Acute and Chronic Acetaminophen Use and Renal Disease: A Case-Control Study Using Pharmacy and Medical Claims

BACKGROUND: Studies have examined the association between acetaminophen (APAP) use and renal disease; however, their interpretation is limited by a number of methodological issues. OBJECTIVES: To study the association between acute and chronic prescription-acquired APAP use and renal disease. METHODS: This was a retrospective case-control study of medical and pharmacy claims of a 10% random sample of the enrollees from the IMS LifeLink Health Plans commercial claims dataset for dates of service from January 1, 1997, through December 31, 2009. Subjects were continuously enrolled and aged 18 years or older. Cases had at least 1 incident claim of renal disease defined by ICD-9-CM codes in the primary diagnosis field. Controls were randomly selected from individuals without evidence of renal disease, liver disease, or asthma in medical claims and matched to cases in a 3-to-1 ratio based on 3 variables (age, gender, and geographic region). APAP exposure, dosage, and duration of use were measured in the 7 and 30 days (acute) and in the 1-year (chronic) look-back periods. Multivariable conditional logistic regression was used to estimate the risk of APAP exposure adjusted for comorbidities, use of other nephrotoxic drugs, and health system factors. RESULTS: There were 4,724 cases and 14,172 controls with a mean (SD) age of 60.8 (17.8) years, and 52.6% were males; 10.9% of cases and 4.2% of controls had APAP exposure in the 30 days pre-index with mean potential maximum daily dosages of 3,846.5 mg and 3,190.8 mg, respectively. Acute APAP exposure was significantly associated with renal disease, and the risk decreased with longer look-back periods (7 days: adjusted odds ratio [OR] = 1.93, 95% CI = 1.61-2.30); 30 days: OR = 1.71, 95% CI = 1.48-1.97). Cumulative APAP dosage greater than 1 kg and APAP use for longer than 30 days in the pre-index year were not significantly associated with an increased risk of renal disease (both P values = 0.900). CONCLUSIONS: Acute prescription-acquired APAP use was associated with renal disease, while chronic use was not. Because this study assessed APAP use in pharmacy claims, further research accounting for over-the-counter APAP use is warranted before the safety of chronic APAP consumption can be firmly established.

ies using large insurance claims datasets have studied the link between APAP use and renal disease.
Administrative claims data include records for a large number of patients for long time periods and can be particularly useful for the study of rare events. 18 Retrospective, interviewbased studies may be subject to recall bias if the cases remember and report their drug exposure more accurately than the controls. However, pharmacy claims record the start and end dates (fill date + days supply) of a prescription and the amount of drug prescribed and are therefore not biased by knowledge about the study outcome. 18 In an analysis of IMS LifeLink Health Plans' pharmacy claims data, Gokhale and Martin (2012) found that the annual mean cumulative APAP dosage increased from 55.3 gm per year in 2001 to 81.9 gm per year in 2008, indicating an increase in chronic prescription-acquired APAP use. 19 This change parallels increases in the use of APAPopioid combination products in the United States. 20 A study of annual APAP use based exclusively on pharmacy claims data reported that approximately 30% of APAP users had a potential maximum daily dosage exceeding the currently recommended maximum daily dosage (4 gm per day), indicating that administrative claims data do capture high risk APAP use. 19 Given the lack of previous claims-based studies examining the association between APAP use and renal disease and the rarity of the outcome, we investigated the association between acute and chronic prescription-acquired APAP use and renal disease using a large, nationally representative commercial insurance dataset.

■■ Methods Study Design and Data Source
This study was part of a larger project examining associations between acute and chronic APAP use and hepatic (liver disease) and nonhepatic (renal disease and asthma) outcomes, using a retrospective case-control study design. Data from 1997-2009 on a 10% random sample of the enrollees in the IMS LifeLink Health Plans were used for this study. This data source consists of claims from more than 98 U.S. managed care organizations and is representative of the commercially insured population in the country with respect to age, gender, and region. The data include pharmacy claims, International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes, procedure codes, and patient enrollment information for more than 6 million individuals followed for an average of 2.5 years. The data contain records for both OTC and prescription APAPcontaining products that have been billed as pharmacy claims; however, it is unlikely that many plans reimburse OTC APAP prescriptions, and less than 5% of APAP claims were for OTC products during the study period. The APAP records in the dataset have been used in a previous examination of APAP use and overuse patterns. 19 most prescribed medicines. 3 Although the drug is considered relatively benign, concerns are increasing over the excessive consumption of APAP. In the United States, 26,000 hospitalizations and 458 deaths due to APAP overdose have been reported annually, 4 and from 1993 through 2007, more than 700,000 emergency room visits were attributable to APAP overdoses. 5 Although the organ primarily affected is the liver, there has also been evidence of renal injury. 6 Some temporal and clinical evidence suggests that liver damage often precedes renal damage, but there are some reports of renal disease without significant hepatic injury, indicating that the mechanisms of organ injury may differ. 6 Ingestion of APAP in dosages exceeding 4 grams (gm) per day can lead to acute renal failure in individuals without risk factors, whereas lower dosages may lead to renal damage in individuals with chronic liver disease, those with concurrent alcohol consumption, and those with increased activity of the cytochrome P-450 enzyme system. 6 While APAP-induced hepatotoxicity has been widely studied, mechanisms of renal toxicity are less clear. Cytochrome P-450 enzymes, glutathione S-transferase, prostaglandin endoperoxidase synthase (PGES), and N-deacetylase are hypothesized to be involved in APAP-induced renal toxicity. 7 Like hepatic cells, renal microsomes also oxidize APAP to an arylating intermediate product via the P-450 dependent mechanism, indicating a biochemical mechanism of toxicity similar to that in the liver. The glutathione (GSH)-conjugate of a secondary metabolite is also thought to be involved in APAP-induced renal disease in CD-1 mice. Renal toxicity due to chronic APAP exposure depends on PGES according to studies on rabbit renal microsomes. It was found that human kidney medulla microsomes also catalyzed the PGES-based metabolic activation of APAP at rates similar to those in rabbit kidney microsomes. Variations in APAP-induced renal toxicity have been observed across different species and gender. 7 Another possible mechanism of renal injury is due to oxidative stress and tumor necrosis factor (TNF)-α production. A study reported oxidative stress-induced renal damage after APAP administration in rats. 8 Incidents of acute renal toxicity after large acute APAP dosages were reported in a case series 9 and a few case reports. 10,11 Several epidemiologic and clinical studies have examined the association between lifetime APAP use and renal disease. [12][13][14][15][16][17] A European autopsy study on 616 individuals reported a decreased prevalence of analgesic nephropathy in spite of the use of APAP-containing analgesics, indicating no renal disease with chronic APAP use. 12 While some epidemiologic case-control studies supported this finding, 14,17 a few others found a positive association between chronic APAP use and renal disease. 13,15,16 However, the interpretation of these studies is limited by a number of methodological limitations, such as the inability to clearly establish temporality of exposure prior to outcome, including other analgesics in exposure measures, and recall bias. To our knowledge, no population-based stud-

Cases
Eligible cases were individuals aged 18 years or older with at least 1 incident primary diagnosis code of acute renal failure (ICD-9-CM codes 584. 5 ; renal sclerosis unspecified (587.xx); renal osteodystrophy (588.0); other specified disorders resulting from impaired renal function (588.8); unspecified disorder resulting from impaired renal function (588.9); nephrogenic diabetes insipidus (588.1); unilateral small kidney (589.0); bilateral small kidney (589.1); and small kidney unspecified (589.9) from January 1, 1998, through December 31, 2009. 21 The sensitivity and specificity of the ICD-9-CM codes for acute renal failure are 35.4% and 97.7%, respectively. 22 We supplemented our case definition to include persons with primary diagnoses for chronic renal disease based on another claimsbased study defining renal disease; 21 this decision was made to increase the sensitivity of our measure at the potential expense of decreasing the specificity. For case selection, only the claims from inpatient hospitalization records, emergency room visits, surgical records, and outpatient visits were used. Claims with record type "ancillary" were excluded because they imply events incidental to direct care of patients (e.g., x-rays, transportation services).
APAP-induced renal disease may or may not be preceded by hepatotoxicity. 6 Therefore, cases of renal disease were checked for evidence of liver disease (acute liver necrosis, hepatitis, hepatic coma, hepatorenal syndrome, and coagulopathy; see Appendix) in the 10 days before the renal disease diagnosis. It is likely that patients would not have been prescribed APAP after a liver disease diagnosis. If the index date was the diagnosis date of renal disease for patients with preceding liver disease, it would be likely that these patients would not consume APAP in the pre-index period after their liver disease diagnosis and bias the results toward no association. To address this potential bias, for patients with liver disease in the 10-day window prior to incident renal disease, the index date was the date of diagnosis of liver disease. For those patients without prior liver disease, the index date was the date of diagnosis of renal disease.
All cases were required to have continuous health plan enrollment in the pre-index year. Since this study was a part of a larger project with 3 outcomes, to keep the methods consistent, cases with diagnoses of liver disease, renal disease, or asthma in the pre-index year were excluded. These exclusion codes (Appendix) contained a broader set of conditions than the case definitions to exclude persons with possible manifestations of each disease. We also excluded cases with previous liver, kidney, or lung transplant; those on immunosuppressant therapy (except corticosteroids); and those with liver, renal, respiratory-tract cancer, or secondary malignancies.

Controls
In order to increase statistical power, 3 controls per case matched on age, gender, and geographic location (East, Midwest, South, and West) were randomly selected from a group of individuals without ICD-9-CM codes for renal disease, liver disease, or asthma (Appendix) in any of the 4 diagnosis fields. Controls were assigned an index date the same as that of the corresponding case and were required to have continuous plan enrollment in the pre-index year. We excluded controls with a previous diagnosis of APAP poisoning (ICD-9-CM code 965.4x). 23 Other exclusion criteria were the same as those for the cases (Figure 1).

APAP Exposure Measures
APAP-containing products were identified using unique Medi-Span Generic Product Identifier (GPI, Medi-Span, Indianapolis, IN) codes. We measured any APAP exposure, dosages, and durations of APAP use for acute (7 and 30 days pre-index) and chronic (365 days pre-index) look-back periods (Table 1). Dosages were calculated as follows: 1. Potential maximum daily dosage (PMDD) in the 7-day and 30-day pre-index periods: This was the highest potential APAP dosage in any 1 day calculated in the pre-index period using the days supply, strength, and quantity fields in the data. Overlapping prescriptions were identified using fill dates and days supply, and the daily dosages were summed to obtain the potential maximum dosage. For example, for a patient with an APAP claim on January 1 for a 13-day supply of 52 tablets at 500 milligrams [mg] per tablet (daily dosage of 2,000 mg per day); a second APAP claim on January 3 for a 7-day supply, 42 tablets at 325 mg per tablet (dosage of 1,950 mg per day); and an index date of January 11, the PMDD in the 7 days pre-index was 3,950 mg.

Potential average daily dosage (PADD) in the pre-index month:
Dosage obtained by summing the APAP dosage contained in all prescriptions in the 30 days pre-index divided by the total days of APAP use. For example, using the scenario above, 3 days supply of the first pharmacy claim occurred after the index date and are not counted (12 tablets

Other Covariates
Using the enrollment information and pre-index medical and pharmacy claims, we obtained data on the following potential risk factors for renal disease: • Medical conditions: These were measured in the 365-day preindex period (in any of the 4 diagnosis fields) for both acute and chronic analyses and consisted of hypertension, 24,25 kidney infections, 26 heart disease, 24,27 substance abuse (alcohol/illicit drug use and abuse), 23,28,29 diabetes, 23,30 metabolic variables (gout and malnutrition), 28,31 and cancer 32 (Appendix). • Drug variables: Drug exposure (at least 1 day supply in the pre-index period) was measured in the 30 days pre-index for analyses of acute APAP exposure and in the 365 days pre-index for analyses of chronic APAP exposure. Use of the following drugs was identified using GPI codes: antibiotics, 28 nonsteroidal anti-inflammatory drugs (NSAIDs), 33,34 diuretics, 33 angiotensin-converting enzyme (ACE) inhibitors, 33 corticosteroids, 33 oral anticoagulants, 33 and miscellaneous drugs. 28,33 • Health system variables: Using the enrollment information, we obtained data on insurance payer/plan type.
Since hypertension and diabetes are important risk factors for kidney disease, 25 we explored the possibility of these diseases being potential effect measure modifiers and included interaction terms between these factors and APAP use. In addition to the previously mentioned covariates, we also included in the chronic APAP-use models a binary term for APAP use in the 30 days pre-index to control for short-term use of APAP. We could not adjust for race, since it was not available in the dataset.

Analysis
We measured baseline descriptive characteristics of the sample in the acute and chronic pre-index periods. Adjusted and unadjusted conditional logistic regression models were used to determine the effect of acute and chronic APAP use on the risk of renal disease. We used likelihood ratio chi-square tests to compare adjusted models with and without the interaction terms. Collinearity among predictor variables was tested using phi coefficients; no exploratory variables had a phi coefficient exceeding 0.5. Odds ratios (ORs) and 95% confidence interval (CI) estimates were calculated. All analyses were performed using SAS version 9.2 (SAS Institute Inc., Cary, NC). This study was approved by the Institutional Review Board of the University of Arkansas for Medical Sciences.

■■ Results
From about 6 million plan enrollees in the data source, we obtained 45,843 incident cases of renal disease, of which 16,163 met the 12-month pre-index continuous enrollment and age criteria ( Figure 1). We excluded those with certain prior medical conditions and drug use to arrive at a final sample of 4,724 cases. The case diagnoses included acute renal failure (n = 1,921), chronic renal failure (n = 1,905), nephritis (n = 631), and other kidney diseases (n = 267). From the same parent population, there were 1,366,555 age-, region-, and gender-matched controls meeting the enrollment criteria. Upon applying further exclusion criteria, we obtained 1,257,240 individuals from whom 3 controls were randomly chosen for every case to get a total of 14,172 controls. The mean (standard deviation [SD]) age of our sample was 60.8 (17.8) years, and 52.6% were male ( Table 2). Prevalence of drug use and medical conditions that increase the risk of renal disease were significantly higher among the cases compared with the controls, and 601 (12.7%) cases had evidence of liver disease in the 10 days before the diagnosis of renal disease.
Descriptive analyses of the APAP-use variables for the cases Acute and Chronic Acetaminophen Use and Renal Disease: A Case-Control Study Using Pharmacy and Medical Claims  ratio test was significant between the models with and without the hypertension-APAP interaction term (P < 0.05; data not shown), suggesting a better model fit with the interaction term. However, the ORs calculated based on the variance-covariance matrix and coefficients obtained from the conditional logistic regression models (as explained by Hosmer and Lemeshow [2000] 35 ) were in opposition to the anticipated direction (APAP exposure in nonhypertensive patients conferring higher risk than APAP exposure with hypertension). Given these findings and the lack of pathophysiological evidence of the joint effect of APAP and hypertension on renal disease, our final models included no interaction terms. Adjusted and unadjusted ORs of acute and chronic APAPuse measures are shown in Table 4. Unadjusted odds of renal disease for patients with any APAP exposure in the 7 and 30 and controls are shown in Table 3. APAP was used for at least 1 day in the pre-index year by 1,366 cases (28.9%) and 2,347 controls (16.6%). Mean cumulative dosages in the pre-index year for cases and controls were 117.92 gm and 83.49 gm, respectively. In the 30 days pre-index, 517 (10.9%) cases and 599 (4.2%) controls used APAP with mean PMDDs of 3,846.50 mg and 3,190.80 mg respectively. The total number of days of APAP use during the pre-index year was higher for cases compared with controls (47.7 days and 35.2 days, respectively). More than 95% of all APAP-containing prescriptions were for opioid-APAP combinations in both cases and controls.

Case-Control Selection Flowchart
We first ran models with interaction terms between APAP use and hypertension and diabetes. The likelihood ratio test between the models with and without the diabetes-APAP interaction term was not significant (data not shown). The likelihood   37 These genetic polymorphisms may be in part responsible for the elevated risks we observed at low levels of APAP use. Unfortunately, we could not investigate this possibility empirically without the ability to measure genetic or even ethnic differences in our data.
days pre-index were 3.17 and 2.80 times that of APAP nonusers, respectively. After controlling for covariates, the ORs decreased to 1.93 (95% CI = 1.61-2.30) and 1.71 (95% CI = 1.48-1.97) for 7-and 30-day pre-index exposure, respectively. Compared with nonusers, there was a 4.60-fold risk (95% CI = 2.87-7.39) of renal disease with PMDD exceeding 4 gm in the 7 days preindex. This was significantly greater than the risk conferred by PMDD of 4 gm per day or less (OR = 1.68, 95% CI = 1.38-2.03; P < 0.001 for trend). Potential maximum and average daily dosages in the 30 days pre-index were significantly associated with renal disease, but a significant dosage-dependent relationship was not observed (P = 0.23 and P = 0.57 for trend, respectively).
To explore other possible dosage thresholds, we re-categorized the APAP daily dosages using the cut-offs of 3.25 gm per day and 2.6 gm per day (instead of 4 gm per day), but the risk at lower dosages continued to be higher than that of nonusers (data not shown).
Cumulative dosage of at least 1 kilogram (kg) in the preindex year increased the renal disease risk by 13% (OR = 1.13, 95% CI = 1.01-1.26) compared with APAP nonuse (Table 4). Only 15 (0.3%) cases and 20 (0.1%) controls had a cumulative dosage exceeding 1 kg, and we obtained a nonsignificant estimate for this category (P = 0.900). We performed a power analysis using the Epicalc package in version 2.6.1 of R statistical software (open-source software available at http://www.rproject.org/) 36 and found only a 15.6% power to detect a significant difference between cases and controls for cumulative dosage exceeding 1 kg.
Having the last day of APAP use (recency based on fill date and days supply) within 0-30 days pre-index was associated with a significantly greater risk of renal disease (OR = 1.84, 95% CI = 1.59-2.13) compared with having the last day of APAP use between the 31st and 365th day pre-index (OR = 1.13, 95% CI = 1.01-1.26, P for trend < 0.001; Table 4). The total duration of APAP use up to 30 days was significantly associated with renal disease, while duration longer than 30 days was not.

■■ Discussion
In the present study sample, any APAP exposure in the 7 and 30 days pre-index increased the risk of renal disease by about 70%-90% compared with no APAP use, and the risk was 3.6 times greater when the PMDD exceeded 4 gm in the 7-day preindex window. The risk associated with APAP use decreased with a longer look-back period of 30 days. Although APAP is generally considered to be safe at therapeutic dosages (up to 4 grams per day), we found a 60% increased risk of renal disease with these dosages in spite of controlling for potential confounders.
It has been suggested that the risk of renal disease with therapeutic APAP dosages is high among individuals with genetic alterations in acetylation processes, which are often

Prescription-Acquired Acetaminophen Exposure, Dosages, and Durations of Use for Cases and Matched Controls in the Pre-Index Period
than recommended dosages, the observed renal disease risk is concerning. Actions to curb the use of more than the recommended dosage of APAP are warranted. Decisions about APAP use should be based on a comparison of the benefit-risk of APAP and alternative drugs, such as NSAIDS and narcotic analgesics not containing APAP. However, these alternatives are also associated with a number of adverse events, 41,42 and a thorough risk-benefit assessment using NNH, quality-of-life studies, alcohol use, and ethnicity considerations, among other factors, should guide recommendations about APAP use. The International Society for Pharmacoeconomics and Outcomes Research (ISPOR) Risk-Benefit Management Working Group also suggests calculating relative-value adjusted NNH (RV-NNH), which incorporates patient preferences for avoidance of negative clinical outcomes. 43 To aid clinical decision making about APAP use, we estimated the number needed to harm (NNH) 38 based on the ORs in our study and literature-based incidence rate of renal disease (0.18%). 39,40 The NNH should, however, be interpreted with caution, since it is based on an OR, which is only an approximation of the relative risk. For the 7 days pre-index, the NNH was 819 (95% CI = 541-1,466) with PMDD up to 4 grams and 156 (95% CI = 88-299) with PMDD greater than 4 grams, corresponding to absolute risk increases of 0.11% and 0.64%, respectively. This result implies that within a period of 7 days, only 156 individuals would have to be treated with APAP at a PMDD of more than 4 gm per day to observe 1 additional case of renal disease. With no known clinical benefit of APAP at dosages exceeding 4 gm, and given our findings combined with those of other studies that found APAP toxicity at more

Conditional Logistic Regression Analyses of Renal Disease: Adjusted and Unadjusted Odds Ratios for Acetaminophen Exposure Variables
patients and identified no instances of hepatic or renal failure or serum creatinine levels at or above 1.5 times the upper limit of the reference range at APAP dosages of 4 gm per day for up to 12 months (annual cumulative dosage of 1,460 kg). 47 Another clinical trial in 88 osteoarthritis patients randomized to APAP (dosage 2.6 gm per day of APAP for up to 2 years; equivalent to annual cumulative dosage 0.95 kg) also did not report the occurrence of any hepatic or renal adverse events. 48 A European clinical autopsy study on 616 adults and 2 epidemiologic casecontrol studies examining analgesic use (including APAP) also support the lack of association between lifetime APAP use and renal disease. 12,14,17 Contrary to these results, several case-control studies suggest an increased risk of renal disease with chronic APAP use. 13,15,16 However, the epidemiologic studies were based on self-reports of lifetime APAP exposure and may be subject to recall bias. Our claims-based study found no association between chronic APAP use and renal disease and adds to the majority of the literature suggesting the lack of such an association. However, given the conflicting evidence, further research is needed to firmly establish the safety of chronic APAP consumption.
There are reports in the literature about chronic highdosage APAP use without any ill effects. 49 Although regeneration of hepatocytes is hypothesized to be the reason for an auto-protective effect against APAP-induced hepatotoxicity, 50 according to our knowledge there is no published evidence about auto-protection against APAP-induced renal disease. Pathophysiological research is warranted to determine if such a mechanism exists.

Limitations
Given the ubiquitous availability of APAP in the U.S. OTC and prescription markets, significant challenges exist in accurately assessing APAP exposure in any research setting, and our results, like those of all past epidemiologic studies of APAP use, are limited by potential exposure misclassification bias. First, we could not account for OTC (or prescription) APAP use not recorded as a pharmacy claim; therefore, the calculated APAP dosages are likely understated. Since the APAP product market consists of 52% OTC products, 51 we may have accounted for only one-half of all APAP use with our data, which would suggest that our calculated APAP dosages may be understated substantially. On the other hand, our PMDD calculations were based on days supply and fill dates, and we assumed that overlapping prescriptions were used concurrently. This method could have overestimated the PMDD in some instances. Overestimates could also have occurred for APAP-opioid combinations that were diverted and not consumed by the recipient. About 95% of the APAP prescription products used in the present study were APAP-opioid combinations, many of which would be prescribed "as needed," and this use pattern could overstate the actual APAP dosages consumed. For these APAP has a narrow therapeutic-to-toxic ratio, and the U.S. Food and Drug Administration (FDA) advisory committees have recommended lowering the maximum daily dosage from the current 4 gm per day to 3.25 gm per day. 2,44 Although this change has not yet been implemented, dosage-lowering efforts are being undertaken by some manufacturers of APAP products. For example, Johnson & Johnson's McNeil Consumer Healthcare Division recently announced lowering the maximum recommended daily dosage on its OTC APAP product label to 3 gm per day. 45 Our results demonstrate an increased renal disease risk at daily dosages above and below 4 gm and therefore offer support to lowering the dosage; however, further research is needed to confirm the safety of the potential new dosage limit.
The FDA also recently mandated limiting the maximum APAP strength in each tablet of prescription-combination products to 325 mg. 46 We previously tested the potential impact of this policy and found that if the dose is limited to 325 mg, the proportion of APAP overusers would reduce by more than one-half. 19 After more robust confirmatory evidence has been acquired, such regulatory changes to APAP products should be considered to help reduce acute APAP overuse and possible resultant renal disease.
Improved labeling of APAP products could also be considered as a means of preventing APAP overuse. Some examples are clear instructions on the maximum number of tablets/doses that can be consumed in a day, enhancing the prominence of the word "acetaminophen" instead of using abbreviations such as APAP, and including warnings about APAP toxicity and concurrent use with other acetaminophen-containing products. 2 Cumulative dosages of more than 1 kg and longer durations of APAP use (more than 30 days) were not associated with renal disease in our sample. While insufficient statistical power could be a reason for this finding, some alternative explanations should be considered. A previous study of trends in APAP use and overuse, also based on IMS claims data, found that 1% of APAP users in calendar year 2008 (1,707 APAP users) had annual cumulative APAP dosage greater than 1 kg. 19 Of these, 853 individuals had pharmacy claims for more than 13 kg of APAP, which is approximately equivalent to using 35 gm per day for 365 days. Similar patterns were observed in previous years from 2001-2007. However, in spite of using the same claims database, we found few cases of renal disease among individuals with chronic APAP use in the present study. The fact that these persons prescribed cumulative APAP did not meet our case definition lends some support to the lack of an association for long-term chronic use.
The majority of the existing evidence based on clinical and some epidemiologic studies of chronic APAP use supports a lack of association between cumulative lifetime APAP dosage and renal disease. 12,14,17,47,48 One clinical trial reported APAP to be effective and well tolerated among 287 osteoarthritis per day. Lowering the recommended maximum daily APAP dosage, improved labeling of APAP products, restructuring pharmacy benefits to deter high daily APAP or APAP-opioid combination use, and regulatory changes to APAP products should be considered to reduce potential APAP overuse and resultant renal disease. In keeping with published clinical and epidemiologic evidence, we found that the renal disease risk decreased with longer look-back periods, and chronic use was not associated with renal disease. However, our interpretation is limited by potentially incomplete APAP exposure ascertainment, and further research is warranted to establish the risks and safety of APAP consumption. reasons, we used the terms "potential" maximum and average daily dosages; however, this potential overestimation of dosage is likely a smaller concern relative to the possible underestimation due to unrecorded OTC APAP use. For these reasons, our observed renal disease risk at dosages up to 4 grams per day should be interpreted with caution, and further studies with complete information on APAP dosages and genetic predisposition are warranted to confirm this risk at lower dosages of APAP.
Second, although the exact extent of OTC APAP use captured in this data source cannot be determined, there undoubtedly were a nontrivial number of persons misclassified as nonexposed when they used OTC APAP exclusively. If this misclassification was nondifferential (i.e., equal proportions of cases and controls misclassified), our results would be biased towards the null. It is possible that the cases with higher comorbidity burden would have more access to providers, and this channeling and differential misclassification would bias the results away from the null. We recognize that accurately assessing APAP exposure is challenging, and our data offer a new perspective on the associations between APAP use and renal disease with some advantages over past survey-based epidemiologic studies that are subject to recall bias and an inability to accurately identify the dates and dosages of use.
Third, since our data source did not include clinical measures, we exclusively used ICD-9-CM codes to define renal disease cases. Given the low sensitivity (35.4%) of claims-based ICD-9-CM diagnosis codes for acute renal failure, we might not have captured all true cases with renal failure. 22 Fourth, in the absence of a well-established definition of "chronicity" of APAP use, our time window of 1 year was somewhat arbitrary but has been used by a clinical trial that studied the adverse events of long-term APAP use. 47 Fifth, since we required the cases in our sample to be free from a diagnosis of renal disease and related conditions only in the pre-index year, the incident cases in our study are actually "incident episodes" of renal disease. However, given the relatively rare occurrence of renal disease, its chronicity, and the regular follow-up care, this methodological decision may not have a substantial impact on study results. Sixth, since this was an insurance database, we did not have data on factors such as body-mass index, obesity, smoking status, ethnicity, and educational level, and we cannot exclude the possibility of omitted variable bias or the possibility of coding errors that could affect our exposure measures as well as our outcome and covariate measures. Finally, limitations typical of a case-control design should be considered. Specifically, we could not directly calculate incidence ratios of renal disease, and the ORs obtained are only approximations of the relative risk.