Association of Nonsteroidal Anti-inflammatory Drug Prescriptions With Kidney Disease Among Active Young and Middle-aged Adults

Key Points Question What is the association between prescribed dosages of nonsteroidal anti-inflammatory drugs and later incident kidney disease among active young and middle-aged adults? Findings In this cohort study of 764 228 US Army soldiers, prescriptions of more than 7 daily defined doses of nonsteroidal anti-inflammatory drugs per month were associated with modest but significant increases in the adjusted hazard ratios of acute and chronic kidney disease diagnoses. Meaning Prescribers should be cognizant of potential kidney disease risks associated with higher doses of nonsteroidal anti-inflammatory drugs among active young and middle-aged adults; dosage reduction represents an approach that may decrease associated kidney disease outcome rates.


Introduction
Nonsteroidal anti-inflammatory drugs (NSAIDs) are widely used in the United States in prescription and over-the-counter forms, 1 with more than 70 million NSAID prescriptions written annually. 2In 2010, more than 29 million US adults were estimated to be regular NSAID users-an increase of 41% from 2005. 3 A recent study of self-reported over-the-counter and prescribed ibuprofen therapy noted that 90% of those using ibuprofen took it regularly, 37% took another NSAID in addition to ibuprofen, and 11% exceeded the recommended daily limit of ibuprofen. 4inicians who prescribe or recommend NSAIDs should weigh the benefits vs the risks for kidney health.Both selective and nonselective NSAIDs adversely affect the kidneys through prostaglandin-related effects. 5Potential insults include impaired renal blood flow and clinically significant cytotoxic effects. 6Signs and symptoms associated with NSAID use that can complicate blood pressure management, such as hypertension and edema, are relatively infrequent 5 but important.
][9][10][11][12][13] Particularly regarding chronic and end-stage kidney disease, NSAID-related research has often focused on specific areas, such as disease progression. 14,15For younger healthy individuals, some studies provide statements of reassurance about the overall risks of NSAIDs 16 and, in particular, about their renal effects. 17wever, evidence on this demographic group is relatively sparse.This limited information may be because NSAID use is less common among young and middle-aged adults, 1 and the expected population rate of clinically significant kidney disease due to NSAIDs is less than 1%. 18udying the NSAID-kidney disease association among working-aged adults therefore requires a large group with robust NSAID use.United States Army soldiers are a useful study population given recent research indicating that 69% or more of this sizable population may use NSAIDs. 19In addition, prior studies have raised concerns about kidney disease risk among NSAID users who engage in endurance exercise, [20][21][22] as renal blood flow may fall to as little as 25% of resting values during strenuous activity. 23The Army population is one in which endurance activities, such as running 24 and long-distance rucksack marching, 25 are regularly undertaken, so this group provides a unique window on NSAIDs and kidney disease among active persons.Other advantages of using a military population include standardized, comprehensive administrative and medical data, as well as preservice, annual, and combat duty-associated health screenings 26 that facilitate recognition of incident diseases.
We therefore used data on the total active-duty US Army to estimate the independent associations between prescribed oral NSAID use and incident acute kidney injury (AKI) and chronic kidney disease (CKD).Renal effects of NSAIDs have been shown to be dose dependent. 18Increased frequency and duration of NSAID use amplify the risk of nonrenal adverse effects. 18,27Accordingly, we devised methods to study NSAID exposure volume over time while controlling for major factors of potential relevance to kidney dysfunction.

Population and Data
This retrospective cohort study was conducted with longitudinal data on the active-duty US Army collected from January 1, 2011, to December 31, 2014.Data were combined from official sources (eTable 1 in the Supplement) and stripped of identifiers.Analyses were conducted from August 1 to November 30, 2018.The institutional review board of Stanford University approved this study, which underwent secondary review by the Human Research Protections Office of the Defense Health Agency.A waiver of consent was granted because the research (1) involves no more than minimal risk to the participants, (2) does not affect the rights or welfare of the participants, and (3) could not

Dependent Variables
During 2011-2014, the Army used the International Classification of Disease, Ninth Revision, Clinical Modification (ICD-9-CM) system.We identified outcomes from diagnoses in outpatient and inpatient care by using ICD-9-CM codes for AKI (584.x,586, 580.9) and CKD (581.x,0][31] A dedicated data system (eProfile) is an additional repository outside the health record per se in which soldiers' duty-limiting health conditions are tracked. 32We therefore also defined outcomes by using eProfile entries noting relevant kidney conditions.

Independent Variables Demographic Factors
Multiple demographic factors were included to control for potential confounding.Sex and Hispanic ethnicity were binary variables.Running age in years and self-reported race were categorical covariates.

Administrative Factors
We controlled for socioeconomic status by using each participant's running military pay grade. 33tal service time was updated each person-month.Combat duty was included as a covariate because of the associated potential for an increase in outcome risk due to injury or surgery.

Biomedical Factors
We used an NSAID exposure variable based on dispensed prescription medications and agentspecific World Health Organization-defined daily doses (DDDs), 34 which represent estimates of typical maintenance doses for adults. 35The categorical variable represented the mean of the total monthly NSAID DDDs dispensed in the 6 months preceding each observation.This rolling window was used to capture exposures that were sufficiently long but also recent with regard to expected kidney effects; the present month was excluded to reduce the potential for overdose as a causal mechanism.
The ICD-9-CM codes were used to identify histories of the potentially contributory conditions hypertension 36 (401.x,402.x, 405) and type 1 or 2 diabetes 37 (250.x).We included a covariate for a association with kidney injury appears to be well established. 38We further controlled for body mass index 39 by using its standard categories, 40 plus a category for missing data.Other potentially contributory conditions (eg, systemic lupus erythematosus 41 ) were explored, but were deemed too infrequent in this population for inclusion.

Statistical Analysis
To characterize the types and quantities of NSAID exposures, we tabulated counts of specific agent classes dispensed to study participants and percentages thereof.Preregression analyses included χ 2 tests of distribution differences for selected covariates.To estimate the independent associations of NSAIDs with the kidney outcomes, we used dedicated Cox proportional hazards regression models for AKI and CKD.We also computed the adjusted risk of each outcome for participants in each of the NSAID exposure categories.These figures were calculated by totaling the products of the Cox regression coefficients and the covariate values, which permitted a computation of the absolute risk differences among the NSAID exposure groups.We additionally performed Wald tests for the interaction between selected medical conditions and NSAID use.In all analyses, 2-sided α < .05defined statistical significance.All analyses were conducted using Stata statistical software, version 14.2 (StataCorp).
There were 763 752 participants eligible for the AKI analysis, among whom 2356 (0.3%) experienced incident AKI events.Among the AKI outcomes, 13 (0.6%) were detected from eProfile data rather than diagnoses in the electronic health record.Of 763 654 individuals eligible for the CKD
Histories of diabetes or rhabdomyolysis were present among fewer than 1% of the participants, while hypertension was more prevalent at up to 8.8%.We did, however, observe statistically significant differences in the distributions of biomedical and demographic factors comparing groups with and without NSAID exposure (Table 2).The proportion of women increased from 12.5% of those without NSAID use to 18.3% of those in the highest use group.Individuals who received the  2).Statistically significant differences in distributions were also observed for each of the military-specific factors.For example, increasing duration of military service was associated with increased NSAID use.
Specifically, those with greater than 12 years of service made up 19.4% of the no NSAIDs group and 30.4% of the highest NSAIDs group (eTable 2 in the Supplement).
a Cox proportional hazards regression models used in analyses.The models additionally controlled for military service time, pay grade, and combat experience.eTable 3 in the Supplement provides related findings.Downloaded from jamanetwork.comby guest on 05/04/2024 of 7 or more DDDs per month was associated with significant increases in the adjusted hazard ratios (aHRs) of both AKI (aHR, 1.2; 95% CI, 1.1-1.4)and CKD (aHR, 1.2; 95% CI, 1.0-1.3).Based on postregression-adjusted risk computations, the highest NSAID exposure level was associated with annual case excesses per 100 000 exposed individuals of 17.6 cases for AKI and 30.0 cases for CKD.
Mean NSAID exposure of 1 to 7 DDDs was associated with smaller hazard increases that were not significant.
To address the issue of whether the selected medical condition covariates might interact with NSAID use, we conducted a formal test of the statistical significance of each such interaction (hypertension, diabetes, and rhabdomyolysis).Only the interaction between prior hypertension and NSAIDs in the CKD analysis was statistically significant (aHR, 0.7; 95% CI, 0.5-0.9).This finding provides some evidence that, in this population, the association between NSAIDs and CKD is significantly weaker among those with prior hypertension than those without.
Other findings included elevated hazards of incident AKI and CKD with increasing age and among men and African American participants.Our CKD findings differed from those seen in the United States Renal Data System, where women demonstrated a higher CKD rate. 42Hispanic soldiers had a lower hazard of AKI compared with non-Hispanic individuals.

Strengths and Limitations
Strengths of this study include the use of standardized, detailed data on a large population and the ability to exclude those with prior disease.Our results may generalize reasonably well to nonmilitary adults of similar ages, but exposures among service members might differ substantially from those of civilians.In addition to required physical exertion, the life of most Army soldiers includes regular field training in outdoor settings.Most US Army installations are in the warm US south, 43 and recent combat deployments have taken place in largely hot and arid regions.Therefore, intermittent dehydration that further depletes fluid volume and increases the strain on the kidneys 44 may be unusually frequent or substantial among soldiers.Our study's results may most closely apply to civilians with strenuous, potentially dehydration-producing occupations, such as athletes, firefighters, and farm, construction, and industrial workers.
This research was subject to the limitations of diagnosis coding, including general imprecision.
One concern associated with diagnosis code validity could be case underdetection, 45,46 which may arise when procedure codes, such as for kidney transplantation, are entered rather than kidney disease codes. 47This issue is likely less important in our study population because early and accurate identification of serious conditions is a key duty of military clinicians to ensure adherence to medical service standards for training and duty. 26A diagnosis would usually occur well before advanced procedures, such as hemodialysis or transplantation, are required.Also, our data sets afforded somewhat augmented event detection owing to clinician entries in the eProfile record system.
We nonetheless acknowledge that our CKD case detection mechanisms may have been reduced by our relatively short follow-up times, as clinical diagnoses and eProfile entries may have occurred afterward for some participants.Misclassification of AKI as CKD and vice versa constitutes another specific possible form of potential imprecision in our data, but the similar findings for the outcomes reduce this concern.We also acknowledge that the sensitivity and specificity of diagnosis codes may further vary in unknown ways, such as across exposure strata.More generally, as in any observational study, residual confounding is possible.However, the wide array of demographic, job-related, and health-related control variables used should reduce concerns.
Other limitations of the study arise from our reliance on dispensed NSAID prescriptions to quantify drug exposure.Whereas our data captured clinicians' instructions, there was no mechanism to observe the details of individual NSAID use.We would expect this approach to have created conservative association estimates because if prescription NSAID intake varied from the total quantity prescribed, it was presumably lower.However, we were unable to account for over-thecounter NSAID use, which could have offset this phenomenon.
Of the participants, 238 168 (31.2%) were new to the Army during the observation period.
These individuals differed in gross exposure to the military environment from those with greater total service times.Furthermore, the presence of experienced soldiers in the data set represents a possible selection sieve, as these soldiers have served for potentially many years.We included the covariates for age, service time, and combat experience specifically to provide control for these factors.
Recently, a more cautious tone has permeated the discussion about NSAID use, 45,48,49 with concerns including the potential delayed or inhibited healing associated with pain management. 50npharmacologic interventions are increasingly emphasized, 51 and research evidence on such options is available. 52Our findings provide additional support for the need for expanded research on alternative treatment options for pain and a greater focus on patient education about the risks and benefits of higher doses of NSAIDs.

Table 1 .
Top NSAIDs Dispensed to 764 228 Study Participants a Suffix elements and compounds associated with the active component of applicable agents (eg, sodium) were omitted for simplicity.b There were 1 630 694 prescriptions dispensed in total during the observed time.Percentages may not total 100 owing to rounding.

Table 2 .
Description of the 764 228 Participants at the Final Observation a Association of Nonsteroidal Anti-inflammatory Drug Prescriptions With Kidney Disease in Adultsgreatest NSAID volumes were twice as likely to be obese, composing 23.6% and 12.4% of the highest and lowest NSAID categories, respectively.Individuals who received the greatest NSAID volumes were also twice as likely to have histories of hypertension (8.8% vs 3.6% of the highest and lowest NSAID categories) and diabetes (0.9% vs 0.3% of the highest and lowest NSAID categories).African American participants were more highly represented among those who received the highest level of prescription NSAIDs than those who received none (22.9% vs 19.6%) (Table b eTable 2 in the Supplement provides other descriptive data on military service time, pay grade, and combat experience.cThePvaluesindicate results of χ 2 tests comparing factor distributions for those that were and were not found in each NSAID exposure category.JAMA Network Open | NephrologyJAMA Network Open.2019;2(2):e187896.doi:10.1001/jamanetworkopen.2018.7896(Reprinted) February 15, 2019 5/12 Downloaded from jamanetwork.comby guest on 05/04/2024