Does increased body mass index lead to elevated prostate cancer risk? It depends on waist circumference

We examined the association between obesity and prostate cancer based on both body mass index (BMI) and waist circumference (WC) using the National Health Insurance System (NHIS) database for the entire male population of Korea. A total of 1,917,430 men who underwent at least one health examination in 2009 without a previous diagnosis of any other cancer were tracked through December 2015. The hazard ratio (HR) and 95% confidence interval (CI) value for the association between prostate cancer and obesity were analyzed using multiple Cox regression model. Since there was a statistically significant interaction between WC and BMI, a multiple HR for prostate cancer was estimated with stratifying both WC and BMI to control the interaction between WC and BMI. Without considering WC as an adjustment factor, very weak association between BMI and prostate cancer development risk was observed. When WC was considered as an adjustment factor, no significant change in the HRs for prostate cancer development beyond the reference BMI was observed in the group with WC < 85 cm in the multivariable-adjusted models. However, in the group with WC ≥ 85 cm, the HRs for prostate cancer increased as the BMI increased beyond the reference BMI. In addition, there was a discrepancy in the trend of prostate cancer development according to BMI among the groups with different categories for WC. In groups with abdominal obesity, a significant linear relationship was observed between increasing BMI and prostate cancer risk. Higher the WC category, the stronger was the association with BMI, signifying that the association of BMI with risk of prostate cancer development depends on abdominal obesity.

between obesity and prostate cancer risk have also been analyzed in multiple studies worldwide [7][8][9][10]. However, these studies have varied substantially in their results depending on the research methods used.
The existing discrepancies might be because most of the studies used body mass index (BMI) to analyze the relationship between obesity and prostate cancer. Though BMI provides the most common estimate of obesity in cancer epidemiologic studies, it does not measure adipose mass sufficiently [11]. Especially in men, BMI correlates better with lean mass than with body adiposity, and it is hypothesized that metabolically obese but normal-weight (MONW) individuals might have a normal BMI [12].
Consequently, other measures that could capture centralized fat disposition such as waist circumference (WC) should be estimated in combination with BMI when considering obesity as a risk factor for prostate cancer development. Therefore, in this study, we examined the association between obesity and prostate cancer based on both BMI and WC using the National Health Insurance System (NHIS) database for the entire male population of Korea.

Data source
The NHIS database of Korea is a public database comprising of the eligibility database, the national health screening database, the health care utilization database, and the long-term care insurance database [13]. This database includes the insurance claim code based on the International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM). Prostate cancer is coded C61. To identify clearly the diagnosis of prostate cancer, patients who had not undergone trans-rectal prostate biopsy were excluded from the study. In addition, the national health screening database was used to obtain information about WC, BMI, and other variables (hypertension, diabetes, dyslipidemia, alcohol consumption status, and smoking status).

Study population
Of the 5,860,389 men who underwent at least one health examination in 2009, men with prostate cancer or other cancers diagnosed before 1 st January 2009 (n = 253,157) were excluded. Moreover, patients aged < 50 years (n = 3, 684,955) were excluded because prostate cancer is rare in this age group. After excluding people with missing WC or BMI data from health examination databases (n = 4847), a total of 1,917,430 men without a previous diagnosis of any other cancer were tracked through 31 th December 2015. The study design and disposition of the subjects are shown in Fig. 1.

Statistical analysis
The baseline characteristics of subjects have been presented as the number (%) for categorical variables. Incidence rate is expressed as the number of newly diagnosed cases of prostate cancer per 100,000 person-years of the follow-up period. The hazard ratio (HR) and 95% confidence interval (CI) value for prostate cancer of obesity were analyzed using a multiple Cox proportional hazard model. The model was adjusted for common variables such as age, diabetes, hypertension, dyslipidemia, alcohol consumption, and smoking status; the obesity indices (BMI and WC) were alternatively added on common variables for adjustment in multiple Cox proportional hazard model. There was a statistically significant interaction between WC and BMI (P value < 0.05). To control the interaction between WC and BMI on the incidence of prostate cancer, a multiple HR for prostate cancer was estimated with stratifying both obesity indices (WC; four classes and BMI; five classes) were classified into, respectively. Statistical significance for linear trends in the HRs for prostate cancer of BMI and WC were verified at a level of 0.05. SAS software (version 9.4, SAS Institute, Cary, NC, USA) was used for statistical analyses.

Results
Baseline clinical characteristics of study population according to categories for WC Among a total of 1,917,430 men, 22,584 (1.18%) incident cases of prostate cancer developed between the beginning of 2009 and the end of 2015 (Table 1). Approximately 1.31% of the population categorized as WC ≥ 90 cm was diagnosed with prostate cancer, while 1.13% of the population categorized as WC < 90 cm was diagnosed with prostate cancer. In the obesity group categorized as WC ≥ 90 cm, 3.05% had a normal weight according to the BMI, and 20.65% of the men categorized as the non-obesity group (WC < 90 cm) were obese class 1 according to the BMI. In addition, the obesity group was more likely to have hypertension (54.45% vs. 36.25%), dyslipidemia (26.72% vs. 18.14%), and diabetes (20.65% vs. 13.65%).

Risk of prostate cancer according to BMI or WC
The HR for prostate cancer was lowest in subjects with a BMI < 18.5 and highest in those with 23.0 ≤ BMI < 25 in the multivariate-adjusted model (HR 0.93, 95% CI 0.85 to 1.02 and HR 1.12, 95% CI 1.08 to 1.16). However, the HRs for prostate cancer showed a decreasing trend with BMI over 25. Even in the patients with BMI ≥ 30, the 95% CI reflected the relative risk of 1.00. So, when we did not consider WC as an adjustment factor, there was a very weak association between BMI and prostate cancer development risk. The risk of prostate cancer according to WC also showed no statistically significant difference among the groups without stratification by BMI (Table 2). We also found that there was a statistically significant interaction between WC and BMI in this cohort (P value < 0.05).

The association between BMI and prostate cancer development according to WC
Since there was a statistically significant interaction between WC and BMI, a multiple HR for prostate cancer was estimated with stratifying both WC and BMI to control the interaction between WC and BMI. In the group with WC < 85 cm, there was no significant change in the HRs for prostate cancer development beyond the reference BMI in the multivariable-adjusted model (P for trend = 0.158; Table 3 and Fig. 2). The HR (95% CI) was 1.05 (1.01, 1.09) in the overweight group, which was characterized by a BMI over 23, and 0.99 (0.92, 1.07) in class 1 obese group, which was characterized by a BMI over 25.
However, in the groups with WC ≥ 85 cm (groups with 85 ≤ WC < 90, 90 ≤ WC < 95, and WC ≥ 95), the HRs for prostate cancer increased as the BMI increased beyond the reference BMI (P for trend < .001). In addition, there was a discrepancy in the trend of prostate cancer development according to BMI among the groups with different categories for WC (Fig. 2). The increasing rate of prostate cancer development risk according to BMI was greatest in the group with WC ≥ 95. On the other hand, HRs for prostate cancer according to WC stratified by BMI showed that impact of BMI on the risk of prostate cancer according to WC was weak (Table 4). Figure 3 shows the HRs for prostate cancer development according to BMI (comparison of BMI < 25 and BMI ≥ 25) when stratified based on abdominal obesity (comparison of WC < 90 and WC ≥ 90). The HR for prostate cancer was significantly higher in WC ≥ 90 groups regardless of BMI than the reference group (P value < .001). However, in the WC < 90 groups, HR did not show a significant difference even if BMI was higher than 25 (P value = 0.301).

Discussion
The main findings of this population-based study are (1) without considering WC as an adjustment factor, very weak association between BMI and prostate cancer development risk was observed; (2) when WC was considered as an adjustment factor, there were significant linear relations between increasing BMI and prostate cancer risk in the groups with abdominal obesity; (3) the higher the WC category, the stronger association with BMI was noted; and (4) this means that the association of BMI with risk of prostate cancer development depends on abdominal obesity. Although the terms "overweight" and "obese" are similar, the difference between the two depends on the BMI. Although BMI is the most widely used method of measuring obesity, an imperfect measurement does not accurately reflect adiposity. BMI indicates overweight relative to height, but it does not discriminate between fat mass and lean body mass. Body composition is variable among individuals with the same BMI [11,12]. Therefore, majority of the reported studies based on BMI provide an agreement that obesity is a risk factor for aggressive prostate cancer development, whereas other studies provide contrary results about localized or overall prostate cancer risk [15][16][17]. The inconsistent findings might be due to limitations in BMI as a calculation of adiposity.
In the present study, low WC and high BMI may be indicative of high lean body mass. In the low WC group (WC < 85 cm), adiposity of men with high BMI is expected to be smaller due to high lean body mass. Furthermore, there was a discrepancy in the trend of prostate cancer development according to BMI among the groups with different categories for WC. This finding signifies that in the higher WC group, men with high BMI have more adiposity than in the lower WC group, which increases the risk of prostate cancer development.
The distinctive feature of our study is that though few studies have simultaneously evaluated the associations between BMI, WC, and prostate cancer risk in the west [18,19], to the best of our knowledge, this is the first study to evaluate the risk for patients in an Asian population with prostate cancer stratified by BMI and WC. Especially, it is noteworthy that our results are based on the Asian population. Due to ethnic differences in body composition like muscle mass and fat distribution    between Westerners and Asians [20,21], research on evaluating risk of prostate cancer development based on obesity in Asians is necessitated. Because obesity also affects cancer screening in various aspects, research on the relationship between obesity and cancer development is important. Increasing BMI is associated with a decrease in serum PSA, which may minimize the diagnosis of prostate cancer based on PSA screening. When considering indications for prostatic biopsy in obese men, we should be aware that the effect of hemodilution might reduce PSA levels [22,23]. Obesity has been correlated with large prostate volume [24]. In case of larger prostate volume, it becomes difficult to get a diagnosis of prostate cancer in the same number of samples. Additionally, with adiposity impeding physical examinations, obese men may face difficulties in undergoing a DRE due to body habitus hindering prostate access [25]. Chu et al. reported that the predictive value of DRE is dependent on obesity and is significantly higher among obese men than normal-weight men [26]. Because DRE may be limited by adiposity in obese men, cancers that are large enough to be palpable may be more meaningful. Consequently, the clinician should remind these effects during the diagnostic process in patients with high BMI or high WC.
The major limitation of our study is that detailed information, such as digital rectal examination (DRE) finding, serum PSA level, prostate volume, cancer stage, Gleason grade, and molecular pathology, were not available for this nationwide cohort; thus, we could not adjust for these factors. The treatment patterns and the overall prognosis differed among regionally localized intermediate, high, and very high-risk prostate cancer. In addition, as previous studies have reported different effects of obesity [15][16][17], failure to perform stage analysis is an important limitation of our study. Therefore, detailed information should be thoroughly considered in future etiologic research.

Conclusions
Without considering WC as an adjustment factor, very weak association between BMI and prostate cancer development risk was observed. However, when we considered WC as an adjustment factor, in the groups with abdominal obesity, significant linear relationship was observed between increasing BMI and prostate cancer risk, and the higher the WC category, the stronger association with BMI was noted. This means that the association of BMI with risk of prostate cancer development depends on abdominal obesity. Therefore, WC should be mutually estimated while considering obesity as a risk factor of prostate cancer development.