Prevalence, Incidence and Risk Factors of Tamoxifen-related Non-alcoholic Fatty Liver Disease

A Systematic Review and Meta-analysis

Bora Lee; Eun-Ae Jung; Jeong-Ju Yoo; Sang Gyune Kim; Cheon-Beom Lee; Young Seok Kim; Soung Won Jeong; Jae Young Jang; Sae Hwan Lee; Hong Soo Kim; Baek Gyu Jun; Young Don Kim; Gab Jin Cheon

Disclosures

Liver International. 2020;40(6):1344-1355. 

In This Article

Results

Characteristics of Included Studies

Based on database searches of titles and abstracts, we identified 165 potentially relevant studies. Among them, 96 studies were excluded due to duplication and 45 studies were excluded because they were reviews (n = 5), case reports (n = 16), animal studies (n = 12), cell studies (n = 1), studies with overlapping populations (n = 2), inadequate data on the outcomes of interest (n = 8) and tamoxifen use for prophylactic purposes (n = 1). Finally, 24 unique studies were eligible for inclusion in the meta-analysis (Figure 1). As shown in Table 1, of these 24 studies, 1, 15 and 8 were an RCT, observational cohort studies and cross-sectional studies respectively. Eight studies had control groups, such as aromatase inhibitors, compared to tamoxifen. Information on the enrolled participants and comparators in each study are described in Table 1.

Figure 1.

Flowchart showing the process of identifying studies included and excluded in the systematic review

Fatty liver was diagnosed by imaging studies, either USG or CT. Most studies enrolled patients with or without menopause. One study enrolled pre-menopausal women only and two studies enrolled post-menopausal women only. The median age of the participants was 52.6 years (range, 43.5–60.7 years). The median follow-up period was 3.2 years (range, 0.3–8.0 years). The median duration of tamoxifen medication was 2.6 years (range, 0.3–5.0 years). The dosage of tamoxifen was 20 mg daily for all participants. Of the 24 studies, 18, 4 and 2 were conducted in Asia, the Middle East and the West respectively.

Prevalence and Incidence of Tamoxifen-induced Fatty Liver

We derived the pooled prevalence using a random-effects model and a QE model using the eight cross-sectional studies (Table 2). The prevalence of fatty liver among breast cancer patients taking tamoxifen was 40.25 per 100 patients [95% confidence interval (CI): 31.35–49.84] by the random-effect models and 50.36 per 100 patients (95% CI: 30.04–70.55) by the QE model.

The incidence rate (IR) was calculated from 16 studies (Table 3). The incidence rate ratio (IRR) was calculated from eight studies consisting of two groups, tamoxifen and control groups (Table 3). When all these studies were integrated, the IR of tamoxifen-induced fatty liver was 12.37 per 100 person-years (95% CI: 7.22–21.20) by the random-effects model and 11.71 per 100 person-years (95% CI: 6.7–21.53) by the QE model. The incidence rate ratio (IRR) was 3.12 (95% CI: 2.05–4.75, P = .016) (Table 3). In other words, tamoxifen increased the incidence of fatty liver 3.12 times compared to the control. Forest plots of all studies are shown in Figure 2A.

Figure 2.

Meta-analysis of the risk of fatty liver associated with tamoxifen. Forest plot comparing patients taking tamoxifen vs those not taking tamoxifen for (A) all studies and (B) subgroup analysis according to the region (eastern vs western), (C) subgroup analysis according to the diagnostic method (ultrasound only vs ultrasound plus CT vs CT only)

Subgroup Analysis Depending on Region and Diagnostic Tools

We then sub-analysed whether the occurrence of fatty liver differed by region or diagnostic method. Tamoxifen significantly increased fatty liver incidence in both the eastern and western regions. The IRR of the eastern population was 3.05 (95% CI: 1.97–4.73, P < .001) and the IRR of the western region was 5.80 (95% CI: 0.77–43.71; P -value, not applicable). Forest plots of the subgroup analysis according to regions are shown in Figure 2B.

Next, we sub-analysed the results depending on the diagnostic tools (Figure 2C). The diagnostic methods for fatty liver were divided into three groups as follows: (a) USG only, (b) CT only and (c) combined methods of USG and CT. Regardless of the diagnostic method, tamoxifen increased fatty liver development compared to the control group. Interestingly, IRR was calculated to be higher in the order of the CT only group (IRR 6.25, 95% CI: 0.91–43.16, P = .135), the combined use of USG and CT (IRR 3.21, 95% CI: 1.89–5.45, P = .544) and the USG only group (IRR 2.71, 95% CI: 1.36–5.39, P = .007). This is because the specificity of CT is higher than USG for diagnosing fatty liver.

Sensitivity Analysis and Meta-regression Analysis

Since the heterogeneity values (I 2) of the included studies were high, a sensitivity analysis was performed (Figure S1). After each study was removed in turn, the pooled IRRs from the random-effects model ranged from 2.96 to 3.62, which were not very different than the original value of 3.12. The funnel plot results are presented in Figure S2. Although the funnel plot in Figure S2 shows some asymmetry, the assessment of publication bias was difficult since the number of studies was small (less than 10 studies).

Next, we performed a meta-regression analysis to explain the heterogeneity between the studies and examine the influence of other factors on clinical outcomes. Because age and medication duration might play an important role in the development of tamoxifen-induced fatty liver, meta-regression was performed for age and medication duration. Although the development of fatty liver showed a tendency to slightly increase with age, meta-regression analysis revealed no significant relationship between age (P = .963) or medication duration (P = .680) and fatty liver development (Table S1). The bubble plot of the meta-regression analysis is shown in Figure S3.

Meta-analysis of Risk Factors

First, we identified all the risk factors described in the studies (Table S2). In our meta-analysis, a total of six risk factors were accessible for calculation: age, body mass index, menopause, hypertension, diabetes mellitus and hypercholesterolaemia (Figure 3; Table S3). Higher BMI increased the incidence of tamoxifen-induced fatty liver (HR: 1.15, 95% CI: 1.09–1.22, P < .001). On average, the BMI of patients with fatty liver development was about 2.23 kg/m2 higher than that of patients without fatty liver development (Figure S4). Hypercholesterolaemia also significantly increased the incidence of fatty liver (HR: 1.01, 95% CI: 1.0–1.02, P = .046). However, age, menopause, hypertension and diabetes failed to show significant associations with fatty liver development.

Figure 3.

Risk factors associated with tamoxifen-induced fatty liver

processing....