The results of this population-based cohort study demonstrate significant associations between the use of zolpidem and the increased risk for overall cancer, as well as oral, liver, lung, breast, esophageal, bladder, and kidney cancers. These unexpected findings may arouse public interest regarding safety issues in zolpidem use.
According to the Taiwan Cancer Registry, since 1982 cancer has been the leading cause of death among the general population in Taiwan. The age-adjusted incidence rate has increased steadily, and in 2007 it reached 270 new cases per 100,000 people.14
However, data from Surveillance Epidemiology and End Results indicate a different trend, with overall cancer incidence rates reportedly decreasing by 0.7% per year between 1999 and 2006 for all racial and ethnic groups combined.15
Because this issue continues to be a challenge for public health in Taiwan, it has gained the attention of the government; the result has been more population-based investigations regarding cancer preventive epidemiology. The NHI program provides adequate data for such studies because of its comprehensive health coverage. The NHIRD contains ambulatory service records, hospital service records, and prescription claim data. It allows researchers to select specific study groups and matched comparison groups, with ensured representation of the underlying population groups. We recently used NHIRD data to evaluate the risk of malignant tumors in patients with end-stage renal disease and found some positive risk. A report on this study has been previously published.16
The current study used a similar design to determine whether zolpidem use relates to the risk of cancer.
To the best of our knowledge, the current study was the first population-based study of zolpidem users in Taiwan (N=14,950 patients). To create a comparison group, we randomly matched each zolpidem user with 4 people from the general population who did not use zolpidem.
As is the case with benzodiazepine use, the probability of a patient using zolpidem increased with age, female sex, and concomitant use of other psychotropics17,18
; our results confirmed this pattern. Kripke11,19,20
has extensively studied the association between hypnotics and cancer risk, and his published work addresses this issue. On the basis of epidemiological data and laboratory studies, Kripke suggests that new hypnotics may increase cancer risk.
Zolpidem use appeared to promote viral infections, which may have indicated suppression of the immune function. Either immunosuppression or viral infections might increase cancer development.11,20
Cancers known to be related to viral infections include oral, liver, and cervical cancers.21-23
Our results demonstrate that zolpidem users have a significantly higher rate of subsequent oral and liver cancer development but not cervical cancer. One possible reason to explain this finding is the effect of national cancer screening. Cervical, oral, breast, prostate, and colorectal cancers are the main targets of Taiwan's free cancer screening program. We assumed that zolpidem users visit physicians relatively frequently and would be screened for cancer whenever necessary. Regarding the carcinogenesis of cervical cancer, cervical intraepithelial neoplasia (CIN) and carcinoma in situ (CIS) lesions are known to precede the development of invasive cancer. Frequent Papanicolaou smear examinations would detect cervical CIS, CIN, and dysplasia, enabling the patient to receive treatment before the disease progresses into invasive cancer. We did not include CIN and CIS as end points for cervical cancer, and this omission may have decreased the number of cases of invasive cervical cancer that we extracted from the database.
The cancer screening effect also partially applies for oral and breast cancers, which showed significantly higher rates in zolpidem users. More cancers can be detected if more frequent cancer screening is conducted among users of zolpidem. Prostate and colorectal cancers, however, did not show any significant difference between zolpidem users and nonusers.
Zolpidem can reduce the arousal response to nocturnal acid exposure and can also increase the duration of each esophageal acid reflux event. Patients with gastroesophageal reflux disease who take zolpidem thus experience significantly higher esophageal exposure to gastric acid, which increases the likelihood of developing esophageal cancer.12
Our data support this theory, and analysis revealed a significantly higher rate of esophageal cancer among zolpidem users. Other cancers that had significantly higher rates in zolpidem users than nonusers were lung, bladder, and kidney cancers. To date, we do not have any biological hypotheses to explain these associations.
Zolpidem users may also take benzodiazepine; thus, benzodiazepine may also be associated with cancer risk. In addition, sleep disorder may be associated with cancer risk. We wanted to investigate whether the relationship between zolpidem use and cancer risk may actually arise from an interaction between zolpidem and other risk factors. clarifies this concern and shows that zolpidem independently increases the overall cancer risk. The dose level of zolpidem also plays an important role. As indicated in , higher doses of zolpidem were associated with greater cancer risk, especially for patients who took zolpidem without taking benzodiazepine. indicates that the group receiving benzodiazepine plus zolpidem had a lower HR than the group receiving zolpidem alone at higher zolpidem dosages (30-299 and ≥300 mg/y but not for the lower zolpidem dosage [1-39 mg/y]). Therefore, the data in should be considered cautiously because the zolpidem-only groups are too small and their HRs are implausible. In addition, complex drug interaction is a possible explanation.
Some readers may suspect reverse causality, which would imply that patients who are already at risk for cancer may have a greater tendency to start using zolpidem than those not at risk for cancer. To clarify this concern, we used the Kaplan-Meier method without adjusting for sex, age, and comorbidity to evaluate it. There was no time lag between the 2 groups (zolpidem vs nonzolpidem). We found that the cancer-free proportion of patients treated with and without zolpidem differed significantly over time (). This finding indicates that cancer risk increases in conjunction with the length of time of zolpidem use; thus, reverse causality seems unlikely.
The current study had some limitations that need to be addressed. First, information on smoking habit, alcohol consumption, body mass index, and family history of cancer was unavailable from the NHIRD. All of these are risk factors for multiple cancers and could plausibly also be associated with zolpidem use. The current study cannot control these possible confounders. However, we have included obesity and alcoholism as risk factors in the models and adjusted for them in the analyses. Second, the organ-specific pattern of cancer occurrence does not appear to correspond to any biological hypothesis of zolpidem action. We think it likely that our findings in part reflect a healthy nonuser effect (ie, those with healthier lifestyles and behaviors may be less likely to require zolpidem treatment or may be more likely to address such problems with nonpharmacological approaches). Third, evidence derived from any cohort study is generally less than that from randomized trials because a cohort study design is subject to many biases related to confounding adjustment. Despite our meticulous study design with adequate control of confounding factors, a key limitation is that bias could still remain if there are unmeasured or unknown confounders. Fourth, the diagnoses in NHI claims primarily serve the purpose of administrative billing and do not undergo verification for scientific purposes. We were unable to contact the patients directly regarding their use of zolpidem because of the anonymity of the identification numbers. In addition, prescriptions issued before 1996 for the drugs under study were not reflected in our analysis. This omission could have underestimated the cumulative dosage and may have weakened the observed association. Fifth, comorbidities were detected between 1998 and 2009. Because the comorbidities do not necessarily precede the use of zolpidem or benzodiazepine, they may be caused by zolpidem and benzodiazepine. This raises a question of whether control for comorbidities incident after zolpidem use is commenced may produce overcontrol and thus underestimation of causal associations. Overcontrol may be a conservative statistical choice. Finally, it is difficult to explain why the HR appears lower when any additional risk factor, except for alcoholism and benzodiazepine, is combined with the risk of zolpidem. Some undetermined mechanisms or intrinsic limitation of this study may lead to the results. Apart from these potential problems, the data on zolpidem prescription and cancer diagnosis were highly reliable.