If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Safe analgesic choices are limited in chronic kidney disease (CKD). We conducted a comparative analysis of harm from opioids versus nonsteroidal anti-inflammatory drugs (NSAIDs) in CKD.
Prospective cohort study.
Setting & Participants
3,939 patients with CKD in the Chronic Renal Insufficiency Cohort (CRIC) Study.
30-day analgesic use reported at annual visits.
A composite outcome of 50% glomerular filtration rate reduction and kidney failure requiring kidney replacement therapy (KRT), as well as the outcomes of kidney failure requiring KRT, hospitalization, and pre–kidney failure death.
Marginal structural models with time-updated exposures.
Participants were followed up for a median of 6.84 years, with 391 (9.9%) and 612 (15.5%) reporting baseline opioid and NSAID use, respectively. Time-updated opioid use was associated with the kidney disease composite outcome, kidney failure with KRT, death (HRs of 1.4 [95% CI, 1.2-1.7], 1.4 [95% CI, 1.1-1.7], and 1.5 [95% CI, 1.2-2.0], respectively), and hospitalization (rate ratio [RR], 1.7; 95% CI, 1.6-1.9) versus opioid nonusers. Similar results were found in an analysis restricted to a subcohort of participants reporting ever using other (nonopioid and non-NSAID) analgesics or tramadol. Time-updated NSAID use was associated with increased risk for the kidney disease composite (HR, 1.2; 95% CI, 1.0-1.5) and hospitalization (RR, 1.1; 95% CI, 1.0-1.3); however, these associations were not significant in the subcohort. The association of NSAID use with the kidney disease composite outcome varied by race, with a significant risk in blacks (HR, 1.3; 95% CI, 1.0-1.7). NSAID use was associated with lower risk for kidney failure with KRT in women and individuals with glomerular filtration rate < 45 mL/min/1.73 m2 (HRs of 0.63 [95% CI, 0.45-0.88] and 0.77 [95% CI, 0.59-0.99], respectively).
Limited periods of recall of analgesic use and potential confounding by indication.
Opioid use had a stronger association with adverse events than NSAIDs, with the latter’s association with kidney disease outcomes limited to specific subgroups, notably those of black race.
Coincident with the concern for NSAID use in CKD is acknowledgment of the opioid epidemic and recommendations to avoid long-term opioid use. In 2013, American health care providers wrote approximately 250 million opioid prescriptions,
Government and health agencies have established pain management guidelines intent on directing prescribers toward more judicious opioid use. Notable examples include the World Health Organization 3-step analgesic ladder for cancer-related pain
The CDC guideline for chronic pain management recommends nonpharmacologic and nonopioid therapies including acetaminophen, select antidepressants, anticonvulsants, and NSAIDs before long-term opioid use.
Hence these analgesic classes are both likely to be used and interchanged in CKD; however, comparative outcomes of using drugs from these analgesic classes are not known. In this analysis of the CRIC Study, our objective is to evaluate the association of opioid and NSAID use with clinical outcomes in patients with CKD not requiring kidney replacement therapy (KRT).
Study Design and Participants
The CRIC Study commenced in 2003, with phase 1 and 2 enrollment completed in 2006, and continued follow-up to date with the design previously described.
This analysis examined 3,939 participants who gave informed consent and were enrolled at 21 to 74 years of age with age-specific estimated glomerular filtration rate (eGFR) eligibility criteria of 20 to 70 mL/min/1.73 m2 from 7 US centers with 13 clinical sites and institutional review board approval at each site. Briefly, CRIC participants underwent annual in-center visits providing demographic information, medical history and status update, vital signs, blood and urine samples, and other survey-based information. GFR was estimated using the 4-variable Modification of Diet in Renal Disease (MDRD) Study equation, the prevailing clinical measure of kidney function at study commencement.
Coordinators recorded participants’ prescription and over-the-counter medications, supplements, and vitamins from 30 days preceding the study visit. To reduce recall bias, participants were asked to maintain a list or bring medications to visits. The drug name, frequency, total daily dosage, dosage units, and administration route were documented. Individual medications and constituents of combinations were identified using the First Databank dictionary (First Databank, Inc).
Classification of Analgesics
The CRIC data file was closed in May 2014 to permit data cleaning and preparation for this analysis. The NSAID category included all oral NSAIDs and cyclooxygenase 2 (COX-2) inhibitors. Aspirin was classified as an NSAID if total daily dosage was >325 mg, dose frequency was more than once daily, or the drug was part of a combination analgesic (excluding aspirin and dipyridamole). Members of the opioid class included all orally administered narcotics as designated in First Databank, such as hydrocodone, codeine, and oxycodone. Methadone represented <0.1% of opioid entries and no buprenorphine/naloxone use was reported. Opioids used as cold or cough remedy were excluded. Members of the “other” (nonopioid non-NSAID) class included predominately acetaminophen. Because tramadol has overlapping but distinct pharmacologic properties from narcotics,
it was considered as a separate class unless taken in combination with an NSAID or an opioid, in which case the medication was classified in the latter’s category. We defined the time-updated opioid and NSAID use at each clinical visit as whether a patient had reported NSAID or opioid use at that visit or any one of the previous CRIC visits. The class of other (nonopioid non-NSAID) analgesics included all other oral analgesics heretofore not classified. Of note, >90% of those entries were acetaminophen alone or in combination with another agent (eg, diphenhydramine). No intravenous or topical analgesics were included in the analysis.
We examined 4 clinical outcomes, including kidney failure requiring KRT, the composite of kidney failure with KRT and 50% reduction in eGFR from baseline, pre–kidney failure death, and number of pre–kidney failure hospitalizations between 2 consecutive annual visits. GFR was estimated annually.
For kidney failure with KRT and the composite kidney disease outcome, participant follow-up was censored at the time of death, withdrawal, loss to follow-up, or end of the follow-up period, whichever occurred first. For pre–kidney failure death and number of hospitalizations, participant follow-up was censored at the time of KRT initiation, withdrawal, loss to follow-up, or end of the follow-up period, whichever occurred first.
We considered several clinically relevant covariates including baseline factors: sex, race, education level, and income reported at study entry. Time-dependent covariates included age, any alcohol drinking, comorbid conditions (diabetes, cardiovascular disease, hypertension, asthma, nonskin cancer, hyperkalemia, and arthritis), GFR, urinary protein-creatinine ratio (UPCR), response on the Beck Depression Inventory, symptom severity, and Kidney Disease Quality of Life 36 questionnaire (KDQOL-36) burden and symptoms subscales, 12-Item Short Form Health Survey (SF-12) physical composite (including a question asking how much pain impeded activities of daily living), SF-12 mental composite, nephrologist visits, and other analgesic use (nonopioid/non-NSAID analgesic and tramadol) collected at annual visits. Urinary protein and creatinine excretion were also measured using standard assays. UPCRs from 24-hour and spot urine specimens were combined to a single UPCR variable.
For descriptive analyses, χ2 tests and t tests compared discrete characteristics and continuous variables, respectively, across groups. We examined the association between time-updated opioid and NSAID use and the study outcomes while controlling for time-dependent covariates. We applied joint marginal structural models because several time-dependent covariates including eGFR could be both a consequence and a predictor for analgesic use. The challenges of making causal inferences from observational data have been previously discussed and illustrated with steps of fitting joint marginal structural models described.
In brief, we used a pooled logistic regression model to predict the probability of time-updated NSAID use at each visit based on NSAID use and opioid use at the previous visit and covariates at the previous visit. Another pooled logistic regression model was applied to predict the probability of time-updated opioid use at each visit based on NSAID use at both the current and previous visits, opioid use at the previous visit, and covariates at the previous visit. Inverse probability weights were computed and stabilized. To control for informative censoring, the inverse probability of censoring weight was also computed and stabilized. The final weight was the product of the NSAID and opioid exposure and the censoring weight. The final weight was also truncated at the 99th percentile. Finally, we fit a weighted discrete failure time model for each of the survival outcomes through generalized estimating equations and using the final weight developed from the logistic regression models. Only baseline covariates were included in the discrete failure time models. For hospitalizations, we fit a weighted Poisson regression model using generalized estimating equation and the final weight.
We performed the analyses using the full cohort and a subcohort including participants who ever used another (nonopioid non-NSAID) analgesic or tramadol at baseline or during follow-up as a surrogate for need of pain relief. We also performed stratified analyses using demographic variables and key predictors of kidney outcomes: baseline age (<65 and ≥65 years), sex, race (nonblack and black), diabetes status, eGFR (≤45 vs >45 mL/min/1.73 m2), and UPCR dichotomized at the sample median.
To demonstrate that our results were robust and not due to unmeasured confounding, we examined the association of opioids and NSAIDs with risk for incident diabetes as a negative control outcome among CRIC participants without diabetes at enrollment and using the joint marginal structural models as described.
All analyses were performed using SAS, version 9.4 (SAS Institute Inc).
The 3,939 participants had a median follow-up of 6.84 years. There was a total of 24,838 visits for time to kidney failure with KRT or pre–kidney failure death, and 24,552 visits for the composite outcome of kidney failure with KRT and 50% reduction in eGFR. Tables 1 and 2 show the overall baseline characteristics of CRIC participants grouped by reported baseline opioid and NSAID use. Comparing the 391 (9.9%) participants who reported baseline opioid use with the 3,548 (90.1%) who did not, the former group was more likely to be female; be black; have an annual income of ≤$50,000; and have a history of rheumatoid arthritis, cardiovascular disease, asthma, and nonskin cancer and were less likely to report alcohol drinking. Compared with the 3,327 (84.5%) participants who did not report baseline NSAID use, the 612 (15.5%) who did were more likely to be aged 45 to 64 years, female, nonblack, a college graduate or with higher education, and with an income >$50,000, and report drinking alcohol. Those reporting baseline NSAID use also were more likely to have a history of asthma, have higher eGFRs, and less likely to have diabetes, cardiovascular disease, hypertension, previous hyperkalemia, and/or prior visit with a nephrologist. Opioid users were more likely to have depressive symptoms than nonusers and lower scores for the KDQOL-36 and its components domains (Table 2). However, NSAID users had higher KDQOL-36 scores compared with nonusers.
Table 1Baseline Characteristics of CRIC Participants Overall and by Opioid and NSAID Use
Table 3 displays crude rates of outcomes in the year following observed visits classified by opioid and NSAID use alone, in combination, or use of neither class of analgesic. The crude incidence of death was highest in the opioid-only group and opioid and NSAID group, with crude rates (per 100 person-years) of 3.5 (95% confidence interval [CI], 2.8-4.3) and 3.5 (95% CI, 2.2-5.5), respectively. The crude incidence of kidney failure with KRT in the opioid and other (nonopioid non-NSAID) groups were 4.2 (95% CI, 3.4-5.2) and 4.3 (95% CI, 4.0-4.6) per 100 person-years, respectively, appearing to be higher than the crude rates in the NSAID-only group (1.9 [95% CI, 1.4-2.5] per 100 person-years) and the opioid and NSAID group (2.5 [95% CI, 0.1-1.5] per 100 person-years). A similar pattern was observed for the composite kidney disease outcome, with crude rates in the opioid and other analgesic groups of 5.9 (95% CI, 5.3-6.3) and 5.9 (95% CI, 5.5-6.3) per 100 person-years, respectively. For hospitalizations, the opioid-only group had the highest crude rate of hospitalizations, while the NSAID-only group had the lowest crude rate (108.6 [95% CI, 99.1-118.9] vs 58.0 [95% CI, 51.5-65.3] per 100 person-years).
Table 3Crude Rate of Events (per 100 person-years) Following Annual Visits Classified by Reported Opioid and NSAID Use
Pre–Kidney Failure Death
Kidney Failure With KRT
Composite Kidney Disease Outcome
Pre–Kidney Failure Hospitalization
Rate per 100 Visits (95% CI)
Rate per 100 Visits (95% CI)
Rate per 100 Visits (95% CI)
Rate per 100 Visits (95% CI)
Opioid (no NSAID)
NSAID (no opioid)
Other (nonopioid non-NSAID)
Opioid & NSAID
Note: Total number of visits used for pre–kidney failure death is 24,838. Total number of visits for kidney failure with KRT and pre–kidney failure hospitalization is 20,899. Total number of visits for the composite kidney disease outcome is 20,613.
Table 4 shows the association of time-updated opioid and NSAID exposure with outcomes in the full cohort with the risk estimates expressed as hazard ratios (HRs) for kidney disease outcomes and as rate ratios (RRs) for hospitalization. Time-updated opioid use was associated with increased adjusted risk for all 4 outcomes relative to never using opioids during CRIC participation, with HRs of 1.4 (95% CI, 1.2-1.7), 1.4 (95% CI, 1.1-1.7), and 1.5 (95% CI, 1.2-2.0) and RR of 1.7 (95% CI, 1.6-1.9) for the kidney disease composite, kidney failure with KRT, death, and hospitalization, respectively. Time-updated NSAID use was associated with a modestly increased hazard of the kidney disease composite and risk for hospitalizations relative to never using NSAID with HR of 1.2 (95% CI, 1.0-1.5) and RR of 1.1 (95% CI, 1.0-1.3). However, there was no significant association between time-updated NSAID use and kidney failure with KRT or death.
Table 4Associations of Time-Updated Cumulative NSAID and Opioid Exposure With Outcomes, Adjusting for Time-Dependent Covariates in the Full Cohort and a Subcohort Comprising Participants Who Ever Used Other Analgesics or Tramadol During CRIC Study
Table 4 demonstrates the associations of time-updated opioid and NSAID exposure with outcomes in the subcohort comprising participants ever exposed to other (nonopioid non-NSAID) analgesics or tramadol during the study. The strength of association between opioid use and the outcomes was comparable to the full cohort. The association between NSAID use and hospitalization, kidney failure with KRT, the kidney disease composite, and death was no longer significant.
Forest plots (Fig 1; Table S1) display the varying HRs for the association of each analgesic group and the outcomes within subgroups designated by age, sex, race, diabetes, GFR, and UPCR at baseline. The association between time-updated NSAID use and the composite kidney disease outcome was stronger in blacks versus nonblacks (HRs of 1.31 [95% CI, 1.01-1.69] and 0.83 [95% CI, 0.64-1.09], respectively; P = 0.02 for effect modification). The association of time-updated NSAID use and kidney failure with KRT also varied across sex and baseline eGFRs with a higher HR for males versus females (HRs of 1.21 [95% CI, 0.91-1.61] and 0.63 [95% CI, 0.45-0.88], respectively; P = 0.004 for effect modification) and significantly lower risk for kidney failure with KRT in the lower versus higher eGFR subgroups (HRs of 0.77 [95% CI, 0.56-1.0] and 1.38 [95% CI, 0.89-2.14], respectively, for eGFR ≤45 and >45 mL/min/1.73 m2; P = 0.02 for effect modification). The association of pre–kidney failure hospitalization with opioid use (Fig S1A; Table S1) was higher in the lower versus higher baseline UPCR subgroups (RRs of 1.90 [95% CI, 1.66-2.17] and 1.54 [95% CI, 1.36-1.74], respectively, for UPCRs below and above the median; P = 0.02 for effect modification). Varying association between NSAID use and pre–kidney failure hospitalization had no significant effect modification (Fig S1B; Table S1).
Sensitivity analyses examined 1,442 CRIC participants without diabetes at enrollment. In this subgroup, neither opioid nor NSAID use were associated with incident diabetes (HRs of 1.27 [95% CI, 0.87-1.84] and 1.03 [95% CI, 0.73-1.45], respectively).
To further explore the impact of potential unmeasured confounders on the associations between opioids and NSAIDs and outcomes, we computed E values for potential unmeasured confounders for each risk estimate and the corresponding CI (Table S2).
E values ranged from 2.2 to 2.8 for the risk estimates determined for the associations between opioids and all outcomes and between 1.4 and 2.6 for the corresponding lower CIs. E values were in the higher portion of that range for the subcohort risk estimates but closer to the null for the weaker associations reported for NSAIDs and outcomes.
We also explored potential confounding of other drug groups that may be used as analgesics, including anxiolytics and antiepileptics (Table 1). However, only antiepileptics were associated with the outcome of death. Hence, we repeated the analysis including antiepileptics with the time-dependent covariates input to the marginal structural models for death. The results were essentially unchanged from those reported in Table 4.
In this cohort of adults with CKD, we demonstrated that reported opioid use within 30 days of ascertainment and treated as a time-updated exposure was associated with a substantial risk for adverse kidney disease outcomes, death, and hospitalization. This was in contrast with the unexpected and modest relationship of NSAID use with adverse outcomes. The association between NSAID use and adverse kidney disease events was most prominent in blacks, with a potentially beneficial association with outcomes observed in subgroups including women and those with lower eGFRs.
Physicians have long reported associations between various analgesics and kidney disease. “Analgesic nephropathy” is characterized by papillary necrosis, chronic interstitial nephritis, and progressive decreases in GFRs.
In our report of NSAID use in the CRIC Study, a quarter of study participants reported NSAID use at baseline or at least 1 annual visit, with a substantial proportion of users reporting treatment over the study.
Interpretation of the findings warrant consideration of the limitations inherent to its design. With observational analyses, one cannot overlook confounding by indication whereby use or nonuse of one or the other analgesic is driven by factors that may be associated with the outcome of interest rather than the primary exposure, in this case, analgesic choice. To minimize confounding by indication, we used causal inference models with inverse probability weighting by expected analgesic use in the examination of the association of analgesics with outcomes. More extensive characterization of the time-updated exposure including variations in dosage and drug discontinuation was limited by the modest sample size. Additionally, the CRIC Study was limited by its lack of a detailed pain assessment including measures of severity and type of pain. However, we used a wide array of available measures of severity of illness and function embedded in the KDQOL-36 and SF-12, which include a gauge of pain’s impedance of work and ability to perform activities of daily living.
Notably, analgesic use exposure ascertainment in CRIC was restricted to self-report, limited to the 30 days preceding an annual visit, and did not necessarily reflect actual use over more distant intervals. Previous studies have examined the fidelity of self-report of NSAID and acetaminophen use when compared with urine drug screening.
Although the study evaluated the use of NSAIDs and opioids versus nonusers of these drugs and the subgroup who were ever treated with any analgesics including the broader range of pain modulators such as acetaminophen, it did not assess the independent effect of the latter because this group served as the analysis reference group. Also, one cannot rule out the possibility that analgesic choice may have been different during the years of this cohort before the opioid epidemic was more broadly recognized. However, we expect the reported associations would only be minimally influenced by secular trends in use.
Of note, this is the first study we are aware of examining the comparative harm of NSAID versus opioid use in CKD. Both classes of agents have recognized risk profiles that are likely amplified in CKD, justifying close consideration of their risk versus benefit. Perhaps most importantly, the equipoise may be avoided with consideration of nonpharmacologic analgesic interventions that often show promising effectiveness in pain syndromes.
In conclusion, our study findings suggest that opioid use is associated with greater harm in CKD than NSAIDs, with a substantial increase in risk for death and poor kidney outcomes. The adverse effects of NSAIDs appear to be less consistent across subgroups with evidence for patient strata in which NSAID use is at least neutral and possibly beneficial. Further studies are needed to confirm such variable findings. Although a prospective trial comparing analgesics in patients with CKD with comparable degrees of pain and indications for analgesics is desirable, such a study is unlikely. Future guidance for strategies for patients with non–KRT-requiring CKD therefore will be based on comparative harm studies such as this and further studies are needed to verify the reported findings.
CRIC Study Investigators
Lawrence J. Appel, MD, MPH, Harold I. Feldman, MD, MSCE, Alan S. Go, MD, Jiang He MD, PhD, John W. Kusek, PhD, James P. Lash, MD, Panduranga S. Rao, MD, Mahboob Rahman, MD, and Raymond R. Townsend, MD.
Authors’ Full Names and Academic Degrees
Min Zhan, PhD, Rebecca M. Doerfler, PhD, Dawei Xie, PhD, Jing Chen, MD, Hsiang-Yu Chen, MS, Clarissa J. Diamantidis, MD, Mahboob Rahman, MD, Ana C. Ricardo, MD, James Sondheimer, MD, Louise Strauss, BSN, Lee-Ann Wagner, MD, Matthew R. Weir, MD, and Jeffrey C. Fink, MD, MS.
Research idea and study design: MZ, JCF; data acquisition: DX, HYC; data analysis/interpretation: MZ, MR, ACR, JS, LS, L-AW, JCF; data curation: RMD, JCF; data visualization: RMD, JCF, JC; statistical analysis: MZ; supervision or mentorship: CJD, MRW. Each author contributed important intellectual content during manuscript drafting or revision and agrees to be personally accountable for the individual’s own contributions and to ensure that questions pertaining to the accuracy or integrity of any portion of the work, even one in which the author was not directly involved, are appropriately investigated and resolved, including with documentation in the literature if appropriate.
Drs Fink, Doerfler, and Zhan were supported by National Institutes of Health (NIH)/ National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) R01 DK090008 . Funding for the CRIC Study was obtained under a cooperative agreement from NIDDK ( U01DK060990 , U01DK060984 , U01DK061022 , U01DK061021 , U01DK061028 , U01DK060980 , U01DK060963 , and U01DK060902 ). In addition, this work was supported in part by the Perelman School of Medicine at the University of Pennsylvania Clinical and Translational Science Award (CTSA) NIH/ National Center for Advancing Translational Sciences (NCATS) UL1TR000003 , Johns Hopkins University UL1 TR-000424 , University of Maryland GCRC M01 RR-16500 , Clinical and Translational Science Collaborative of Cleveland , UL1TR000439 from the NCATS component of the NIH and NIH Roadmap for Medical Research, Michigan Institute for Clinical and Health Research (MICHR) V 2014.07.28 UL1TR000433 , University of Illinois at Chicago CTSA UL1RR029879 , Tulane COBRE for Clinical and Translational Research in Cardiometabolic Diseases P20 GM109036 , Kaiser Permanente NIH/ National Center for Research Resources UCSF-CTSI UL1 RR-024131 . The funders did not have a role in study design; data collection, analysis, or reporting; or the decision to submit for publication.
The authors declare that they have no relevant financial interests.
Received July 5, 2019. Evaluated by 3 external peer reviewers and a statistician, with editorial input from an Acting Editor-in-Chief (Editorial Board Member Jane Schell, MD). Accepted in revised form December 11, 2019. The involvement of an Acting Editor-in-Chief to handle the peer-review and decision-making processes was to comply with AJKD’s procedures for potential conflicts of interest for editors, described in the Information for Authors & Journal Policies.