- Email: [email protected]
Medication use and survival in diabetic patients with kidney cancer: A population-based cohort study
Pharmacological Research 113 (2016) 468–474
Contents lists available at ScienceDirect
Pharmacological Research journal homepage: www.elsevier.com/locate/yphrs
Perspective
Medication use and survival in diabetic patients with kidney cancer: A population-based cohort study Madhur Nayan a , Erin M. Macdonald b , David N. Juurlink b,c , Peter C. Austin b,d,e , Antonio Finelli a , Girish S. Kulkarni a , Robert J. Hamilton a,∗ , for the Canadian Drug Safety and Effectiveness Research Network (CDSERN) a Division of Urology, Departments of Surgery and Surgical Oncology, Princess Margaret Cancer Centre, University Health Network and the University of Toronto, Toronto, Canada b Institute for Clinical Evaluative Sciences, Toronto, Canada c Department of Internal Medicine, Sunnybrook Health Sciences Centre, University of Toronto, Toronto, Canada d Institute of Health Management, Policy and Evaluation, University of Toronto, Toronto, Canada e Schulich Heart Research Program, Sunnybrook Research Institute, Toronto, Canada
a r t i c l e
i n f o
Article history: Received 21 June 2016 Received in revised form 12 September 2016 Accepted 23 September 2016 Available online 24 September 2016 Keywords: Kidney neoplasms Diabetes mellitus Hydroxymethylglutaryl-CoA reductase inhibitors Anti-inflammatory agents Non-steroidal Metformin Survival
a b s t r a c t Survival rates in kidney cancer have improved little over time, and diabetes may be an independent risk factor for poor survival in kidney cancer. We sought to determine whether medications with putative anti-neoplastic properties (statins, metformin and non-steroidal anti-inflammatory drugs (NSAIDs)) are associated with survival in diabetics with kidney cancer. We conducted a population-based cohort study utilizing linked healthcare databases in Ontario, Canada. Patients were aged 66 or older with newly diagnosed diabetes and a subsequent diagnosis of incident kidney cancer. Receipt of metformin, statins or NSAIDs was defined using prescription claims. The primary outcome was all-cause mortality and the secondary outcome was cancer-specific mortality. We used multivariable Cox proportional hazard regression, with medication use modeled with time-varying and cumulative exposure analyses to account for intermittent use. During the 14-year study period, we studied 613 patients. Current statin use was associated with a markedly reduced risk of death from any cause (adjusted hazard ratio 0.74; 95% CI 0.59–0.91) and death due to kidney cancer (adjusted hazard ratio 0.71; 95% CI 0.51–0.97). However, survival was not associated with current use of metformin or NSAIDs, or cumulative exposure to any of the medications studied. Among diabetic patients with kidney cancer, survival outcomes are associated with active statin use, rather than total cumulative use. These findings support the use of randomized trials to confirm whether diabetics with kidney cancer should be started on a statin at the time of cancer diagnosis to improve survival outcomes. © 2016 Elsevier Ltd. All rights reserved.
1. Introduction Kidney cancer is the most lethal and third most common urological malignancy, with an estimated 61,650 new cases and 14,080 deaths in the United States in 2015 [1]. Moreover, the incidence of kidney cancer has been rising in most countries [2], most likely because of greater use of diagnostic imaging and increasing rates
∗ Corresponding author at: Division of Urology, Department of Surgery, Princess Margaret Cancer Centre, University Health Network, 610 University Ave 3-130, Toronto, Ontario, M5G 2M9, Canada. E-mail address: [email protected] (R.J. Hamilton). http://dx.doi.org/10.1016/j.phrs.2016.09.027 1043-6618/© 2016 Elsevier Ltd. All rights reserved.
of obesity and hypertension, established risk factors for the disease [3]. Despite an increase in incidence, survival rates have improved only marginally [4]. Furthermore, several studies demonstrate that patients with kidney cancer and pre-existing diabetes have poorer survival than those without diabetes [5,6]. In light of the increasing prevalence of diabetes [7] and studies suggesting that diabetes may be an independent risk factor for kidney cancer [8,9], research into improving outcomes in diabetics with kidney cancer is an increasingly important topic in oncology. Several common medications have recently garnered interest for their putative anti-neoplastic effects, most notably statins [10], metformin [11,12] and non-steroidal anti-inflammatory drugs
M. Nayan et al. / Pharmacological Research 113 (2016) 468–474
469
Fig. 1. Cohort of interest: patients with incident kidney cancer following diagnosis of incident diabetes.
(NSAIDs) [13]. To date, however, limited data exist on the influence of these medications in kidney cancer [14–25]. We sought to evaluate the association between use of statins, metformin, and NSAIDs and survival in patients with incident kidney cancer in the setting of type 2 diabetes. 2. Methods 2.1. Setting We conducted a population-based retrospective cohort study of Ontarians aged 66 or older with kidney cancer in the setting of diabetes. This study was approved by the Research Ethics Board of Sunnybrook Health Sciences Centre, Toronto, Ontario. 2.2. Data sources We used the Ontario Diabetes Database to identify patients with diabetes [26], and the Ontario Cancer Registry [27] to identify patients with kidney cancer (ICD-9: 189.0; ICD-10: C64). We identified medication use through prescription claims of the Ontario Drug Benefit Database [28], which contains comprehensive records of prescription drugs dispensed to all Ontario residents aged 65 or older. We excluded subjects during their first year of eligibility for prescription drug coverage (age 65) to avoid incomplete assessment of medication use. We obtained hospitalization data from the Canadian Institute for Health Information Discharge Abstract Database [29], which contains detailed clinical information regarding all hospital admissions in Ontario. We used the Ontario Health Insurance Plan database to identify claims for physician services, and obtained basic demographic data and date of death from the Registered Persons Database, a registry of all Ontario residents eligible for health insurance. These databases were linked in an anonymous fashion using encrypted health card numbers. Details regarding all databases used and their validity are provided in Supplemental Methods 1. 2.3. Study participants We identified patients with incident diabetes aged 66 or older. Among these patients, we studied those with kidney cancer first diagnosed after the diagnosis of diabetes (Fig. 1). All subjects had universal access to physician services, hospital care and prescription drug coverage. We accrued patients from April 1st, 1998, following them until December 31st, 2012 (for kidney cancer-specific mortality) and December 31st, 2014 (for all-cause mortality). These dates were based on the most recent update of the database used for each outcome. We deemed study subjects to have localized kidney cancer at presentation if they underwent surgery (OHIP billing codes S411, S412, S413, S415, S416), radiofrequency ablation (OHIP billing code J069) or cryotherapy (OHIP billing code S400) as the first treatment following diagnosis of kidney cancer. Other subjects were deemed to have advanced disease if they received immunotherapy (OHIP billing codes G381, G281, G345, G359, G075, G390, G388)
or no intervention as their first treatment following diagnosis of kidney cancer. This classification is consistent with other studies conducted in Ontario [30,31]. We excluded patients who died within 30 days of intervention (surgery, radiofrequency ablation, or cryotherapy), since this is unlikely to reflect the effects of medication exposure. We also excluded patients with any diagnosis of malignancy (excluding non-melanoma skin cancers) prior to their kidney cancer diagnosis in order to exclude patients whose overall survival may be influenced by a pre-existing malignancy. Finally, we excluded patients whose histology indicated a primary malignancy other than kidney cancer. 2.4. Exposure assessment We quantified medication exposure from the diagnosis of kidney cancer to the end of follow-up using prescription dates and the number of days supplied in each, as done previously [32,33]. This allowed for calculation of the duration of cumulative exposure for each day of follow-up, as well as accounting for periods of intermittent use. 2.5. Outcome assessment The primary outcome was all-cause mortality and the secondary outcome was kidney cancer-specific mortality. For each outcome, patients were followed until their date of last contact with health services, death or the end of the study period, whichever occurred first. 2.6. Statistical methods We conducted time-to-event analyses using multivariable Cox proportional hazard regression to estimate the effect of drug exposure on the risk of the primary and secondary outcomes. For the secondary outcome, we focused on the cause-specific hazard of death from kidney cancer, accounting for the competing risk of death from other causes. Covariates in the model were selected a priori and include age at kidney cancer diagnosis, sex, duration of diabetes, comorbidity (defined by the Johns Hopkins Adjusted Clinical Groups score [34]), year of kidney cancer diagnosis (to account for temporal changes in exposure and outcomes), disease stage, rurality, socioeconomic status and exposure to each medication after kidney cancer diagnosis. We evaluated the association between medication exposure and outcomes in two ways. First, we studied medication use as a time-varying covariate denoting whether or not the subject was actively receiving the medication of interest, determined using prescription claims. Second, we examined cumulative medication use as a time-varying covariate (Supplemental Methods 2). Both of these models account for intermittent use; however, they evaluate different hypotheses. In the time-varying analyses, a patient can contribute survival time to both the exposed and unexposed group, based on the date and duration of their prescription, offsetting the selection bias associated with a simple comparison of
470
M. Nayan et al. / Pharmacological Research 113 (2016) 468–474
Fig. 2. Cohort derivation.
users and non-users. As such, the time-varying analyses evaluate the hypothesis of whether current use of a medication, as a binary variable, is associated with survival. In contrast, cumulative use analyses evaluate whether increasing use is associated with survival. This approach method mitigates confounding by indication because it compares patients with varying degrees of total use [35]. For the analysis of cumulative duration of use, we used fractional polynomials [36] to determine the function form that best described the relationship between cumulative exposure to each medication and the hazard of mortality. All statistical analyses were performed using SAS (version 9.3; SAS Institute, Cary, NC) and used a two-sided p value of 0.05 as the threshold for statistical significance. 3. Results 3.1. Cohort characteristics Over the 14-year study period, we identified 815 patients aged 66 or older with incident diabetes and a subsequent diagnosis of kidney cancer. Of these, 613 patients met inclusion criteria (Fig. 2). The characteristics of the cohort are described in Table 1. Of these, 409 (67%) died during follow-up, including 194 (32%) who died from kidney cancer. Medication use following kidney cancer diagnosis is shown in Supplemental Table 1. 3.2. All-cause mortality In the primary analysis, current use of statin therapy was associated with a significant reduction in all-cause mortality (adjusted hazard ratio 0.74; 95% CI 0.59–0.91) (Table 2). Other characteristics independently associated with all-cause mortality included sex, age at kidney cancer diagnosis, disease stage, socioeconomic status and rural residence (Table 2). In contrast, current use of NSAIDs or metformin was not significantly associated with all-cause mortality. For no study medication was cumulative exposure significantly associated with all-cause mortality (Supplemental Table 2). 3.3. Kidney cancer-specific mortality In the primary analysis, current use of statins was associated with a significant reduction in kidney cancer-specific mortality (adjusted hazard ratio 0.71; 95% CI 0.51–0.97) (Table 3). No
such association was observed with any other study medication. Other characteristics independently associated with kidney cancerspecific mortality included duration of diabetes, disease stage and rural residence (Table 3). Cumulative duration of medication use was not associated with kidney-cancer mortality for any of the study medications (Supplemental Table 2). 4. Discussion This population-based study of patients with type 2 diabetes found that in a time-varying analysis, current use of a statin following a diagnosis of kidney cancer was associated with a 26% reduced risk of all-cause mortality and a 29% reduced risk of kidney cancerspecific mortality. However, an association between increasing use of statins and survival outcomes was not observed. In no analysis was exposure to receipt of NSAIDs or metformin associated with improved survival. One possible explanation for our results demonstrating a protective association with current use of statin but not increasing cumulating use is that statins have a short half-life (0.5–3 h), and the putative anti-neoplastic effects may be limited to the duration that statins are present in the circulation. This may be similar to the effects observed upon the discontinuation of tyrosine kinase inhibitors, such as sunitinib (half-life 40–60 h) or pazopanib (halflife 31 h), which results in accelerated tumour growth and tumour flare in kidney cancer patients [37]. In this manner, the outcomes may be more related to the current use of a statin (time-varying), rather than the anti-neoplastic effects of statins cumulating with each additional use (cumulative use). Importantly, an alternate explanation for our finding of no association between increasing statin use and survival outcomes is that cumulative use analyses assign equal weight to remote exposures and more recent exposures, which may not be appropriate. While our study is the first to evaluate the association between statin exposure and survival outcomes in diabetic patients with kidney cancer, previous studies evaluating statins in non-diabetics with kidney cancer yield conflicting results [16,21–24]. The majority of these defined statin users based on medication use at the time of surgery. Classifying exposure in this manner does not account for intermittent use and can therefore lead to bias because medication use can change over time. Indeed, one study found that the association between statin use and survival outcomes in kidney cancer differed when comparing use at the time of surgery with the
M. Nayan et al. / Pharmacological Research 113 (2016) 468–474
471
Table 1 Baseline Patient Characteristics (n = 613). Characteristics
All patients (n = 613)
Male gender (No (%)) Age at diabetes diagnosis, years (mean (SD)) Age at KC diagnosis, years (mean (SD)) Duration of diabetes to KC diagnosis, years (mean (SD)) Time from KC diagnosis to end of follow-up, years (mean (SD)) Primary Treatment (No (%)) Surgery or Radiofrequency ablation No primary treatment or Immunotherapy a Comorbidity score (No (%)) Low Intermediate High b SES status (No (%)) Income Quintile 1 (lowest) Income Quintile 2 Income Quintile 3 Income Quintile 4 Income Quintile 5 (highest) Rural residence (No (%)) KC-specific death (No (%)) Overall mortality (No (%))
367 (59.9) 73.8 (6.0) 77.9 (6.4) 4.1 (3.3) 3.3 (3.4) 272 (44.4) 341 (55.6) 22 (3.6) 90 (14.7) 501 (81.7) 132 (21.5) 124 (20.2) 130 (21.2) 135 (22.0) 92 (15.0) 76 (12.4) 194 (31.6) 409 (66.7)
Abbreviations: KC, kidney cancer; SES, socioeconomic status. a Comorbidity scores were calculated by using Johns Hopkins Adjusted Clinical Groups Case-Mix System assigning a specific weight to each adjusted diagnostic group (low, weighted adjusted diagnostic group score 5 or lower; intermediate, 6–9; high, 10 or higher). b Income quintiles from median income in neighborhoods from 1 (low) to 5 (high). Table 2 Cox proportional hazard models evaluating all-cause mortality (time-varying analysis denoting effect of current exposure). Characteristics
Univariate
Multivariable
HR (95% CI)
HR (95% CI)
1.03 (0.85–1.26) 1.06 (1.05–1.08) 1.03 (0.99–1.06) 4.79 (3.83–5.99)
1.25 (1.02–1.54) 1.03 (1.01–1.05) 0.96 (0.93–1.00) 4.42 (3.49–5.60)
1.00 (reference) 1.05 (0.60–1.84) 0.81 (0.48–1.37)
1.00 (reference) 1.26 (0.71–2.25) 1.05 (0.62–1.79)
Socioeconomic status(lowest) Income quintile 1 Income quintile 2 Income quintile 3 Income quintile 4 Income quintile 5 (highest) Rural residence Year of KC diagnosis
1.00 (reference) 0.78 (0.58–1.04) 0.78 (0.59–1.04) 0.66 (0.49–0.88) 0.67 (0.49–0.93) 1.59 (1.21–2.08) 1.02 (0.98–1.04)
1.00 (reference) 0.85 (0.64–1.15) 0.89 (0.66–1.19) 0.68 (0.50–0.92) 0.65 (0.47–0.90) 1.56 (1.18–2.07) 1.00 (0.97–1.04)
Medication exposure after KC diagnosis Statins NSAIDs Metformin Insulin Insulin secretagogue Thiazolidinedione
0.74 (0.61–0.91) 1.06 (0.81–1.38) 0.99 (0.78–1.25) 1.33 (0.71–2.50) 1.33 (1.02–1.72) 1.44 (0.46–4.47)
0.74 (0.59–0.91) 1.04 (0.79–1.36) 1.09 (0.85–1.40) 1.87 (0.98–3.59) 1.16 (0.88–1.53) 2.88 (0.90–9.28)
Male gender Age at KC diagnosis (years) Duration of diabetes to KC diagnosis (years) Disease stage (advanced vs. localized) a Comorbidity score Low Intermediate High b
Abbreviations: IQR, interquartile range; KC, kidney cancer; SES, socioeconomic status. a Comorbidity scores were calculated by using Johns Hopkins Adjusted Clinical Groups Case-Mix System assigning a specific weight to each adjusted diagnostic group (low, weighted adjusted diagnostic group score 5 or lower; intermediate, 6–9; high, 10 or higher). b Income quintiles from median income in neighborhoods from 1 (low) to 5 (high).
time-varying analysis [16]. Furthermore, the results of these studies may not be generalized to non-diabetics; a recent study found that statin use among diabetic patients with pancreatic cancer was associated with significantly improved overall survival while there was no benefit of statin use in non-diabetics [38]. Although we found that current use of statins was associated with improved survival outcomes, use of metformin, another putative anti-neoplastic medication, was not. Three previous studies in kidney cancer have demonstrated similar results; [19,25,39] however, these studies classified medication exposure based on use
at the time of surgery and did not account for intermittent use. While cumulative metformin use has been associated with survival outcomes in prostate cancer [32], it may be possible that the anti-neoplastic effect of metformin on survival outcomes may be dependent on the type of cancer. Indeed, another study found no association between cumulative metformin exposure and survival outcomes in patients with breast cancer [33]. Similarly, it is possible that NSAIDs are beneficial in some malignancies and not in others, as shown previously [40]. To date, the association between use of NSAIDs and kidney cancer survival out-
472
M. Nayan et al. / Pharmacological Research 113 (2016) 468–474
Table 3 Cox proportional hazard models evaluating kidney-cancer specific mortality (time-varying analysis denoting current exposure). Characteristics
Male gender Age at KC diagnosis (years) Duration of diabetes–KC diagnosis (years) Disease stage (advanced vs. localized) a Comorbidity score Low Intermediate High b Socioeconomic status Income quintile 1 (lowest) Income quintile 2 Income quintile 3 Income quintile 4 Income quintile 5 (highest) Rural residence Year of KC diagnosis Medication exposure after KC diagnosis Statins NSAIDs Metformin Insulin Insulin secretagogue Thiazolidinedione
Univariate
Multivariable
HR (95% CI)
HR (95% CI)
0.93 (0.70–1.24) 1.04 (1.02–1.06) 1.00 (0.96–1.05) 5.67 (4.00–8.04)
1.08 (0.81–1.46) 1.01 (0.98–1.03) 0.93 (0.89–0.98) 5.64 (3.92–8.11)
1.00 (reference) 1.30 (0.61–2.78) 0.75 (0.37–1.54)
1.00 (reference) 1.74 (0.80–3.82) 1.09 (0.52–2.28)
1.00 (reference) 0.73 (0.48–1.11) 0.72 (0.47–1.09) 0.67 (0.44–1.09) 0.69 (0.44–1.09) 1.90 (1.32–2.72) 1.04 (0.99–1.08)
1.00 (reference) 0.76 (0.49–1.17) 0.79 (0.51–1.21) 0.69 (0.44–1.06) 0.64 (0.40–1.02) 1.80 (1.23–2.62) 1.05 (0.99–1.10)
0.68 (0.50–0.91) 0.96 (0.65–1.42) 1.12 (0.80–1.55) 0.95 (0.30–2.96) 1.12 (0.75–1.66) 1.84 (0.46–7.40)
0.71 (0.51–0.97) 0.98 (0.66–1.46) 1.25 (0.88–1.77) 1.14 (0.36–3.65) 1.05 (0.69–1.58) 2.48 (0.58–10.65)
Abbreviations: KC, kidney cancer; SES, socioeconomic status. a Comorbidity scores were calculated by using Johns Hopkins Adjusted Clinical Groups Case-Mix System assigning a specific weight to each adjusted diagnostic group (low, weighted adjusted diagnostic group score 5 or lower; intermediate, 6–9; high, 10 or higher). b Income quintiles from median income in neighborhoods from 1 (low) to 5 (high).
comes has not been studied; however, our results do not support their use for this purpose. Our study has several strengths. First, we used a population based-cohort and our study is the largest to date to evaluate putative anti-neoplastic medications among diabetics with kidney cancer. Second, the comprehensive nature of prescription claims data permitted more robust analyses evaluating medication exposure, rather than simply classifying exposure based on use at the time of diagnosis. Third, we simultaneously evaluated multiple putative anti-neoplastic medications which allowed us to evaluate the independent effect of each medication, while controlling for the possible effects of the others. Furthermore, our results demonstrating that active use of neither metformin nor NSAIDs were associated with survival outcomes suggests against a healthy user bias behind the protective association observed for statins. Fourth, by using a cohort of incident diabetics, we were able to evaluate a more homogenous population and minimize selection bias that may be attributable to varying durations of diabetes. Finally, while randomized trials provide the strongest level of evidence, the profile of patients included in trials differ from those in real-world settings. The results of this study, on the other hand, provide insight on the real-world association of the use of putative anti-neoplastic medications on survival outcomes among diabetics with kidney cancer. Some limitations of our study merit emphasis. First, several factors govern the prescribing of medications we studied, and certain details (such as tumour characteristics) were not available. However, these are not true confounders as they are not associated with both exposure and outcome. As such, controlling for these factors will not improve the estimation of the exposure-outcome relationship; rather, it would lead to overadjustment [41]. Additionally, many studies have shown that tumour stage was not significantly different in users vs. non-users of statins or metformin [16,19,21,23,24,39]. While there is conflicting evidence on the association of obesity and kidney cancer survival outcomes [42–45], data on obesity was not available in the databases used. Similarly,
data on glycemic control was not available. Second, we classified disease stage based on the intervention following cancer diagnosis and there may be patients with localized disease that were managed with observation, and patients with advanced disease that underwent a cytoreductive nephrectomy. However, the hazard ratios associated with our method of classifying disease stage suggest against a strong effect of misclassficiation. Third, because we used prescription data as a surrogate for medication use, the actual and estimated exposures may not accord if patients are non-compliant. However, the analyses in this study accounted for intermitted exposure and therefore controlled for lapses between prescription claims that would be expected in non-compliant patients. Fourth, while administrative data are susceptible to misclassification bias, several of the databases used in this study have been validated (Supplemental Methods 1). Fifth, some NSAIDs are available over the counter and it is possible that some exposure was not captured due to this. However, patients in our study have a financial incentive to obtain these drugs by prescription, particularly for long-term use, and previous research shows that administrative data represent a valid indicator of NSAIDs exposure [46]. Finally, the results of this study are limited to incident diabetic patients aged 66 or older with a subsequent diagnosis of kidney cancer. While the generalizability of our findings is unknown, the results provide impetus to study statins in younger diabetic patients with kidney cancer. Despite these limitations, this is the first study to simultaneously evaluate statins, metformin, and NSAIDs in a large population of type 2 diabetic patients with kidney cancer and suggests that statins may have a role in kidney cancer therapy among diabetics.
5. Conclusion Among type 2 diabetic patients, current statin use following a diagnosis of kidney cancer was associated with a significant improvement in all-cause mortality and kidney cancer-specific mortality. However, the benefits of statins were not dependent
M. Nayan et al. / Pharmacological Research 113 (2016) 468–474
on increasing total cumulative use. These findings provide support for randomized trials evaluating the potential effect of statins on survival in diabetic patients with kidney cancer. Funding This study was supported in part by a grant from the Canadian Drug Safety & Effectiveness Research Network (CDSERN), and by the Institute for Clinical Evaluative Sciences (ICES), a non-profit research institute sponsored by the Ontario Ministry of Health and Long-Term Care (MOHLTC). Dr. Juurlink is supported by the Eaton Scholar award, Department of Medicine, University of Toronto. The opinions, results and conclusions reported in this paper are those of the authors and are independent from the funding sources. No endorsement by ICES or the Ontario MOHLTC is intended or should be inferred. Competing interests’ None. Acknowledgements We thank Brogan Inc., Ottawa for use of their Drug Product and Therapeutic Class Database. Parts of this material are based on data and information compiled and provided by CIHI. However, the analyses, conclusions, opinions and statements expressed herein are those of the author, and not necessarily those of CIHI. The authors thank Qing Li, analyst at the Institute of Clinical and Evaluative Sciences, for preparing the data. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.phrs.2016.09. 027. References [1] R.L. Siegel, K.D. Miller, A. Jemal, Cancer statistics, CA: Cancer J. Clin. 65 (1) (2015) 5–29. [2] A. Znaor, J. Lortet-Tieulent, M. Laversanne, A. Jemal, F. Bray, International variations and trends in renal cell carcinoma incidence and mortality, Eur. Urol. (2014). [3] W.H. Chow, L.M. Dong, S.S. Devesa, Epidemiology and risk factors for kidney cancer, Nat. Rev. Urol. 7 (5) (2010) 245–257. [4] P. De, M.C. Otterstatter, R. Semenciw, L.F. Ellison, L.D. Marrett, D. Dryer, Trends in incidence, mortality, and survival for kidney cancer in Canada, Cancer Causes Control (2014) 1986–2007. [5] A. Vavallo, S. Simone, G. Lucarelli, M. Rutigliano, V. Galleggiante, G. Grandaliano, L. Gesualdo, M. Campagna, M. Cariello, E. Ranieri, G. Pertosa, G. Lastilla, F.P. Selvaggi, P. Ditonno, M. Battaglia, Pre-existing type 2 diabetes mellitus is an independent risk factor for mortality and progression in patients with renal cell carcinoma, Medicine (United States) 93 (27) (2014). [6] Y.S. Ha, W.T. Kim, S.J. Yun, S.C. Lee, W.J. Kim, Y.H. Park, S.H. Kang, S.H. Hong, S.S. Byun, Y.J. Kim, Multi-institutional analysis of localized renal cell carcinoma that demonstrates the impact of diabetic status on prognosis after nephrectomy, Ann. Surg. Oncol. 20 (11) (2013) 3662–3668. [7] G. Danaei, M.M. Finucane, Y. Lu, G.M. Singh, M.J. Cowan, C.J. Paciorek, J.K. Lin, F. Farzadfar, Y.-H. Khang, G.A. Stevens, National, regional, and global trends in fasting plasma glucose and diabetes prevalence since 1980: systematic analysis of health examination surveys and epidemiological studies with 370 country-years and 2· 7 million participants, Lancet 378 (9785) (2011) 31–40. [8] S.C. Larsson, A. Wolk, Diabetes mellitus and incidence of kidney cancer: a meta-analysis of cohort studies, Diabetologia 54 (5) (2011) 1013–1018. [9] C. Bao, X. Yang, W. Xu, H. Luo, Z. Xu, C. Su, X. Qi, Diabetes mellitus and incidence and mortality of kidney cancer: a meta-analysis, J. Diabetes Complications 27 (4) (2013) 357–364. [10] S.F. Nielsen, B.G. Nordestgaard, S.E. Bojesen, Statin use and reduced cancer-related mortality, New Engl. J. Med. 367 (19) (2012) 1792–1802. [11] G.W. Landman, N. Kleefstra, K.J. van Hateren, K.H. Groenier, R.O. Gans, H.J. Bilo, Metformin associated with lower cancer mortality in type 2 diabetes ZODIAC-16, Diabetes Care 33 (2) (2010) 322–326.
473
[12] M. Nayan, B. Bhindi, L.Y. Julie, T. Hermanns, A. Mohammed, R.J. Hamilton, A. Finelli, M.A. Jewett, A.R. Zlotta, N.E. Fleshner, The effect of metformin on cancer-specific survival outcomes in diabetic patients undergoing radical cystectomy for urothelial carcinoma of the bladder, in: Urologic Oncology: Seminars and Original Investigations, Elsevier, 2015, p. 386. e387-386. e313. [13] X. Liao, P. Lochhead, R. Nishihara, T. Morikawa, A. Kuchiba, M. Yamauchi, Y. Imamura, Z.R. Qian, Y. Baba, K. Shima, Aspirin use, tumor PIK3CA mutation, and colorectal-cancer survival, New Engl. J. Med. 367 (17) (2012) 1596–1606. [14] B.I. Rini, V. Weinberg, S. Dunlap, A. Elchinoff, N. Yu, R. Bok, J. Simko, E.J. Small, Maximal COX-2 immunostaining and clinical response to celecoxib and interferon alpha therapy in metastatic renal cell carcinoma, Cancer 106 (3) (2006) 566–575. [15] N. Shinohara, A. Kumagai, K. Kanagawa, S. Maruyama, T. Abe, A. Sazawa, K. Nonomura, Multicenter phase II trial of combination therapy with meloxicam, a COX-2 inhibitor, and natural interferon-␣ for metastatic renal cell carcinoma, Jpn. J. Clin. Oncol. 39 (11) (2009) 720–726. [16] R.J. Hamilton, D. Morilla, F. Cabrera, M. Leapman, L.Y. Chen, M. Bernstein, A.A. Hakimi, V.E. Reuter, P. Russo, The association between statin medication and progression after surgery for localized renal cell carcinoma, J. Urol. 191 (4) (2014) 914–919. [17] F. Khimani, N. Shah, B.F. Curley, M.D. Newton, T.F. Hogan, M. Monga, Metformin use and renal cell cancer outcomes among patients with diabetes, J. Clin. Oncol. (2014). [18] N. Shah, F. Khimani, B.F. Curley, M.D. Newton, T.F. Hogan, S. Wen, Y. Hu, M. Monga, Metformin use and survival in diabetic patients with renal cell carcinoma (RCC), J. Clin. Oncol. (2014). [19] S. Psutka, S. Boorjian, C. Lohse, S. Stewart, M. Tollefson, B. Leibovich, R.H. Thompson, The effect of metformin in patients with surgically treated renal cell carcinoma, J. Urol. (2014) e579. [20] S.M. Haake, S.A. Brooks, E. Welsh, A. Berglund, W.J. Fulp, D.T. Chen, E.B. Haura, W.J. Sexton, P.E. Spiess, J. Pow-Sang, K. Rathmell, M.N. Fishman, Molecular subtypes of clear cell renal cell carcinoma: impact of diabetes mellitus, metformin, and immunotherapy on patient outcomes, J. Clin. Oncol. (2014). [21] S.D. Kaffenberger, O. Lin-Tsai, K.L. Stratton, T.M. Morgan, D.A. Barocas, S.S. Chang, M.S. Cookson, S.D. Herrell, J.A. Smith, P.E. Clark, Statin use is associated with improved survival in patients undergoing surgery for renal cell carcinoma, in: Urologic Oncology: Seminars and Original Investigations, Elsevier, 2015, p. 21. e11-21. e17. [22] B.R. Viers, R.H. Thompson, S.P. Psutka, C.M. Lohse, J.C. Cheville, B.C. Leibovich, M.K. Tollefson, S.A. Boorjian, The association of statin therapy with clinicopathologic outcomes and survival among patients with localized renal cell carcinoma undergoing nephrectomy, in: Urologic Oncology: Seminars and Original Investigations, Elsevier, 2015. [23] L.S. Krane, J.M. Sandberg, J.T. Rague, A.K. Hemal, Do statin medications impact renal functional or oncologic outcomes for robot-assisted partial nephrectomy? J. Endourol. 28 (11) (2014) 1308–1312. [24] M. Nayan, A. Finelli, M. Jewett, D. Juurlink, P. Austin, G. Kulkarni, R. Hamilton, Statin use and kidney cancer survival outcomes: a propensity score analysis, Urol. Oncol. (2016) http://www.ncbi.nlm.nih.gov/pubmed/27427224. [25] M. Nayan, A. Finelli, M. Jewett, D. Juurlink, P. Austin, G. Kulkarni, R. Hamilton, Metformin use and kidney cancer outcomes in diabetics: a propensity score analysis, Clin. Genitourin. Cancer (2016) http://www.ncbi.nlm.nih.gov/ pubmed/27424258. [26] J.E. Hux, F. Ivis, V. Flintoft, A. Bica, Diabetes in Ontario: determination of prevalence and incidence using a validated administrative data algorithm, Diabetes Care 25 (3) (2002) 512–516. [27] D.R. Brenner, M.C. Tammemagi, S.B. Bull, D. Pinnaduwaje, I.L. Andrulis, Using cancer registry data: agreement in cause-of-death data between the Ontario Cancer Registry and a longitudinal study of breast cancer patients, Chronic Dis. Can. 30 (1) (2009) 16–19. [28] A. Levy, B. O’Brien, C. Sellors, P. Grootendorst, D. Willison, Coding accuracy of administrative drug claims in the Ontario Drug Benefit database, Can. J. Clin. Pharmacol. 10 (2) (2003) 67. [29] D.N. Juurlink, Canadian Institute for Health Information Discharge Abstract Database: a Validation Study, Institute for Clinical Evaluative Sciences, 2006. [30] J.L. Warren, L. Barbera, K.E. Bremner, K.R. Yabroff, J.S. Hoch, M.J. Barrett, J. Luo, M.D. Krahn, End-of-life care for lung cancer patients in the United States and Ontario, J. Natl. Cancer Inst. 103 (11) (2011) 853–862. [31] R.W. Jang, M.K. Krzyzanowska, C. Zimmermann, N. Taback, S.M. Alibhai, Palliative care and the aggressiveness of end-of-life care in patients with advanced pancreatic cancer, J. Natl. Cancer Inst. 107 (3) (2015) dju424. [32] D. Margel, D.R. Urbach, L.L. Lipscombe, C.M. Bell, G. Kulkarni, P.C. Austin, N. Fleshner, Metformin use and all-cause and prostate cancer-specific mortality among men with diabetes, J. Clin. Oncol. 31 (25) (2013) 3069–3075. [33] I.C. Lega, P.C. Austin, A. Gruneir, P.J. Goodwin, P.A. Rochon, L.L. Lipscombe, Association between metformin therapy and mortality after breast cancer: a population-based study, Diabetes Care 36 (10) (2013) 3018–3026. [34] P.C. Austin, C. van Walraven, W.P. Wodchis, A. Newman, G.M. Anderson, Using the Johns Hopkins Aggregated Diagnosis Groups (ADGs) to predict mortality in a general adult population cohort in Ontario, Canada, Med. Care 49 (10) (2011) 932–939. [35] D. Margel, D. Urbach, N. Fleshner, P.C. Austin, Reply to MMJ Zanders et al, J. Clin. Oncol. 32 (7) (2014) 702. [36] W. Sauerbrei, C. Meier-Hirmer, A. Benner, P. Royston, Multivariable regression model building by using fractional polynomials: description of SAS, STATA and R programs, Comput. Statistics & Data Anal. 50 (12) (2006) 3464–3485.
474
M. Nayan et al. / Pharmacological Research 113 (2016) 468–474
[37] R. Iacovelli, F. Massari, L. Albiges, Y. Loriot, C. Massard, K. Fizazi, B. Escudier, Evidence and clinical relevance of tumor flare in patients who discontinue tyrosine kinase inhibitors for treatment of metastatic renal cell carcinoma, Eur. Urol. 68 (July (1)) (2014) 154–160. [38] Y. Nakai, H. Isayama, T. Sasaki, S. Mizuno, N. Sasahira, H. Kogure, K. Kawakubo, N. Yamamoto, K. Hirano, H. Ijichi, Clinical outcomes of chemotherapy for diabetic and nondiabetic patients with pancreatic cancer: better prognosis with statin use in diabetic patients, Pancreas 42 (2) (2013) 202–208. [39] A.A. Hakimi, L. Chen, P.H. Kim, D. Sjoberg, L. Glickman, M.R. Walker, P. Russo, The impact of metformin use on recurrence and cancer-specific survival in clinically localized high-risk renal cell carcinoma, Can. Urol. Assoc. J. 7 (11-12) (2013) E687–691. [40] P.M. Rothwell, F.G.R. Fowkes, J.F. Belch, H. Ogawa, C.P. Warlow, T.W. Meade, Effect of daily aspirin on long-term risk of death due to cancer: analysis of individual patient data from randomised trials, Lancet 377 (9759) (2011) 31–41. [41] K. Jager, C. Zoccali, A. Macleod, F. Dekker, Confounding: what it is and how to deal with it, Kidney Int. 73 (3) (2008) 256–260.
[42] S.M. Donat, E.W. Salzhauer, N. Mitra, B.V. Yanke, M.E. Snyder, P. Russo, Impact of body mass index on survival of patients with surgically treated renal cell carcinoma, J. Urol. 175 (1) (2006) 46–52. [43] A. Haferkamp, M. Pritsch, J. Bedke, N. Wagener, J. Pfitzenmaier, S. Buse, M. Hohenfellner, The influence of body mass index on the long-term survival of patients with renal cell carcinoma after tumour nephrectomy, BJU Int. 101 (10) (2008) 1243–1246. [44] A.M. Kamat, R.P. Shock, Y. Naya, C.J. Rosser, J.W. Slaton, L.L. Pisters, Prognostic value of body mass index in patients undergoing nephrectomy for localized renal tumors, Urology 63 (1) (2004) 46–50. [45] S. Steffens, C. Von Klot, H. Eggers, C. Seidel, G. Wegener, M. Schrader, M.A. Kuczyk, A.J. Schrader, The role of diabetes mellitus and obesity in localized and metastatic RCC, Eur. Urol. 11 (Suppl. 1) (2012), e1018-e1018a. [46] M.U. Yood, U.B. Campbell, K.J. Rothman, S.S. Jick, J. Lang, K.E. Wells, H. Jick, C.C. Johnson, Using prescription claims data for drugs available over-the-counter (OTC), Pharmacoepidemiol. Drug Saf. 16 (9) (2007) 961–968.
Contents lists available at ScienceDirect
Pharmacological Research journal homepage: www.elsevier.com/locate/yphrs
Perspective
Medication use and survival in diabetic patients with kidney cancer: A population-based cohort study Madhur Nayan a , Erin M. Macdonald b , David N. Juurlink b,c , Peter C. Austin b,d,e , Antonio Finelli a , Girish S. Kulkarni a , Robert J. Hamilton a,∗ , for the Canadian Drug Safety and Effectiveness Research Network (CDSERN) a Division of Urology, Departments of Surgery and Surgical Oncology, Princess Margaret Cancer Centre, University Health Network and the University of Toronto, Toronto, Canada b Institute for Clinical Evaluative Sciences, Toronto, Canada c Department of Internal Medicine, Sunnybrook Health Sciences Centre, University of Toronto, Toronto, Canada d Institute of Health Management, Policy and Evaluation, University of Toronto, Toronto, Canada e Schulich Heart Research Program, Sunnybrook Research Institute, Toronto, Canada
a r t i c l e
i n f o
Article history: Received 21 June 2016 Received in revised form 12 September 2016 Accepted 23 September 2016 Available online 24 September 2016 Keywords: Kidney neoplasms Diabetes mellitus Hydroxymethylglutaryl-CoA reductase inhibitors Anti-inflammatory agents Non-steroidal Metformin Survival
a b s t r a c t Survival rates in kidney cancer have improved little over time, and diabetes may be an independent risk factor for poor survival in kidney cancer. We sought to determine whether medications with putative anti-neoplastic properties (statins, metformin and non-steroidal anti-inflammatory drugs (NSAIDs)) are associated with survival in diabetics with kidney cancer. We conducted a population-based cohort study utilizing linked healthcare databases in Ontario, Canada. Patients were aged 66 or older with newly diagnosed diabetes and a subsequent diagnosis of incident kidney cancer. Receipt of metformin, statins or NSAIDs was defined using prescription claims. The primary outcome was all-cause mortality and the secondary outcome was cancer-specific mortality. We used multivariable Cox proportional hazard regression, with medication use modeled with time-varying and cumulative exposure analyses to account for intermittent use. During the 14-year study period, we studied 613 patients. Current statin use was associated with a markedly reduced risk of death from any cause (adjusted hazard ratio 0.74; 95% CI 0.59–0.91) and death due to kidney cancer (adjusted hazard ratio 0.71; 95% CI 0.51–0.97). However, survival was not associated with current use of metformin or NSAIDs, or cumulative exposure to any of the medications studied. Among diabetic patients with kidney cancer, survival outcomes are associated with active statin use, rather than total cumulative use. These findings support the use of randomized trials to confirm whether diabetics with kidney cancer should be started on a statin at the time of cancer diagnosis to improve survival outcomes. © 2016 Elsevier Ltd. All rights reserved.
1. Introduction Kidney cancer is the most lethal and third most common urological malignancy, with an estimated 61,650 new cases and 14,080 deaths in the United States in 2015 [1]. Moreover, the incidence of kidney cancer has been rising in most countries [2], most likely because of greater use of diagnostic imaging and increasing rates
∗ Corresponding author at: Division of Urology, Department of Surgery, Princess Margaret Cancer Centre, University Health Network, 610 University Ave 3-130, Toronto, Ontario, M5G 2M9, Canada. E-mail address: [email protected] (R.J. Hamilton). http://dx.doi.org/10.1016/j.phrs.2016.09.027 1043-6618/© 2016 Elsevier Ltd. All rights reserved.
of obesity and hypertension, established risk factors for the disease [3]. Despite an increase in incidence, survival rates have improved only marginally [4]. Furthermore, several studies demonstrate that patients with kidney cancer and pre-existing diabetes have poorer survival than those without diabetes [5,6]. In light of the increasing prevalence of diabetes [7] and studies suggesting that diabetes may be an independent risk factor for kidney cancer [8,9], research into improving outcomes in diabetics with kidney cancer is an increasingly important topic in oncology. Several common medications have recently garnered interest for their putative anti-neoplastic effects, most notably statins [10], metformin [11,12] and non-steroidal anti-inflammatory drugs
M. Nayan et al. / Pharmacological Research 113 (2016) 468–474
469
Fig. 1. Cohort of interest: patients with incident kidney cancer following diagnosis of incident diabetes.
(NSAIDs) [13]. To date, however, limited data exist on the influence of these medications in kidney cancer [14–25]. We sought to evaluate the association between use of statins, metformin, and NSAIDs and survival in patients with incident kidney cancer in the setting of type 2 diabetes. 2. Methods 2.1. Setting We conducted a population-based retrospective cohort study of Ontarians aged 66 or older with kidney cancer in the setting of diabetes. This study was approved by the Research Ethics Board of Sunnybrook Health Sciences Centre, Toronto, Ontario. 2.2. Data sources We used the Ontario Diabetes Database to identify patients with diabetes [26], and the Ontario Cancer Registry [27] to identify patients with kidney cancer (ICD-9: 189.0; ICD-10: C64). We identified medication use through prescription claims of the Ontario Drug Benefit Database [28], which contains comprehensive records of prescription drugs dispensed to all Ontario residents aged 65 or older. We excluded subjects during their first year of eligibility for prescription drug coverage (age 65) to avoid incomplete assessment of medication use. We obtained hospitalization data from the Canadian Institute for Health Information Discharge Abstract Database [29], which contains detailed clinical information regarding all hospital admissions in Ontario. We used the Ontario Health Insurance Plan database to identify claims for physician services, and obtained basic demographic data and date of death from the Registered Persons Database, a registry of all Ontario residents eligible for health insurance. These databases were linked in an anonymous fashion using encrypted health card numbers. Details regarding all databases used and their validity are provided in Supplemental Methods 1. 2.3. Study participants We identified patients with incident diabetes aged 66 or older. Among these patients, we studied those with kidney cancer first diagnosed after the diagnosis of diabetes (Fig. 1). All subjects had universal access to physician services, hospital care and prescription drug coverage. We accrued patients from April 1st, 1998, following them until December 31st, 2012 (for kidney cancer-specific mortality) and December 31st, 2014 (for all-cause mortality). These dates were based on the most recent update of the database used for each outcome. We deemed study subjects to have localized kidney cancer at presentation if they underwent surgery (OHIP billing codes S411, S412, S413, S415, S416), radiofrequency ablation (OHIP billing code J069) or cryotherapy (OHIP billing code S400) as the first treatment following diagnosis of kidney cancer. Other subjects were deemed to have advanced disease if they received immunotherapy (OHIP billing codes G381, G281, G345, G359, G075, G390, G388)
or no intervention as their first treatment following diagnosis of kidney cancer. This classification is consistent with other studies conducted in Ontario [30,31]. We excluded patients who died within 30 days of intervention (surgery, radiofrequency ablation, or cryotherapy), since this is unlikely to reflect the effects of medication exposure. We also excluded patients with any diagnosis of malignancy (excluding non-melanoma skin cancers) prior to their kidney cancer diagnosis in order to exclude patients whose overall survival may be influenced by a pre-existing malignancy. Finally, we excluded patients whose histology indicated a primary malignancy other than kidney cancer. 2.4. Exposure assessment We quantified medication exposure from the diagnosis of kidney cancer to the end of follow-up using prescription dates and the number of days supplied in each, as done previously [32,33]. This allowed for calculation of the duration of cumulative exposure for each day of follow-up, as well as accounting for periods of intermittent use. 2.5. Outcome assessment The primary outcome was all-cause mortality and the secondary outcome was kidney cancer-specific mortality. For each outcome, patients were followed until their date of last contact with health services, death or the end of the study period, whichever occurred first. 2.6. Statistical methods We conducted time-to-event analyses using multivariable Cox proportional hazard regression to estimate the effect of drug exposure on the risk of the primary and secondary outcomes. For the secondary outcome, we focused on the cause-specific hazard of death from kidney cancer, accounting for the competing risk of death from other causes. Covariates in the model were selected a priori and include age at kidney cancer diagnosis, sex, duration of diabetes, comorbidity (defined by the Johns Hopkins Adjusted Clinical Groups score [34]), year of kidney cancer diagnosis (to account for temporal changes in exposure and outcomes), disease stage, rurality, socioeconomic status and exposure to each medication after kidney cancer diagnosis. We evaluated the association between medication exposure and outcomes in two ways. First, we studied medication use as a time-varying covariate denoting whether or not the subject was actively receiving the medication of interest, determined using prescription claims. Second, we examined cumulative medication use as a time-varying covariate (Supplemental Methods 2). Both of these models account for intermittent use; however, they evaluate different hypotheses. In the time-varying analyses, a patient can contribute survival time to both the exposed and unexposed group, based on the date and duration of their prescription, offsetting the selection bias associated with a simple comparison of
470
M. Nayan et al. / Pharmacological Research 113 (2016) 468–474
Fig. 2. Cohort derivation.
users and non-users. As such, the time-varying analyses evaluate the hypothesis of whether current use of a medication, as a binary variable, is associated with survival. In contrast, cumulative use analyses evaluate whether increasing use is associated with survival. This approach method mitigates confounding by indication because it compares patients with varying degrees of total use [35]. For the analysis of cumulative duration of use, we used fractional polynomials [36] to determine the function form that best described the relationship between cumulative exposure to each medication and the hazard of mortality. All statistical analyses were performed using SAS (version 9.3; SAS Institute, Cary, NC) and used a two-sided p value of 0.05 as the threshold for statistical significance. 3. Results 3.1. Cohort characteristics Over the 14-year study period, we identified 815 patients aged 66 or older with incident diabetes and a subsequent diagnosis of kidney cancer. Of these, 613 patients met inclusion criteria (Fig. 2). The characteristics of the cohort are described in Table 1. Of these, 409 (67%) died during follow-up, including 194 (32%) who died from kidney cancer. Medication use following kidney cancer diagnosis is shown in Supplemental Table 1. 3.2. All-cause mortality In the primary analysis, current use of statin therapy was associated with a significant reduction in all-cause mortality (adjusted hazard ratio 0.74; 95% CI 0.59–0.91) (Table 2). Other characteristics independently associated with all-cause mortality included sex, age at kidney cancer diagnosis, disease stage, socioeconomic status and rural residence (Table 2). In contrast, current use of NSAIDs or metformin was not significantly associated with all-cause mortality. For no study medication was cumulative exposure significantly associated with all-cause mortality (Supplemental Table 2). 3.3. Kidney cancer-specific mortality In the primary analysis, current use of statins was associated with a significant reduction in kidney cancer-specific mortality (adjusted hazard ratio 0.71; 95% CI 0.51–0.97) (Table 3). No
such association was observed with any other study medication. Other characteristics independently associated with kidney cancerspecific mortality included duration of diabetes, disease stage and rural residence (Table 3). Cumulative duration of medication use was not associated with kidney-cancer mortality for any of the study medications (Supplemental Table 2). 4. Discussion This population-based study of patients with type 2 diabetes found that in a time-varying analysis, current use of a statin following a diagnosis of kidney cancer was associated with a 26% reduced risk of all-cause mortality and a 29% reduced risk of kidney cancerspecific mortality. However, an association between increasing use of statins and survival outcomes was not observed. In no analysis was exposure to receipt of NSAIDs or metformin associated with improved survival. One possible explanation for our results demonstrating a protective association with current use of statin but not increasing cumulating use is that statins have a short half-life (0.5–3 h), and the putative anti-neoplastic effects may be limited to the duration that statins are present in the circulation. This may be similar to the effects observed upon the discontinuation of tyrosine kinase inhibitors, such as sunitinib (half-life 40–60 h) or pazopanib (halflife 31 h), which results in accelerated tumour growth and tumour flare in kidney cancer patients [37]. In this manner, the outcomes may be more related to the current use of a statin (time-varying), rather than the anti-neoplastic effects of statins cumulating with each additional use (cumulative use). Importantly, an alternate explanation for our finding of no association between increasing statin use and survival outcomes is that cumulative use analyses assign equal weight to remote exposures and more recent exposures, which may not be appropriate. While our study is the first to evaluate the association between statin exposure and survival outcomes in diabetic patients with kidney cancer, previous studies evaluating statins in non-diabetics with kidney cancer yield conflicting results [16,21–24]. The majority of these defined statin users based on medication use at the time of surgery. Classifying exposure in this manner does not account for intermittent use and can therefore lead to bias because medication use can change over time. Indeed, one study found that the association between statin use and survival outcomes in kidney cancer differed when comparing use at the time of surgery with the
M. Nayan et al. / Pharmacological Research 113 (2016) 468–474
471
Table 1 Baseline Patient Characteristics (n = 613). Characteristics
All patients (n = 613)
Male gender (No (%)) Age at diabetes diagnosis, years (mean (SD)) Age at KC diagnosis, years (mean (SD)) Duration of diabetes to KC diagnosis, years (mean (SD)) Time from KC diagnosis to end of follow-up, years (mean (SD)) Primary Treatment (No (%)) Surgery or Radiofrequency ablation No primary treatment or Immunotherapy a Comorbidity score (No (%)) Low Intermediate High b SES status (No (%)) Income Quintile 1 (lowest) Income Quintile 2 Income Quintile 3 Income Quintile 4 Income Quintile 5 (highest) Rural residence (No (%)) KC-specific death (No (%)) Overall mortality (No (%))
367 (59.9) 73.8 (6.0) 77.9 (6.4) 4.1 (3.3) 3.3 (3.4) 272 (44.4) 341 (55.6) 22 (3.6) 90 (14.7) 501 (81.7) 132 (21.5) 124 (20.2) 130 (21.2) 135 (22.0) 92 (15.0) 76 (12.4) 194 (31.6) 409 (66.7)
Abbreviations: KC, kidney cancer; SES, socioeconomic status. a Comorbidity scores were calculated by using Johns Hopkins Adjusted Clinical Groups Case-Mix System assigning a specific weight to each adjusted diagnostic group (low, weighted adjusted diagnostic group score 5 or lower; intermediate, 6–9; high, 10 or higher). b Income quintiles from median income in neighborhoods from 1 (low) to 5 (high). Table 2 Cox proportional hazard models evaluating all-cause mortality (time-varying analysis denoting effect of current exposure). Characteristics
Univariate
Multivariable
HR (95% CI)
HR (95% CI)
1.03 (0.85–1.26) 1.06 (1.05–1.08) 1.03 (0.99–1.06) 4.79 (3.83–5.99)
1.25 (1.02–1.54) 1.03 (1.01–1.05) 0.96 (0.93–1.00) 4.42 (3.49–5.60)
1.00 (reference) 1.05 (0.60–1.84) 0.81 (0.48–1.37)
1.00 (reference) 1.26 (0.71–2.25) 1.05 (0.62–1.79)
Socioeconomic status(lowest) Income quintile 1 Income quintile 2 Income quintile 3 Income quintile 4 Income quintile 5 (highest) Rural residence Year of KC diagnosis
1.00 (reference) 0.78 (0.58–1.04) 0.78 (0.59–1.04) 0.66 (0.49–0.88) 0.67 (0.49–0.93) 1.59 (1.21–2.08) 1.02 (0.98–1.04)
1.00 (reference) 0.85 (0.64–1.15) 0.89 (0.66–1.19) 0.68 (0.50–0.92) 0.65 (0.47–0.90) 1.56 (1.18–2.07) 1.00 (0.97–1.04)
Medication exposure after KC diagnosis Statins NSAIDs Metformin Insulin Insulin secretagogue Thiazolidinedione
0.74 (0.61–0.91) 1.06 (0.81–1.38) 0.99 (0.78–1.25) 1.33 (0.71–2.50) 1.33 (1.02–1.72) 1.44 (0.46–4.47)
0.74 (0.59–0.91) 1.04 (0.79–1.36) 1.09 (0.85–1.40) 1.87 (0.98–3.59) 1.16 (0.88–1.53) 2.88 (0.90–9.28)
Male gender Age at KC diagnosis (years) Duration of diabetes to KC diagnosis (years) Disease stage (advanced vs. localized) a Comorbidity score Low Intermediate High b
Abbreviations: IQR, interquartile range; KC, kidney cancer; SES, socioeconomic status. a Comorbidity scores were calculated by using Johns Hopkins Adjusted Clinical Groups Case-Mix System assigning a specific weight to each adjusted diagnostic group (low, weighted adjusted diagnostic group score 5 or lower; intermediate, 6–9; high, 10 or higher). b Income quintiles from median income in neighborhoods from 1 (low) to 5 (high).
time-varying analysis [16]. Furthermore, the results of these studies may not be generalized to non-diabetics; a recent study found that statin use among diabetic patients with pancreatic cancer was associated with significantly improved overall survival while there was no benefit of statin use in non-diabetics [38]. Although we found that current use of statins was associated with improved survival outcomes, use of metformin, another putative anti-neoplastic medication, was not. Three previous studies in kidney cancer have demonstrated similar results; [19,25,39] however, these studies classified medication exposure based on use
at the time of surgery and did not account for intermittent use. While cumulative metformin use has been associated with survival outcomes in prostate cancer [32], it may be possible that the anti-neoplastic effect of metformin on survival outcomes may be dependent on the type of cancer. Indeed, another study found no association between cumulative metformin exposure and survival outcomes in patients with breast cancer [33]. Similarly, it is possible that NSAIDs are beneficial in some malignancies and not in others, as shown previously [40]. To date, the association between use of NSAIDs and kidney cancer survival out-
472
M. Nayan et al. / Pharmacological Research 113 (2016) 468–474
Table 3 Cox proportional hazard models evaluating kidney-cancer specific mortality (time-varying analysis denoting current exposure). Characteristics
Male gender Age at KC diagnosis (years) Duration of diabetes–KC diagnosis (years) Disease stage (advanced vs. localized) a Comorbidity score Low Intermediate High b Socioeconomic status Income quintile 1 (lowest) Income quintile 2 Income quintile 3 Income quintile 4 Income quintile 5 (highest) Rural residence Year of KC diagnosis Medication exposure after KC diagnosis Statins NSAIDs Metformin Insulin Insulin secretagogue Thiazolidinedione
Univariate
Multivariable
HR (95% CI)
HR (95% CI)
0.93 (0.70–1.24) 1.04 (1.02–1.06) 1.00 (0.96–1.05) 5.67 (4.00–8.04)
1.08 (0.81–1.46) 1.01 (0.98–1.03) 0.93 (0.89–0.98) 5.64 (3.92–8.11)
1.00 (reference) 1.30 (0.61–2.78) 0.75 (0.37–1.54)
1.00 (reference) 1.74 (0.80–3.82) 1.09 (0.52–2.28)
1.00 (reference) 0.73 (0.48–1.11) 0.72 (0.47–1.09) 0.67 (0.44–1.09) 0.69 (0.44–1.09) 1.90 (1.32–2.72) 1.04 (0.99–1.08)
1.00 (reference) 0.76 (0.49–1.17) 0.79 (0.51–1.21) 0.69 (0.44–1.06) 0.64 (0.40–1.02) 1.80 (1.23–2.62) 1.05 (0.99–1.10)
0.68 (0.50–0.91) 0.96 (0.65–1.42) 1.12 (0.80–1.55) 0.95 (0.30–2.96) 1.12 (0.75–1.66) 1.84 (0.46–7.40)
0.71 (0.51–0.97) 0.98 (0.66–1.46) 1.25 (0.88–1.77) 1.14 (0.36–3.65) 1.05 (0.69–1.58) 2.48 (0.58–10.65)
Abbreviations: KC, kidney cancer; SES, socioeconomic status. a Comorbidity scores were calculated by using Johns Hopkins Adjusted Clinical Groups Case-Mix System assigning a specific weight to each adjusted diagnostic group (low, weighted adjusted diagnostic group score 5 or lower; intermediate, 6–9; high, 10 or higher). b Income quintiles from median income in neighborhoods from 1 (low) to 5 (high).
comes has not been studied; however, our results do not support their use for this purpose. Our study has several strengths. First, we used a population based-cohort and our study is the largest to date to evaluate putative anti-neoplastic medications among diabetics with kidney cancer. Second, the comprehensive nature of prescription claims data permitted more robust analyses evaluating medication exposure, rather than simply classifying exposure based on use at the time of diagnosis. Third, we simultaneously evaluated multiple putative anti-neoplastic medications which allowed us to evaluate the independent effect of each medication, while controlling for the possible effects of the others. Furthermore, our results demonstrating that active use of neither metformin nor NSAIDs were associated with survival outcomes suggests against a healthy user bias behind the protective association observed for statins. Fourth, by using a cohort of incident diabetics, we were able to evaluate a more homogenous population and minimize selection bias that may be attributable to varying durations of diabetes. Finally, while randomized trials provide the strongest level of evidence, the profile of patients included in trials differ from those in real-world settings. The results of this study, on the other hand, provide insight on the real-world association of the use of putative anti-neoplastic medications on survival outcomes among diabetics with kidney cancer. Some limitations of our study merit emphasis. First, several factors govern the prescribing of medications we studied, and certain details (such as tumour characteristics) were not available. However, these are not true confounders as they are not associated with both exposure and outcome. As such, controlling for these factors will not improve the estimation of the exposure-outcome relationship; rather, it would lead to overadjustment [41]. Additionally, many studies have shown that tumour stage was not significantly different in users vs. non-users of statins or metformin [16,19,21,23,24,39]. While there is conflicting evidence on the association of obesity and kidney cancer survival outcomes [42–45], data on obesity was not available in the databases used. Similarly,
data on glycemic control was not available. Second, we classified disease stage based on the intervention following cancer diagnosis and there may be patients with localized disease that were managed with observation, and patients with advanced disease that underwent a cytoreductive nephrectomy. However, the hazard ratios associated with our method of classifying disease stage suggest against a strong effect of misclassficiation. Third, because we used prescription data as a surrogate for medication use, the actual and estimated exposures may not accord if patients are non-compliant. However, the analyses in this study accounted for intermitted exposure and therefore controlled for lapses between prescription claims that would be expected in non-compliant patients. Fourth, while administrative data are susceptible to misclassification bias, several of the databases used in this study have been validated (Supplemental Methods 1). Fifth, some NSAIDs are available over the counter and it is possible that some exposure was not captured due to this. However, patients in our study have a financial incentive to obtain these drugs by prescription, particularly for long-term use, and previous research shows that administrative data represent a valid indicator of NSAIDs exposure [46]. Finally, the results of this study are limited to incident diabetic patients aged 66 or older with a subsequent diagnosis of kidney cancer. While the generalizability of our findings is unknown, the results provide impetus to study statins in younger diabetic patients with kidney cancer. Despite these limitations, this is the first study to simultaneously evaluate statins, metformin, and NSAIDs in a large population of type 2 diabetic patients with kidney cancer and suggests that statins may have a role in kidney cancer therapy among diabetics.
5. Conclusion Among type 2 diabetic patients, current statin use following a diagnosis of kidney cancer was associated with a significant improvement in all-cause mortality and kidney cancer-specific mortality. However, the benefits of statins were not dependent
M. Nayan et al. / Pharmacological Research 113 (2016) 468–474
on increasing total cumulative use. These findings provide support for randomized trials evaluating the potential effect of statins on survival in diabetic patients with kidney cancer. Funding This study was supported in part by a grant from the Canadian Drug Safety & Effectiveness Research Network (CDSERN), and by the Institute for Clinical Evaluative Sciences (ICES), a non-profit research institute sponsored by the Ontario Ministry of Health and Long-Term Care (MOHLTC). Dr. Juurlink is supported by the Eaton Scholar award, Department of Medicine, University of Toronto. The opinions, results and conclusions reported in this paper are those of the authors and are independent from the funding sources. No endorsement by ICES or the Ontario MOHLTC is intended or should be inferred. Competing interests’ None. Acknowledgements We thank Brogan Inc., Ottawa for use of their Drug Product and Therapeutic Class Database. Parts of this material are based on data and information compiled and provided by CIHI. However, the analyses, conclusions, opinions and statements expressed herein are those of the author, and not necessarily those of CIHI. The authors thank Qing Li, analyst at the Institute of Clinical and Evaluative Sciences, for preparing the data. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.phrs.2016.09. 027. References [1] R.L. Siegel, K.D. Miller, A. Jemal, Cancer statistics, CA: Cancer J. Clin. 65 (1) (2015) 5–29. [2] A. Znaor, J. Lortet-Tieulent, M. Laversanne, A. Jemal, F. Bray, International variations and trends in renal cell carcinoma incidence and mortality, Eur. Urol. (2014). [3] W.H. Chow, L.M. Dong, S.S. Devesa, Epidemiology and risk factors for kidney cancer, Nat. Rev. Urol. 7 (5) (2010) 245–257. [4] P. De, M.C. Otterstatter, R. Semenciw, L.F. Ellison, L.D. Marrett, D. Dryer, Trends in incidence, mortality, and survival for kidney cancer in Canada, Cancer Causes Control (2014) 1986–2007. [5] A. Vavallo, S. Simone, G. Lucarelli, M. Rutigliano, V. Galleggiante, G. Grandaliano, L. Gesualdo, M. Campagna, M. Cariello, E. Ranieri, G. Pertosa, G. Lastilla, F.P. Selvaggi, P. Ditonno, M. Battaglia, Pre-existing type 2 diabetes mellitus is an independent risk factor for mortality and progression in patients with renal cell carcinoma, Medicine (United States) 93 (27) (2014). [6] Y.S. Ha, W.T. Kim, S.J. Yun, S.C. Lee, W.J. Kim, Y.H. Park, S.H. Kang, S.H. Hong, S.S. Byun, Y.J. Kim, Multi-institutional analysis of localized renal cell carcinoma that demonstrates the impact of diabetic status on prognosis after nephrectomy, Ann. Surg. Oncol. 20 (11) (2013) 3662–3668. [7] G. Danaei, M.M. Finucane, Y. Lu, G.M. Singh, M.J. Cowan, C.J. Paciorek, J.K. Lin, F. Farzadfar, Y.-H. Khang, G.A. Stevens, National, regional, and global trends in fasting plasma glucose and diabetes prevalence since 1980: systematic analysis of health examination surveys and epidemiological studies with 370 country-years and 2· 7 million participants, Lancet 378 (9785) (2011) 31–40. [8] S.C. Larsson, A. Wolk, Diabetes mellitus and incidence of kidney cancer: a meta-analysis of cohort studies, Diabetologia 54 (5) (2011) 1013–1018. [9] C. Bao, X. Yang, W. Xu, H. Luo, Z. Xu, C. Su, X. Qi, Diabetes mellitus and incidence and mortality of kidney cancer: a meta-analysis, J. Diabetes Complications 27 (4) (2013) 357–364. [10] S.F. Nielsen, B.G. Nordestgaard, S.E. Bojesen, Statin use and reduced cancer-related mortality, New Engl. J. Med. 367 (19) (2012) 1792–1802. [11] G.W. Landman, N. Kleefstra, K.J. van Hateren, K.H. Groenier, R.O. Gans, H.J. Bilo, Metformin associated with lower cancer mortality in type 2 diabetes ZODIAC-16, Diabetes Care 33 (2) (2010) 322–326.
473
[12] M. Nayan, B. Bhindi, L.Y. Julie, T. Hermanns, A. Mohammed, R.J. Hamilton, A. Finelli, M.A. Jewett, A.R. Zlotta, N.E. Fleshner, The effect of metformin on cancer-specific survival outcomes in diabetic patients undergoing radical cystectomy for urothelial carcinoma of the bladder, in: Urologic Oncology: Seminars and Original Investigations, Elsevier, 2015, p. 386. e387-386. e313. [13] X. Liao, P. Lochhead, R. Nishihara, T. Morikawa, A. Kuchiba, M. Yamauchi, Y. Imamura, Z.R. Qian, Y. Baba, K. Shima, Aspirin use, tumor PIK3CA mutation, and colorectal-cancer survival, New Engl. J. Med. 367 (17) (2012) 1596–1606. [14] B.I. Rini, V. Weinberg, S. Dunlap, A. Elchinoff, N. Yu, R. Bok, J. Simko, E.J. Small, Maximal COX-2 immunostaining and clinical response to celecoxib and interferon alpha therapy in metastatic renal cell carcinoma, Cancer 106 (3) (2006) 566–575. [15] N. Shinohara, A. Kumagai, K. Kanagawa, S. Maruyama, T. Abe, A. Sazawa, K. Nonomura, Multicenter phase II trial of combination therapy with meloxicam, a COX-2 inhibitor, and natural interferon-␣ for metastatic renal cell carcinoma, Jpn. J. Clin. Oncol. 39 (11) (2009) 720–726. [16] R.J. Hamilton, D. Morilla, F. Cabrera, M. Leapman, L.Y. Chen, M. Bernstein, A.A. Hakimi, V.E. Reuter, P. Russo, The association between statin medication and progression after surgery for localized renal cell carcinoma, J. Urol. 191 (4) (2014) 914–919. [17] F. Khimani, N. Shah, B.F. Curley, M.D. Newton, T.F. Hogan, M. Monga, Metformin use and renal cell cancer outcomes among patients with diabetes, J. Clin. Oncol. (2014). [18] N. Shah, F. Khimani, B.F. Curley, M.D. Newton, T.F. Hogan, S. Wen, Y. Hu, M. Monga, Metformin use and survival in diabetic patients with renal cell carcinoma (RCC), J. Clin. Oncol. (2014). [19] S. Psutka, S. Boorjian, C. Lohse, S. Stewart, M. Tollefson, B. Leibovich, R.H. Thompson, The effect of metformin in patients with surgically treated renal cell carcinoma, J. Urol. (2014) e579. [20] S.M. Haake, S.A. Brooks, E. Welsh, A. Berglund, W.J. Fulp, D.T. Chen, E.B. Haura, W.J. Sexton, P.E. Spiess, J. Pow-Sang, K. Rathmell, M.N. Fishman, Molecular subtypes of clear cell renal cell carcinoma: impact of diabetes mellitus, metformin, and immunotherapy on patient outcomes, J. Clin. Oncol. (2014). [21] S.D. Kaffenberger, O. Lin-Tsai, K.L. Stratton, T.M. Morgan, D.A. Barocas, S.S. Chang, M.S. Cookson, S.D. Herrell, J.A. Smith, P.E. Clark, Statin use is associated with improved survival in patients undergoing surgery for renal cell carcinoma, in: Urologic Oncology: Seminars and Original Investigations, Elsevier, 2015, p. 21. e11-21. e17. [22] B.R. Viers, R.H. Thompson, S.P. Psutka, C.M. Lohse, J.C. Cheville, B.C. Leibovich, M.K. Tollefson, S.A. Boorjian, The association of statin therapy with clinicopathologic outcomes and survival among patients with localized renal cell carcinoma undergoing nephrectomy, in: Urologic Oncology: Seminars and Original Investigations, Elsevier, 2015. [23] L.S. Krane, J.M. Sandberg, J.T. Rague, A.K. Hemal, Do statin medications impact renal functional or oncologic outcomes for robot-assisted partial nephrectomy? J. Endourol. 28 (11) (2014) 1308–1312. [24] M. Nayan, A. Finelli, M. Jewett, D. Juurlink, P. Austin, G. Kulkarni, R. Hamilton, Statin use and kidney cancer survival outcomes: a propensity score analysis, Urol. Oncol. (2016) http://www.ncbi.nlm.nih.gov/pubmed/27427224. [25] M. Nayan, A. Finelli, M. Jewett, D. Juurlink, P. Austin, G. Kulkarni, R. Hamilton, Metformin use and kidney cancer outcomes in diabetics: a propensity score analysis, Clin. Genitourin. Cancer (2016) http://www.ncbi.nlm.nih.gov/ pubmed/27424258. [26] J.E. Hux, F. Ivis, V. Flintoft, A. Bica, Diabetes in Ontario: determination of prevalence and incidence using a validated administrative data algorithm, Diabetes Care 25 (3) (2002) 512–516. [27] D.R. Brenner, M.C. Tammemagi, S.B. Bull, D. Pinnaduwaje, I.L. Andrulis, Using cancer registry data: agreement in cause-of-death data between the Ontario Cancer Registry and a longitudinal study of breast cancer patients, Chronic Dis. Can. 30 (1) (2009) 16–19. [28] A. Levy, B. O’Brien, C. Sellors, P. Grootendorst, D. Willison, Coding accuracy of administrative drug claims in the Ontario Drug Benefit database, Can. J. Clin. Pharmacol. 10 (2) (2003) 67. [29] D.N. Juurlink, Canadian Institute for Health Information Discharge Abstract Database: a Validation Study, Institute for Clinical Evaluative Sciences, 2006. [30] J.L. Warren, L. Barbera, K.E. Bremner, K.R. Yabroff, J.S. Hoch, M.J. Barrett, J. Luo, M.D. Krahn, End-of-life care for lung cancer patients in the United States and Ontario, J. Natl. Cancer Inst. 103 (11) (2011) 853–862. [31] R.W. Jang, M.K. Krzyzanowska, C. Zimmermann, N. Taback, S.M. Alibhai, Palliative care and the aggressiveness of end-of-life care in patients with advanced pancreatic cancer, J. Natl. Cancer Inst. 107 (3) (2015) dju424. [32] D. Margel, D.R. Urbach, L.L. Lipscombe, C.M. Bell, G. Kulkarni, P.C. Austin, N. Fleshner, Metformin use and all-cause and prostate cancer-specific mortality among men with diabetes, J. Clin. Oncol. 31 (25) (2013) 3069–3075. [33] I.C. Lega, P.C. Austin, A. Gruneir, P.J. Goodwin, P.A. Rochon, L.L. Lipscombe, Association between metformin therapy and mortality after breast cancer: a population-based study, Diabetes Care 36 (10) (2013) 3018–3026. [34] P.C. Austin, C. van Walraven, W.P. Wodchis, A. Newman, G.M. Anderson, Using the Johns Hopkins Aggregated Diagnosis Groups (ADGs) to predict mortality in a general adult population cohort in Ontario, Canada, Med. Care 49 (10) (2011) 932–939. [35] D. Margel, D. Urbach, N. Fleshner, P.C. Austin, Reply to MMJ Zanders et al, J. Clin. Oncol. 32 (7) (2014) 702. [36] W. Sauerbrei, C. Meier-Hirmer, A. Benner, P. Royston, Multivariable regression model building by using fractional polynomials: description of SAS, STATA and R programs, Comput. Statistics & Data Anal. 50 (12) (2006) 3464–3485.
474
M. Nayan et al. / Pharmacological Research 113 (2016) 468–474
[37] R. Iacovelli, F. Massari, L. Albiges, Y. Loriot, C. Massard, K. Fizazi, B. Escudier, Evidence and clinical relevance of tumor flare in patients who discontinue tyrosine kinase inhibitors for treatment of metastatic renal cell carcinoma, Eur. Urol. 68 (July (1)) (2014) 154–160. [38] Y. Nakai, H. Isayama, T. Sasaki, S. Mizuno, N. Sasahira, H. Kogure, K. Kawakubo, N. Yamamoto, K. Hirano, H. Ijichi, Clinical outcomes of chemotherapy for diabetic and nondiabetic patients with pancreatic cancer: better prognosis with statin use in diabetic patients, Pancreas 42 (2) (2013) 202–208. [39] A.A. Hakimi, L. Chen, P.H. Kim, D. Sjoberg, L. Glickman, M.R. Walker, P. Russo, The impact of metformin use on recurrence and cancer-specific survival in clinically localized high-risk renal cell carcinoma, Can. Urol. Assoc. J. 7 (11-12) (2013) E687–691. [40] P.M. Rothwell, F.G.R. Fowkes, J.F. Belch, H. Ogawa, C.P. Warlow, T.W. Meade, Effect of daily aspirin on long-term risk of death due to cancer: analysis of individual patient data from randomised trials, Lancet 377 (9759) (2011) 31–41. [41] K. Jager, C. Zoccali, A. Macleod, F. Dekker, Confounding: what it is and how to deal with it, Kidney Int. 73 (3) (2008) 256–260.
[42] S.M. Donat, E.W. Salzhauer, N. Mitra, B.V. Yanke, M.E. Snyder, P. Russo, Impact of body mass index on survival of patients with surgically treated renal cell carcinoma, J. Urol. 175 (1) (2006) 46–52. [43] A. Haferkamp, M. Pritsch, J. Bedke, N. Wagener, J. Pfitzenmaier, S. Buse, M. Hohenfellner, The influence of body mass index on the long-term survival of patients with renal cell carcinoma after tumour nephrectomy, BJU Int. 101 (10) (2008) 1243–1246. [44] A.M. Kamat, R.P. Shock, Y. Naya, C.J. Rosser, J.W. Slaton, L.L. Pisters, Prognostic value of body mass index in patients undergoing nephrectomy for localized renal tumors, Urology 63 (1) (2004) 46–50. [45] S. Steffens, C. Von Klot, H. Eggers, C. Seidel, G. Wegener, M. Schrader, M.A. Kuczyk, A.J. Schrader, The role of diabetes mellitus and obesity in localized and metastatic RCC, Eur. Urol. 11 (Suppl. 1) (2012), e1018-e1018a. [46] M.U. Yood, U.B. Campbell, K.J. Rothman, S.S. Jick, J. Lang, K.E. Wells, H. Jick, C.C. Johnson, Using prescription claims data for drugs available over-the-counter (OTC), Pharmacoepidemiol. Drug Saf. 16 (9) (2007) 961–968.