Taiwan is an independent country with more than 23 million citizens (Chan et al., 2016; Chang and Yu, 2016; Chang et al., 2016; Chen and Wu, 2016; Chen S.Y. et al., 2016; Chen Y.F. et al., 2016; Cheng et al., 2016a,b; Hsieh et al., 2016; Hsu and Yin, 2016; Huang and Chang, 2016; Lin and Lin, 2016; Maa and Leu, 2016; Ooi, 2016; Yu et al., 2016; Liang et al., 2017; Liao et al., 2017c; Wen and Yin, 2017). We conducted a population-based case-control study using the database of citizens enrolled in the Taiwan National Health Insurance Program. This insurance program began in March 1995 and the enrollment rate has exceeded 99.6% of the entire population of 23 million citizens living in Taiwan in 2015 (Ministry of Health and Welfare, 2016b). The study was approved by the Research Ethics Committee of China Medical University and Hospital in Taiwan (CMUH-104-REC2-115). The details of the program have been written down in previous studies (Lai et al., 2010; Cheng et al., 2012; Liao et al., 2012; Chen et al., 2013; Hung et al., 2013).

We identified subjects aged 20 years and more with newly diagnosed hepatocellular carcinoma (International Classification of Diseases, Ninth Revision, Clinical Modification, ICD-9 codes 155, 155.0, and 155.2) from 2000 to 2013 as the cases. The date of subjects being diagnosed with hepatocellular carcinoma was defined as the index date. Subjects without the diagnosis of hepatocellular carcinoma were randomly selected as the controls. Both cases and controls were matched with sex, age (5-year interval), and the year of index date. Subjects who had any other cancer (ICD-9 codes 140-208) before the index date were excluded from the study. The definition of subject selection was adapted from previous studies (Lai et al., 2013a,b).

Comorbidities which could be related to hepatocellular carcinoma were selected as follows: alcohol-related disease, cardiovascular disease, chronic kidney disease, chronic obstructive pulmonary disease, diabetes mellitus, hyperlipidemia, hypertension, as well as chronic liver disease including cirrhosis, hepatitis B, hepatitis C, and other chronic hepatitis. All comorbidities were diagnosed based on ICD-9 codes. The validity of ICD-9 codes has been discussed in previous studies (Hung et al., 2016; Lai et al., 2016, 2017a,b; Liao et al., 2016a,b, 2017a,b; Shen et al., 2016; Wong et al., 2016; Lin C.M. et al., 2017; Hung et al., 2017).

Prescription histories of selective serotonin reuptake inhibitors and other medications including metformin and statin were collected in the study. Selective serotonin reuptake inhibitors available in Taiwan during 2000–2013 were included as follows: fluoxetine, fluvoxamine, paroxetine, sertraline, citalopram, and escitalopram. The definition of medication use was found in previous studies (Cheng et al., 2017; Lai et al., 2017d,e; Liao et al., 2017d,e). That is, ever use of medications was classified as a subject who had at least a prescription of medications studied before the index date. Never use of medications was classified as a subject who never had a prescription of medications studied before the index date.

We assessed the differences of the demographic status, selective serotonin reuptake inhibitors use, other medications use, and comorbidities between the cases and the controls by using the Chi-square test for categorized variables. The t-test was used to examine the difference of mean age between the cases and the controls. Initially, all variables were included in the univariable logistic regression model. Only those significantly associated with hepatocellular carcinoma in the univariable logistic regression model could be further included in the multivariable logistic regression model. We measured the odds ratio (OR) and the 95% confidence interval (CI) for hepatocellular carcinoma associated with selective serotonin reuptake inhibitors use. All analyses were performed using the SAS statistical software (version 9.2; SAS Institute, Inc., Cary, NC, United States). The results were considered statistically significant when two-tailed P-values were less than 0.05.

In Table 1, we identified 4901 cases with newly diagnosed hepatocellular carcinoma in 2000–2013 and 19604 controls. Both cases and controls had similar distributions of sex and age. The mean ages (standard deviation) were 62.5 (12.3) years in cases with hepatocellular carcinoma and 62.3 (12.4) years in controls, without statistical significance (t-test, P = 0.44). The cases with hepatocellular carcinoma were more likely to have a higher proportion of ever use of selective serotonin reuptake inhibitors than the controls (7.57% vs. 6.05%, P = 0.001). In addition, the proportions of ever use of metformin, alcohol-related disease, chronic kidney disease, chronic liver disease, chronic obstructive pulmonary disease, diabetes mellitus, and hypertension were higher in the cases than the controls, with statistical significance (P < 0.001 for all).

In Table 2, all variables were initially included in the univariable logistic regression model. Only those significantly associated with hepatocellular carcinoma in the univariable logistic regression model could be further included in the multivariable logistic regression model. Therefore, only metformin use, statin use, alcohol-related disease, chronic kidney disease, chronic liver disease, chronic obstructive pulmonary disease, diabetes mellitus, hyperlipidemia, and hypertension could be further included for adjustment. After adjustment for co-variables, the multivariable logistic regression model showed that the adjusted OR of hepatocellular carcinoma was 0.92 for subjects with ever use of selective serotonin reuptake inhibitors (95% CI 0.79, 1.08), comparing with never use. The statistical power of our study to detect a clinically important difference is small (statistical power = 0.24).

In Table 3, among subjects with any one of comorbidities including alcohol-related disease, chronic liver disease, and diabetes mellitus, the adjusted OR of hepatocellular carcinoma was 0.89 for subjects with ever use of selective serotonin reuptake inhibitors (95% CI 0.75, 1.06), comparing with never use.

There are some strengths in this study. Our study focuses on patients with any one of the comorbid conditions associated with hepatocellular carcinoma. The results are well-structured and explained. The readers are more convinced to the results.

There are some limitations in this study. First, due to the limitation of case-control study, a causal-effect relationship cannot be established. Second, due to the inherent limitation of the database used, the status of alcohol consumption is not recorded. Thus, we used alcohol-related disease for analysis. This point was mentioned in previous studies (Lai et al., 2017c; Lin H.-F. et al., 2017). Third, due to the inherent limitation of the database used, we could not determine whether selective serotonin reuptake inhibitors were used as a mono-therapy for the treatment of depressive disorder or anxiety disorder.