Journal of Clinical Pharmacy and Therapeutics, 2014, 39, 354–360

doi: 10.1111/jcpt.12151

Dose effect of thiazolidinedione on cancer risk in type 2 diabetes mellitus patients: a six-year population-based cohort study H. C. Lin*† MD, Y. T. Hsu‡ MS, B. H. Kachingwe‡ MS, C. Y. Hsu§ PhD, Y. S. Uang‡ PhD and L. H. Wang‡¶ PhD *Department of Pediatrics, School of Medicine, College of Medicine, Taipei Medical University, Taipei, †Department of Laboratory Medicine, Taipei Medical University Hospital, Taipei, ‡School of Pharmacy, College of Pharmacy, Taipei Medical University, Taipei, §Graduate Institute of Biomedical Informatics, College of Medical Science and Technology, Taipei Medical University, Taipei, and ¶Department of Pharmacy, Taipei Medical University Hospital, Taipei,Taiwan

Received 12 February 2014, Accepted 18 February 2014

Keywords: cancer risk, cohort study, dose effect, thiazolidinedione, type 2 diabetes mellitus

mediated by insulin include stimulation of aromatase activity, thereby increasing bioavailable oestrogen2 and inhibition of the production of sex-hormone-binding globulin, which would result in higher levels of free oestrogens and androgens.3 In addition, insulin increases the amount of bioavailable insulin-like growth factor-1 by inhibiting the production of insulin-like growth factor binding protein.4 Thiazolidinediones (TZDs) are selective ligands of the nuclear transcription factor peroxisome proliferator-activated receptor gamma (PPARc) that are used to treat DM.5 TZDs induce anticancer actions through three main PPARc-dependent pathways: (i) cell-cycle arrest, (ii) induction of apoptosis and (iii) inhibition of cell invasion. In addition to PPARc-dependent actions, PPARc-independent effects induced by TZDs may also inhibit tumorigenesis.6 Antitumorigenic effects of TZDs were demonstrated in breast,7 colon,8 lung,9,10 thyroid11 and hepatocellular cancer cells.12 However, rodent studies indicate that PPARc agonists can potentiate tumorigenesis; mice treated with high doses of TZDs showed increase in the number of colon polyps.13,14 There is also concern that pioglitazone use increases the risk of bladder cancer.15,16 Therefore, the biological plausibility of the role of TZDs in human cancer is uncertain. Results of epidemiologic studies of the risk of cancer among patients with DM are conflicting.13,17,18 A recent meta-analysis of randomized clinical trials of rosiglitazone also reported no statistically significant effect on the cancer risk.19 Due to insufficient clinical data regarding the TZDs effects on cancer risk, especially among Asian populations, we conducted a hypothesis-generating, retrospective study in Taiwan to assess the risk of development of all kinds of cancer among DM patients treated with TZDs relative to those treated with other anti-DM drugs and those not prescribed any anti-DM drugs.

SUMMARY What is known and objective: Prior studies found that thiazolidinediones (TZDs) might have tumour-suppressor activity mediated through cell-cycle arrest, induction of apoptosis and inhibition of cell invasion. The main objective of this study was to investigate the effects of TZDs on the risk of cancer among patients with type 2 diabetes mellitus (DM). Methods: Patients diagnosed with DM between 1 January 1998 and 31 December 2002 were identified from the Longitudinal Health Insurance Database (LHID) within the Taiwan National Health Insurance (NHI) programme. Using Cox regression models, we assessed the association between prescribed TZDs and cancer risk, TZDs’ dose effect and the association between TZDs and specific cancer types. Hazard ratios (HR) were adjusted for potential confounders (age, gender, income, Charlson score index, metformin and insulin use). Results and discussion: The adjusted HRs for those prescribed TZD were 0
74 (95% CI 0
43–1
26, P = 0
27), 0
39 (95% CI 0
33– 0
45, P < 0
001) and 0
49 (95% CI 0
27–0
89, P = 0
02), respectively, relative to non-DM patients, DM patients prescribed other anti-DM drugs besides TZDs and DM patients not prescribed any anti-DM drugs. In addition, the effects of TZDs were shown to be significantly dose dependent (P for trend < 0
001). The risk of breast, brain, colorectal, ear–nose– throat, kidney, liver, lung, lymphatic, prostate, stomach, and uterus cancer was significantly lower in those prescribed TZDs. What is new and conclusions: The results showed a decrease in cancer risk in diabetic patients using TZD, and the association was dose dependent.

WHAT IS KNOWN AND OBJECTIVE

METHODS

Cancer and type 2 diabetes mellitus (DM), two common diseases that share many risk factors, greatly impact health worldwide. Epidemiological evidence suggests that DM confers significant risks of many forms of cancer.1 Type 2 DM is caused by peripheral insulin resistance and is usually characterized by b cell hyperplasia and hyperinsulinemia. Hyperinsulinemia may lead to cancer through its direct effects on cancer cells through the insulin receptor or insulin-like growth factor I receptor. Indirect pathways

Subjects We conducted a population-based retrospective cohort study among residents of Taiwan. The study included 32 891 patients identified from the Longitudinal Health Insurance Database (LHID), with at least three consecutive episodes of diagnosed type 2 DM (International Classification of Disease version 9 (ICD9) codes 250), between 1 January 1998 and 31 December 2002; three episodes were required to increase the validity of diagnosis. Only patients who had not been diagnosed with DM or cancer before 1 January 1998 were included to ensure that only newly identified cases were included. This study was exempted from full review by

Correspondence: Li-Hsuan Wang, School of Pharmacy, College of Pharmacy, Taipei Medical University, No. 250, Wu-Hsing St., Taipei 11031, Taiwan. Tel.: 886 2 2737 2181 ext 8434; fax: 886 2 2378 1926; e-mail: [email protected]

© 2014 John Wiley & Sons Ltd

354

H. C. Lin et al.

Dose effect of thiazolidinedione on cancer risk

the Institutional Review Board because the identification numbers of all individuals in the National Health Insurance (NHI) database were encrypted to protect the privacy of the individuals, and the data set was released by NHI for public research purpose. We classified the type 2 DM patients into the following groups: Group 1: patients who were prescribed pioglitazone or rosiglitazone; Group 2: patients who were prescribed other anti-DM drugs besides TZDs; these drugs included metformin, sulfonylureas, acarbose, insulin analogues and others and Group 3: patients who were not prescribed any anti-DM drugs. The non-DM control group: for each study, group patient (groups 1–3), four non-DM and non-cancer control patients were randomly selected from the remaining population, resulting in 131 564 matched controls.

RESULTS AND DISCUSSION From the LHID data set of the NHI, we identified 32 891 type 2 DM patients meeting the study inclusion criteria; they were classified into three groups according to prescribed anti-DM therapy (Fig. 1). There were significant differences in baseline characteristics (age, sex income and Charlson score index) and metformin and insulin use between the study and control group (Table 1). Rarely, the non-diabetic patients used insulin or metformin, possibly for other indications, such as septic shock or polycystic ovary syndrome. We compared the cancer risk in the different groups, non-DM patients, patients prescribed other anti-DM drugs besides TZDs and patients not prescribed any anti-DM drugs with those prescribed TZDs. Although the hazard ratio was 1
44 (95% CI: 1
24–1
68) for the TZDs prescribed group compared with non-DM patients group, the risk of cancer did not differ significantly after adjusting for potential confounders (adjusted HR = 0
74, 95% CI: 0
43–1
26). However, significantly lower adjusted HRs were found when the TZDs prescribed group was compared with patients prescribed other anti-DM drugs (besides TZDs) and those not prescribed any anti-DM drugs (Table 2). In the analysis of dose effect of TZDs on cancer risk, patients prescribed TZDs had lower cancer incidence and longer time until cancer occurrence compared with patients prescribed other medications or not prescribed any anti-DM drugs. We evaluated the effects of accumulated doses on the cancer risk and found that the hazard ratios decreased as the prescribed DDDs increased, so there appears to be a significant dose effect (Table 3). The dose effects of rosiglitazone and pioglitazone on the cancer risk were analysed separately, and similar dose effects were observed (data not shown). We also analysed the association between prescribed TZDs and cancer types. The results revealed that for breast; brain; colorectal; ear, nose and throat; renal, hepatic, lung, prostatic, stomach and uterine cancers, the hazard ratios were significantly lower in those prescribed TZDs compared with patients not prescribed any antiDM drugs and patients prescribed other anti-DM drugs (Table 4). The occurrence of cancer during the follow-up period is shown in Fig. 2; the results indicate significantly different rates of cancer occurrence between those prescribed TZDs and patients not prescribed any or prescribed other anti-DM drugs. Although type 2 DM and cancer share many risk factors and diabetes has been consistently associated with an increased risk of several more common site-specific cancers, for many less common cancers, there are limited data to support this association.13,22 Our results indicate that when TZDs is prescribed to DM patients, the difference in cancer risk between DM and non-DM patients become non-significant (adjusted HR: 0
74, 95% CI: 0
43–1
26, P = 0
27). Meanwhile, we have demonstrated a dose-related effect of TZDs on cancer risk, and we further evaluated the effect of TZDs on the risk of many kinds of cancer. Many in vitro studies have demonstrated that TZDs inhibit the growth of human cancer cell lines. However, the results of clinical studies are inconsistent. Home et al.23 evaluated the malignancy data within the A Diabetes Outcome Progression Trial (ADOPT) and Rosiglitazone Evaluated for Cardiovascular Outcomes and Regulation of Glycaemia in Diabetes (RECORD) randomized controlled trials and found a non-significant reduction in malignancies in patients with type 2 DM treated with rosiglitazone and metformin compared with those treated with sulfonylurea. A meta-analysis, performed by Colmers et al.24 that included twenty

Data sources Taiwan NHI is a universal, state-operated health programme that covers >99% of the national population. We analysed the computerized prescription records of LHID, and it consisted of the sampled 5% population of the NHI programme. The LHID included all claim data for these 1 073 891 subjects and contained diagnosis codes, drug prescriptions, hospital visits, including detailed clinical and demographic information of all hospital admissions and ambulatory visits. Dosages of TZD Complete information about all prescriptions of TZDs was extracted from the NRI prescription database. Data collected included the date of prescription, the daily dose and the number of days supplied. For the TZD treatment group, we calculated the total dosage prescribed during the follow-up period. The defined daily doses (DDD) were calculated from prescription data according to WHO’s definition.20 Outcome measures Each patient and each control were followed up for 6 years except the cancer outcome occurred. The primary outcome was the effect of cancer risk of TZD in type 2 diabetic patients, and the secondary outcome was dose effects of TZDs and the association of TZDs with cancer types. Cancer was diagnosed according to ICD-9 codes 140–239. Statistical analysis Statistical analyses were performed using SAS version 9
1 (SAS Institute, Cary, NC, USA). Pearson’s chi-squared tests were applied to evaluate differences in socio-demographic characteristics (age, gender and income) and potential confounders (Charlson score index, metformin and insulin use) between diabetic patients and non-diabetic patients. The comorbidity score, the Charlson score index, predicts 10-year survival and was validated in several previous studies.21 Cox’s proportional hazards regression models were fit to estimate the hazard of cancer for the study and comparison groups after adjusting for patient’s age, gender, income, Charlson score index, metformin and insulin use. The dose effect of the TZD was analysed by trend of the hazard ratios. Cancer occurrence rate during the followed period was calculated, and differences in the cancer-free survival rates between groups were examined by the log-rank test. A two-tailed P < 0
05 was considered statistically significant for all analyses.

© 2014 John Wiley & Sons Ltd

Journal of Clinical Pharmacy and Therapeutics, 2014, 39, 354–360 355

H. C. Lin et al.

Dose effect of thiazolidinedione on cancer risk

Longitudinal Health Insurance Database (LHID) 2005 N = 1,000,000 No cancer diagnosis as of January 1, 1997

Type 2 diabetes (3 consensus diagnoses between Jan 1997 – Dec. 2002)

Type 2 Diabetic Patients, n = 32,891

Matched for age and gender

No previous cancer diagnosis Type 2 Diabetes Drug Therapy, n = 28,225

Non-DM Patients N = 131,564

Other Anti-DM Drugs N = 21,069

No Anti-DM Drug Therapy N = 4,666

TZD Only N = 7,156

Fig. 1. Flowchart of study group.

Table 1. Baseline characteristics of the control (non-DM) and study (DM) groups

Variables

Age, mean (SD) Male n (%) Charlson score index, mean (SD) Income, mean (SD) Antidiabetic medications Insulin n (%) Metformin n (%)

Controls

Study patients with diabetes mellitus

Non-DM patients (n = 131 564)

Other anti-DM drugs (n = 21 069)

No anti-DM drugs (n = 4666)

Prescribed TZDs (n = 7156)

P value

27
0 65 982 0
07 16 490

58
5 10 619 2
19 12 231

55
4 2153 2
22 15 901

55
9 3282 2
19 13 672