (95% CI 0.20 to 3.28) and 3.69 mm (95% CI 1.08 to 6.31) for LVED and LVES, respectively. FRAC was −6.09% (95% CI −9.90% to −2.29%) before and −6.03% (95% CI −9.81% to −2.26%) after adjustment compared with those who were not current NSAID users (table 2). The ORs for LVSF for current NSAID users who started <14 days before their echocardiography were compared with the ORs for those who were not current NSAID users (0.94 (95% CI 0.60 to 1.45) before and 0.94 (95% CI 0.60 to 1.46) after adjustment). Participants who were current users of NSAIDs on the echocardiography date for >14 days had a significantly increased LVED of 1.94 mm (95% CI 0.78 to 3.10) before adjustment and 1.96 mm (95% CI 0.82 to 3.11) after adjustment. However, FRAC and LVES were hardly changed in comparison with non-users and the differences were non-significant. Past NSAID use did not show significant effect estimates compared with never users.
Effect of NSAID use on left ventricular function in the Rotterdam Study: effect estimates, 95% CIs and p values
To check for potential confounding we conducted an additional analysis in which we adjusted for body mass index and use of antihypertensive drugs (ATC code C02), β blockers (ATC code C07), calcium antagonists (ATC code C08) and ACE inhibiters (ATC code C09). None of these covariables changed the effect estimates of the echocardiographic parameters by more than 10%.
A dose-effect relationship was not observed within users. The F tests did not show statistically significant results when DDD was entered into the model (p=0.650 for LVED, p=0.929 for LVES and p=0.719 for FRAC). Comparison of two different regression models with dummy variables for duration of use only and dummy variables that additionally incorporated dosage showed that the latter was inferior for LVED, LVES and FRAC.
In participants for whom MDRD could be calculated, the MDRD did not change the effect estimates of the echocardiographic parameters by more than 10% and no interaction was found between NSAID use and renal function.
Discussion
To our knowledge, this report is the first to examine the early effects of NSAIDs on echocardiographic parameters in the general population. We found a significant association between echocardiographic parameters (LVED, LVES and FRAC) and recently started NSAIDs. Participants who were current users of NSAIDs for <14 days before their echocardiography had raised LVED and LVES values compared with those who were not current NSAID users, and FRAC was significantly reduced. A potential explanation for these findings is that the left ventricle has to deal with a higher circulating blood volume because of the influence of the NSAIDs on vascular homeostasis in the kidneys. The change in the parameters probably reflects transient compensation for the higher circulating blood volume and thus cardiac load. Except for an effect on LVED, participants who were current users of NSAIDs for >14 days before their echocardiography did not show a significant difference from those who were not current NSAID users. LVES was much lower in current users of NSAIDs for >14 days than in those using NSAIDs for <14 days. The increased FRAC in participants who were current users of NSAIDs for >14 days before their echocardiography possibly reflects a compensatory response to the decreased LVES compared with those currently using NSAIDs for <14 days before their echocardiography. Whether this compensation is in line with the Frank Starling curve or follows other biological compensatory pathways could not be studied. However, a dose-effect relationship was not observed.
Strengths and limitations of the study
An important advantage of the Rotterdam Study is the prospective ascertainment of risk factors and outcome variables over a relatively long period of follow-up for which extensive information is available for use in the analysis. Information bias is therefore unlikely. Although a proportion of the target sample of the Rotterdam Study did not participate, a strength of the Rotterdam Study with its population-based character is the relatively high participation rate (78%) which reduces the potential impact of selection bias. Because of the character of the study population, these results may be generalisable only to the older population. Unfortunately, only one echocardiography was available per participant, so we were not able to study whether these echocardiographic parameters were permanent or improved after discontinuation of NSAIDs. For further investigation of underlying mechanisms, more than one echocardiography would be needed which would make it possible to investigate the changes induced by and during NSAID use and thereafter within individuals. Another potential limitation is that we do not know whether all users with a current prescription actually used the NSAIDs. However, as this would lead to underestimation of the true exposure, our estimates are conservative. Furthermore, NSAIDs are available free of prescription (over-the-counter use) but, as this would result in non-differential misclassification, this would also result in conservative estimates. Another issue could be that participation in this study was voluntary but, since our response rates are relatively high (78%), selection bias is less likely. Because of the observational character of our data, confounding cannot be excluded completely. Since we observed no effect for past use on studied echocardiographic parameters, however, confounding by indication with a lasting effect on cardiac function is unlikely. Although we cannot completely exclude confounding by indication in current users for indications with a short-term effect on cardiac function, it should be emphasised that echocardiograms were taken in ambulatory elderly subjects as part of the Rotterdam Study and not based on clinical indications. Furthermore, the use of NSAIDs is contraindicated in patients with heart failure so, without adjustment for heart failure, the results may even be an underestimation of the true risk. The results were therefore adjusted for people who had a diagnosis of heart failure before echocardiography but, as shown in table 1, this only applied to a small number of participants.
In conclusion, this study is the first to investigate the effect of NSAIDs on echocardiographic parameters in community-dwelling elderly people and shows that there is a transient association between short-term use of NSAIDs and left ventricular function of the heart. These findings are in line with earlier studies which suggested that NSAIDs increased the risk of hospitalisation because of heart failure. Further echocardiographic studies are needed to gain more insight into how NSAIDs affect these preclinical parameters and the subsequent risk of heart failure.