Study Population

We identified all individuals in this study that were diagnosed cancers for the first time (International Classification of Disease, Clinical Modification, Ninth Revision [ICD-9-CM] codes 104-208) in between January 1, 2001 and December 31, 2008 who were eligible cases and used the date of the cancer diagnosis as the index date (S1 in Appendix http://links.lww.com/MD/A181). The individuals without any cancer diagnosis during 12 years of the study served as controls. For each case, we selected 6 controls randomly among all individuals in the sample population, a propensity-score was matched for sex, age at cancer diagnosis, and year of diagnosis. Controls were assigned an index date identical to the date of diagnosis for the corresponding case.

Benzodiazepines Exposure

Information regarding patients’ medications was retrieved from the pharmacy prescription database. BZDs were classified as Anatomical Therapeutic Chemical (ATC) code N05BA (Anxiolytics), N05CD (Hypnotics and sedatives), N03AE (Anti-epileptics), and N05CF (Benzodiazepine related drugs) (Table S1 in Appendixhttp://links.lww.com/MD/A181). In each filled prescription for study participant, we recorded only oral drugs with drug name, dispensing data, and the total amount of the recommended defined daily dose (DDD)19 (ie, the assumed average maintenance dose per day).

The daily dose for BZD users was estimated as dose1 divided by |t1 − t2|, where dose1 is the prescription of a BZD before the date of cancer diagnosis, then measured the average defined daily doses (ie, the average milligrams dispensed, divided by each defined daily dose for specific BZD) (Table S1 in Appendixhttp://links.lww.com/MD/A181). The value of |t1 − t2| is the time duration of each BZD prescription prescribed before the index date (Figure S1 in Appendix http://links.lww.com/MD/A181). BZD doses were analyzed for defined daily dose per day in the following categories: 0.00 (reference), less than 0.10, 0.10 to 0.39, 0.40 to 0.69, 0.70 to 0.99 and more than or equal 1.00.

The BZD exposure was analyzed only before the cancer diagnosis/index date. We also considered whether individuals have had ever exposed to BZDs or not. In addition, we performed further analysis to compare individuals with cancer if they ever took a BZDs before their cancer diagnosis and compared with those who had never taken it (Figure S2 in Appendix http://links.lww.com/MD/A181).

Patients who had BZD prescriptions prescribed at least 2 months during the study period, were classified as BZD users (Figures S1 and S3 in Appendix http://links.lww.com/MD/A181). Exposure to drug was classified in windows size (ie, 61–90 days, 91–180 days, 181–1 years, 1–2 years, and over 2 years) before the index date. An additional category was created for “no users” where patients had been never or <2 months prescribed any benzodiazepines.20

Covariate Assessment

Propensity score was calculated using a logistic regression as proposed by Rosenbaum and Rubin21,22 to estimate the probabilities for patient classifications into the cancer (case) and non-cancer (control) groups as shown in Table ​Table1.1. The potential confounders were included in the study. The use of drugs known or suspected to modify the risk of some cancers, including aspirin (ATC: B01AC06, N02BA01, N02BA51), non-steroidal anti-inflammatory drugs (NSAIDs) (M01A, excluding M01AX), statins (C10AA) and angiotensin-II antagonists (C09C and C09D) were included in the study which might have potentials to influence for their carcinogenic effects.15 Exposure to these confounder drugs was defined as if it was dispensed at least twice per year within a period of 3 years to the date of diagnosis.

TABLE 1

Baseline Characteristics of Cancer Cases and Their Controls

Since the chance for cancer can be confounded by competing risk, therefore we also identified comorbidities that may be associated with mortality based on diagnostic codes from outpatient datasets prior to the outcome of interest. All diseases were included in the Charlson Comorbidity Index and analyzed except for human immunodeficiency virus (HIV).23

Moreover, the other confounding factors could influence to the risk of some cancers such as location (ie, Regions) and socio-economic status (ie, SES—based on the total amounts of payment to Taiwan's National Health Insurance) which were included in this study.

Data Analysis

We excluded from analyses patients with cancers who were <20 years of age, because such patients are unlikely prescribed benzodiazepines in Taiwan.

Two statistical approaches were used to analyze the data in the study. Firstly, all subjects of both case and control groups were measured the BZDs use 3 years before the date of diagnosis/index date (Figure S1 in Appendix http://links.lww.com/MD/A181). The conditional logistic regression were adjusted for potential confounders and used to investigate the association between exposure to the different drugs and risk for cancer. Our interest was to identify safer BZDs individual or each class on the occurrence of overall cancer. The results are expressed in an adjusted odds ratios (AORs) with 95% CI (confidence intervals).

Secondly, all 297,500 subjects were followed from the initial BZDs dispense date or the first visit date of the cohort database until a cancer diagnosis/index date or until the time subject was censored for loss to follow-up, or termination of insurance or to the end of 2009 (Figure S4 in Appendixhttp://links.lww.com/MD/A181). Subjects who were prescribed a BZD for at least 2 months before the date of diagnosis, defined as BZDs cohort use. Cox regression models with the time (in days) as the time scale were used to calculate hazard ratio with 95% CI. Multivariable Cox models were adjusted for these confounders listed in Table ​Table11.

We used the SPSS 20 software to perform data analysis and the results calculations were expressed as the estimated numbers together with 95% CI. Based on statistical power at 0.9, type I error rate at 0.05, and the individual numbers in both groups, the detectable risk difference was estimated to be 0.01.

Benzodiazepines Use and Cancer Risk

The multivariable-adjusted hazard ratio for overall cancers among BZDs users, as compared with patients who had never used BZDs, was 1.14 (95% CI, 1.10–1.17) (Figure ​(Figure1).1). We also found that the use of chlorodizepoxide (HR, 0.98; 95% CI, 0.92–1.04), diazepam (HR, 0.96; 95% CI, 0.92–1.00), lormetazepam (HR, 1.08; 95% CI 0.99–1.17), medazepam (HR, 1.01; 95% CI 0.84–1.21), nitrazepam (HR, 1.06; 95% CI, 0.98–1.14), oxazepam (HR, 1.05; 95% CI, 0.94–1.17) were not significantly risk for cancers, as compared with no BZDs use. These results were observed similar with the 3-year of benzodiazepines use before the cancer diagnosis (Table ​(Table2).2). Moreover, both multivariable-adjusted hazard ratios and adjusted odd ratios result were found consistent that oxazolam, zolpidem, and clonazepam (HR [95% CI], 1.15 [1.09–1.22]; AOR [95% CI], 1.22 [1.13–1.31]) were associated with a higher risk for cancers, as compared with no BZDs use. We also observed the risk of cancer for overall, individual and classes of benzodiazepines in both male and female (Figures S5 and S6 in Appendix http://links.lww.com/MD/A181).

FIGURE 1

Benzodiazepines (classes and individuals) and their association with overall cancer risk.

TABLE 2

Safe and Unsafe Benzodiazepines for Cancer Risk

Figure ​Figure22 presents the further analysis with types of specific cancer associated with benzodiazepines exposure. We observed the high risk for brain (AOR, 1.98; 95% CI, 1.58–2.47), colorectal (AOR, 1.25; 95% CI, 1.17–1.34), lung (AOR, 1.10; 95% CI, 1.01–1.20), oesophagus (AOR, 1.59; 95% CI, 1.26–2.00), and prostate (AOR, 1.36; 95% CI, 1.23–1.51) among BZDs users. However, there were not significant association found for ovary (AOR, 1.14; 95% CI, 0.92–1.42), stomach (AOR, 1.12; 95% CI, 0.97–1.28), and cervical cancer (AOR, 0.88; 95% CI, 0.76–1.02).

FIGURE 2

Overall benzodiazepines and their association with specific cancer risk.

Safe Benzodiazepines

From all BZDs, the chlordiazepoxide, diazepam, medazepam, oxazepam, and nitrazepam were observed to be safer drugs that means these drugs did not have any association with cancer. Our results are consistent with Rosenberg et al12 that chlordiazepoxide and diazepam have no risk for cancer however, contradict with Horrobin and Trosko27 that diazepam is possibly cancer promoters, and Iida et al9 that oxazepam use is risk for liver cancer. Since, diazepam is most frequently prescribed BZD in Taiwan to treat anxiety, panic attacks and insomnia. It appears to be safer from our findings as compare to other BZDs which could be because of BZDs varies in their therapeutic spectrum and activity 28.

Unsafe Benzodiazepines

For unsafe BZDs, we observed clonazepam, lorazepam, alparazolam, bromazepam, zolpidem, and zopiclone have high risk for cancer as examined with DDD and exposure duration. We found that clonazepam users have had 15% higher risk to develop cancer among all other BZD drugs. Our findings consistent with Rosenberg et al12 at some extent for few drugs which are safer but contradict with Kripke11 and Kripke et al16 investigations to have threefold greater cancer risk in hypnotic users. However, the benzodiazepines are relatively safer drugs as it rarely cause serious adverse effects.29 We assume that this could be due to aggregated risk for long term use of BZDs in relation to polypharmacy or metabolic related drugs which could have effect.30 Therefore, we need more attention to compute the aggregated risk of multiple drugs uses.

Benzodiazepines Use and Risk for Specific Cancer

We also observed that benzodiazepines exposure increased the overall cancer risk up to 21%, specifically for brain 98%, colorectal 25%, lung 10%, esophagus 59%, prostate 36%, bladder 39%, liver 18%, pancreas 41% and other cancers 27%. However, cervical, ovary, and stomach cancers were not observed statistically significant. These findings are important and have a positive impact for benzodiazepines users as it is commonly prescribed drugs. Our findings are consistent with Rosenberg et al,12 Kripke and Langer,31 and Cronin-Fenton et al32 that the BZDs use have only selected cancer risks but contradict with Pottegård et al15that there is no association found in Danish BZDs users.

In Taiwan, the bladder cancer incidence is particularly high and reported as the sixth common cancer in the world33 that we observed significantly risk for cancer among BZDs users. Moreover, our findings for breast cancer are consistent with Karmali et al7 and Horrobin and Trosko27 animal studies that breast cancer significantly associated with BZDs exposure but contradict with Halapy et al13 that there is no association for breast cancer. We also found similar results with Kao et al2 and Coogan et al34 who reported that BZDs use have a significant risk for prostate cancer in men but there were no risk for ovarian and cervical cancer in women. Although, some researchers might think that anxiety leads to cancers instead of BZDs or other drugs. However, Kao et al2 studied in individuals without anxiety using benzodiazepines still have had higher risk for developing cancers in Taiwan. Therefore, we assume that risk of cancers could be associated with individual BZD, which might have some relationship only with particular cancers etiology need to be identified. Since it was also reported that BZDs might be useful as adjuncts to some specific cancer chemotherapies.

We recommend that the therapeutic effectiveness of BZDs should be monitored closely for long-term users. Furthermore, the metabolism of these drugs should be investigated in relation to their carcinogenicity in accordance to multiple drugs use and multiple diseases. We believe that some BZDs are safer among others and should need to investigate them on large population.