Introduction

Breast cancer has the highest incidence and mortality among all cancers in females worldwide, accounting for an estimated 1.7 million cases (25% of all new cancer cases) and 521 900 deaths (15% of total cancer-related deaths) in 2012 [1]. Tamoxifen is a selective estrogen receptor (ER) modulator. Its estrogen antagonist activity has been widely used to treat ER-positive breast cancer for approximately 4 decades [2]. Tamoxifen therapy for 5 years can reduce the annual mortality rate by 31% and prevent cancer over the next 20 years [3, 4]. Additionally, tamoxifen acts as an estrogen agonist in the bones, the endometrium, the coagulation system, and lipoprotein metabolism and exerts beneficial effects by maintaining the bone density and lowering the cholesterol level [2, 5] Tamoxifen is generally a safe and well-tolerated drug, but has certain adverse effects such as endometrial cancer, thromboembolism, and menopausal symptoms of clinical concern [2]. Moreover, 9 case reports have described tamoxifen-induced AP with severe hypertriglyceridemia and one event of death [6, 7].

AP is an acute inflammation of the pancreas and is the leading cause of hospitalization among all gastrointestinal disorders in the United States (274 119 discharges were reported in 2009; USD 2.6 billion per year are spent on in-hospital costs) and several other countries [8, 9]. AP-associated inflammation may expand to local and remote extra pancreatic tissues. It is a major public health concern owing to its significant potential mortality and morbidity, with an increasing rate of incidence and hospitalization [8–10]. In addition, several routinely prescribed drugs increase the risk of AP; medication use was reported as the cause of 0.1%– 5% of AP cases [11, 12]. Increasing time trends were reported in a proportion of first-attack AP patients on pancreatitis-related drug therapy from 2003 to 2012 [11].

Based on the WHO database from 1968 to 1993, 525 different drugs were suspected to cause AP [13]. However, overdependence on spontaneous adverse drug reactions or case reports is likely to exaggerate the true risk, which might therefore remain controversial without deriving any conclusion regarding the correlations between drugs and the risk of AP [13, 14]. Thus, the aforementioned case reports have yielded uncertain results regarding the association between tamoxifen use and the risk of AP. To address this clinical concern, we designed a formal pharmacoepidemiological study with a cohort design using nationwide population-based data to investigate whether tamoxifen use is correlated with the risk of AP in Taiwanese patients with breast cancer.

Methods

Results

Table 1 lists the baseline characteristics, comorbidity statuses and treatment of the breast cancer patients in all tamoxifen users or tamoxifen nonusers patients and the PS-matched cohorts. In non- matching cohorts, the mean age in the tamoxifen users group was 51.4 ± 11.6 years and 53.4 ± 12.0 years in the tamoxifen nonusers group, respectively. DM, HCV, hyperlipidemia, CAD, COPD, and asthma were more prevalent in the tamoxifen nonusers group than in the tamoxifen users group. There’s significant difference of medication history of fluorouracil use between the two groups (54.5% vs. 59.7%; P<0.001). No significant differences were observed for age, comorbidity and medication statuses between the tamoxifen users and the PS-matched tamoxifen nonusers groups.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Distributions of demographic variables and comorbidities in breast cancer patients with and without propensity score matching.

https://doi.org/10.1371/journal.pone.0173089.t001

After adjustment for age, ARD, DM, HBV, HCV, gallstones, hypertriglyceridemia, obesity, hyperlipidemia, CAD, COPD, and asthma and medication use including fluorouracil and doxorubicin, the risk of AP was not significant between tamoxifen users and tamoxifen nonusers (adjusted HR = 0.94, 95% CI = 0.74–1.19) (Table 2) in the non-matching cohorts. The results revealed no dose–response trend between tamoxifen use and the risk of AP (adjusted HR = 0.98, 95% CI = 0.96–1.00). The comorbidities DM and gallstones were associated with a significantly increased risk of AP. Similar trends were observed in PS-matched cohorts (S1 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Crude and adjusted hazard ratios and 95% confidence intervals of the incidence of acute pancreatitis in tamoxifen users and nonusers according to a time-dependent Cox regression model without propensity score matching.

https://doi.org/10.1371/journal.pone.0173089.t002

Discussion

In this present study, we observed no statistically significant correlation between tamoxifen use and the risk of AP, but detected a marginal reduction in the risk of AP after tamoxifen use (adjusted HR = 0.93, 95% CI = 0.73–1.18). We did not observe any dose–response trend between tamoxifen use and the risk of AP (adjusted HR = 0.98, 95% CI = 0.96–1.00). To our knowledge, the present study is the first population-based cohort study to examine the correlation between the use of tamoxifen and the risk of AP, and no other data could be compared with its results. We reviewed related articles to validate the biological plausibility of the finding as follows:

First, tamoxifen exerts estrogen agonist activities on serum lipid metabolism. Kim et al and Singh et al reported that all nine patients with tamoxifen-induced AP had hypertriglyceridemia (as an intermediary) [6, 7, 20]. Moreover, in a 9-year follow-up cohort study, Lim et al used the NHIRD to demonstrate that tamoxifen use in patients with breast cancer was associated with a decreased risk of hyperlipidemia; however, the researchers failed to mention whether tamoxifen use was associated hypertriglyceridemia or the risk of AP [21]. Subsequently, our review of studies analyzing the effect of tamoxifen use on serum hypertriglyceridemia concentration revealed the following important observations: Twelve case reports on tamoxifen-induced severe hypertriglyceridemia (serum triglycerides >1000 mg/dL) demonstrated that only 4 patients showed induced pancreatitis and most of these patients had a personal or family history of hyperlipidemia [22]. In addition, a retrospective study assessed the effect of tamoxifen use on changes in triglyceride levels and suggested that triglyceride levels significantly increased only in patients with a disturbed triglyceride-rich lipoprotein metabolism (susceptible patients) [23]. Furthermore, there was evidence that tamoxifen can induce hypertriglyceridemia in susceptible patients [22]. Because the NHIRD does not record the status of triglyceride-rich lipoprotein metabolism, we could confirm that tamoxifen use is no significantly correlated with AP in patients with breast cancer, but we could not confirm whether susceptible patients carry an insignificant risk (HR) of AP or whether non susceptible patients carry no risk of AP. Additional prospective studies are warranted to address this concern.

Second, tamoxifen, in addition to its partial estrogen agonist activity on lipoprotein metabolism, can induce severe hypertriglyceridemia and pancreatitis [2, 22]. Because of its estrogen antagonist activity through competitive inhibition at the estrogen binding sites on ERs, the drug is used in ER-positive breast cancer [2]. However, whether the estrogen antagonist activity exerts a protective effect on the pancreas remains unknown. Therefore, we researched some relevant literature to address this concern. A review article reported that ERs and estrogen-binding proteins are present in the healthy human pancreas [24]. Tamoxifen, when combined with other chemotherapeutics, was effective in treating pancreatic cancer in phase II trials, independent of the hormone receptor status [25,26]. Moreover, another study reported that tamoxifen exerted a cytotoxic effect against tumor cells of the pancreas, independent of the hormone receptor status [27]. All these studies indicate the effectiveness of tamoxifen in pancreatic cancer cells, with no mention of the protective effect of the drug on the pancreas and its association with a decreased risk of AP. The estrogen antagonist activity of tamoxifen may override the estrogen agonist activity, and a protective effect was demonstrated in some cases. This partly explains why we observed a marginal decrease with no statistical significance in the risk of AP after tamoxifen use. However, additional prospective studies are warranted to address the concern regarding the correlation between the estrogen antagonist activity of tamoxifen use and the risk of AP.

Alcohol is a known cause of acute and chronic pancreatitis. We used alcohol-related diseases (including alcoholic psychoses, alcohol dependence syndrome, alcohol abuse, alcoholic fatty liver, acute alcoholic hepatitis, alcoholic cirrhosis, alcoholic liver damage, excessive blood level of alcohol, and personal history of alcoholism) instead of alcohol use. Unexpectedly, on the bases of the multivariable analysis, alcohol-related diseases (adjusted HR = 1.56, 95% CI, 0.50–4.88) was not associated with AP. (Table 2) There might be the difficulty of documenting heavy alcohol use but prior to the onset of alcoholic liver disease. The ICD-9 codes document alcohol-related diseases, but does not necessarily include regular alcohol use prior to a presentation of alcohol disease. A prospective research need to clarify this issue.

Our study has several strengths. The NHIRD database, its extensiveness, and its ability to be used in addressing several clinical questions were extensively reviewed by Hsing et al. [28]. Breast cancer is considered a catastrophic illness by the NHI, and most medical copayments and drug cost-sharing are exempted. Therefore, tamoxifen use does not tend to change with the social class, reducing the selection bias in our study. Although the use of AP diagnoses coded according to the ICD-9-CM was documented by Lai et al and Hung et al in well-known lituretures, this claims dataset lack of serum amylase (or lipase) level, and abdominal imaging (eg, computed tomography scan) findings to confirm the AP diagnoses [17, 19]. The concern about misclassification of AP diagnosis might be occurred with a low probability, because the disease diagnosis without valid supporting clinical findings may be considered a medical fraud by NHI with a penalty of 100-fold of the payment claimed by the treating physician or hospital. However, the high coverage rate of the Taiwan NHI and high accessibility to Taiwan’s medical care service might make those omissions negligible. Therefore, the AP diagnosis by ICD-9 code in this study should be reliable. Finally, data on tamoxifen use did not depend on interviews and questionnaires, thus preventing any self-reporting errors and avoiding recall bias.

Nonetheless, some limitations should be considered when interpreting our study data. First, the NHIRD does not provide detailed information on potential covariates such as lifestyle habits, diet, and obesity (body mass index, waist circumference data) [19]. However, our study patients were limited to patients with breast cancer, thus partially reducing selection bias. The NHIRD data covers more than 99% of residents in Taiwan and the reimbursement policy is universal. Thus, it was unlikely for these covariates to differ significantly between the drug users and nonusers. Second, partial selection bias might have occurred because the tamoxifen users were more closely monitored and thus had a higher likelihood of being diagnosed with AP or undergoing serum triglyceride level evaluation and correction. However, our results demonstrated a null association, with a marginal decrease in the risk of AP after tamoxifen use. Third, we could not determine whether the patients actually consumed tamoxifen, because the NHIRD simply records the prescriptions rather than the actual dose consumed by patients. Because tamoxifen is usually prescribed to patients diagnosed with breast cancer, the prescription rate and patient compliance were probably relatively high owing to the low risk–benefit ratio of the drug. In addition [2], in a large-scale population-based study, the effects of tamoxifen prescription noncompliance can be neglected. Fourth, genetics may be a factor that predisposes people to AP [14, 18]. Despite using a large sample, information on serum triglyceride levels and genetic testing results was unavailable for our dataset; further, these findings only represent the Taiwanese ethnic group [19].

In conclusion, no significant correlation between tamoxifen use and the risk of AP was observed in patients with breast cancer. According to the foregoing discussion, additional prospective studies are warranted to address the concerns.

Supporting information

Acknowledgments

This study is supported in part by Taiwan Ministry of Health and Welfare Clinical Trial Center (MOHW105-TDU-B-212-133019).