Ethnicity Is an Independent Determinant of Age-Specific PSA Level: Findings from a Multiethnic Asian Setting

Objectives To study the baseline PSA profile and determine the factors influencing the PSA levels within a multiethnic Asian setting. Materials and Methods We conducted a cross-sectional study of 1054 men with no clinical evidence of prostate cancer, prostate surgery or 5α-reductase inhibitor treatment of known prostate conditions. The serum PSA concentration of each subject was assayed. Potential factors associated with PSA level including age, ethnicity, height, weight, family history of prostate cancer, lower urinary tract voiding symptoms (LUTS), prostate volume and digital rectal examination (DRE) were evaluated using univariable and multivariable analysis. Results There were 38 men (3.6%) found to have a PSA level above 4 ng/ml and 1016 (96.4%) with a healthy PSA (≤4 ng/ml). The median PSA level of Malay, Chinese and Indian men was 1.00 ng/ml, 1.16 ng/ml and 0.83 ng/ml, respectively. Indians had a relatively lower median PSA level and prostate volume than Malays and Chinese, who shared a comparable median PSA value across all 10-years age groups. The PSA density was fairly similar amongst all ethnicities. Further analysis showed that ethnicity, weight and prostate volume were independent factors associated with age specific PSA level in the multivariable analysis (p<0.05). Conclusion These findings support the concept that the baseline PSA level varies between different ethnicities across all age groups. In addition to age and prostate volume, ethnicity may also need to be taken into account when investigating serum PSA concentrations in the multiethnic Asian population.


Introduction
Prostate cancer screening is widely practiced amongst primary care physicians and specialists, particularly in affluent western countries such as the United States and Canada [1,2]. Latest reports of the European Randomized Study of Screening for Prostate Cancer (ERSPC) and Prostate, Lung Colorectal and Ovarian (PLCO) Cancer Screening Trial indicate that the prostate-specific antigen (PSA) testing increased the rate of overdiagnosis of patients with indolent and nonaggressive forms of prostate cancer to an estimated 17% to 50% [3,4]. This is a major concern as a man with cancer that would remain asymptomatic for the remainder of his life is unlikely to benefit from screening or treatment. Therefore, the U.S. Preventive Services Task Force currently recommended against PSA screening in asymptomatic man [5]. New methods that improve the sensitivity and specificity of serum PSA level are essential particularly in identifying men at risk of high grade, aggressive prostate cancer.
One of the promising, non-invasive approaches is the application of prostate cancer risk prediction model which incorporates other indications for biopsy, in addition to serum PSA level. Evidence from the Prostate Cancer Prevention Trial (PCPT), the European Randomized Study of Screening for Prostate Cancer derived Prostate Risk Indicator (SWOP-PRI) and the Montreal cohort studies demonstrated that variables including age, ethnicity, family history of prostate cancer, prostate volume and digital rectal examination (DRE) could improve the positive predictive value of PSA [6][7][8].
With insights from western population-based studies, it is of our interest to study and improve the understanding of PSA profiles amongst Asian men. Malaysia is a high middle-income country in Southeast Asia with a multi-ethnic population encompassing mainly Malays, Chinese, Indians and indigenous races. We conducted a cross-sectional population-based study in the Klang Valley region, Malaysia to assess the distribution of PSA level across different age as well as ethnic groups and determine the association between age, ethnicity, family history of prostate cancer, height, weight, lower urinary tract voiding symptoms (LUTS), DRE and prostate volume on elevated PSA levels.  13.2%) and 'other ethnicities' (8; 0.7%). Of these, only subjects with no history of prostate cancer, previous history of prostate surgery or 5a-reductase inhibitor treatment of known prostate conditions were selected for further analysis (N = 1160). All of them were aged between 40 and 79 years. Men from the 'other ethnicities' were excluded from the study because the number was too small to enable a meaningful analysis.

Study subjects and samples collection
Demographic and clinical data such as ethnic background, family history of prostate cancer information, height, weight and urological voiding history (International prostate symptom score, IPSS) [9] was ascertained using a structured questionnaire (proforma) requiring input from patients interviews, as well as medical examinations. All men underwent digital rectal examination (DRE) and transrectal ultrasonography (TRUS) to detect abnormalities in the prostate gland and determine the prostate volume. Blood sample of each individual was collected and analysed for total PSA levels in three main centers using ADVIA Centaur PSA assay (Siemens Healthcare Diagnostic Inc, Muenchen, Germany), ARCHITECT total PSA assay (Abbott Laboratories, Abbott Park, Illinois, USA) and AxSYM total PSA assay (Abbott Laboratories, Abbott Park, Illinois, USA). The correlation between the three assays had been validated in a previous study [10].
Based on serum PSA, DRE and TRUS findings, 144 men found with PSA above 4 ng/ml were recommended for TRUS-guided prostate biopsy. Only approximately 40% (57/144) of these subjects agreed to undergo biopsy, of which 19 men (33.3%) were diagnosed with prostate cancer. With the exclusion of men with confirmed prostate cancer and those who refused the biopsy, 1054 men (90.9%) with no evidence of prostate cancer formed the study population.

Data analysis
The relationship between baseline serum PSA level and age was tested using Spearman rank correlation. For each ethnicity, the serum PSA level was calculated for the 25 th , 50 th and 75 th percentiles and stratified by age groups of 40-49, 50-59, 60-69 and 70-79 years. We then compared the PSA levels in our population with four other measurements performed in previous studies (Table 1). Similar approach was used to demonstrate the prostate volume and prostate density of the study population ( Table 2). Subjects were then categorized into two groups using the age-specific median PSA level as a cutoff value. Of note, men with a serum PSA level above the age specific median level had a higher risk for prostate cancer compared to those of PSA below this level [11][12][13]. Potential variables associated with serum PSA level were compared between these two groups, including ethnicity, family history of prostate cancer, height, weight, the presence of LUTS, DRE (presence or absence of a nodule) and prostate volume. Logistic regression was employed to study the associations of these variables, alone and in combination, with the age-specific PSA level. Approximately 25% of the men had missing values on one or more covariates which were likely to be missing at random. Missing values were imputed using multiple imputation [14] in SPSS for Windows version 21.0 (SPSS Inc., Chicago, Illinois, USA). All variables of the multivariable logistic regression model were included in the imputation model and 10 imputation sets were created. Two-tailed P value ,0.05 was termed as statistically significant.
There was a positive trend between serum PSA level and age (Table 1), achieving a formal statistical significance (r = 0.247; p = 0.000, Spearman rank correlation). Comparing the 10-year age groups, Indian had a relatively lower median PSA level than Malay and Chinese who shared a comparable median PSA value across all age groups (Table 1). This trend was also consistently observed in prostate volume in which Chinese and Malays had a higher prostate volume than the Indians (Table 2). Although the median PSA value and prostate volume of Japanese were fairly lower compared to our study population, both populations showed a similar value of the median PSA density across all age groups (Tables 1 and 2). In addition, Caucasian tends to have a relatively higher median PSA level and prostate volume, but a lower PSA density, than Asians (Tables 1 and 2).
Subjects were then grouped into two categories using the agegroup specific median PSA level as cut-off point. Summary estimates obtained after multiple imputation were similar to those in complete cases. Ethnicity, weight and prostate volume were found to be significantly associated with age-specific PSA level (Table 3). For instance, men of Indian ethnicity were 40% less Japanese [17] Korean [20] Chinese (China) [19] Chinese  likely to have an above-median age-specific PSA level than the Malays (p = 0.007). On the other hand, subjects with a prostate volume $30 ml was 3.5-fold more likely than those of a lower prostate volume to have a PSA level above the age-specific median (p,0.001). With every 1 kg increase in body weight, there was a 1% decline in the probability of having an above-median agespecific PSA level (p = 0.003). We did not find any significant associations between age-specific PSA level and family history of prostate cancer, height, LUTS or DRE. In the multivariable analysis, it was found that ethnicity, weight and prostate volume remained to be significantly and independently associated with age-specific median PSA level (Table 4).

Discussion
Findings from this study provide an insight into the baseline PSA profile, prostate volume and PSA density of a large cohort of multiethnic Asian men. We demonstrated that there were significant ethnic variations in age-specific PSA levels, independent of body weight and prostate volume.
The elevated serum PSA concentration is often seen in men with benign prostatic hyperplasia (BPH), prostatitis or prostate cancer in some cases. Based on the current study, the detection rate of prostate cancer amongst men with PSA.4 ng/ml who underwent TRUS-guided biopsy was 1.8% only (19/1054), raising a concern of the existing PSA threshold that is based on Caucasian populations. Firstly, the prostate cancer incidence in Asia has increased in the last two decades with a higher proportion of patients being diagnosed with high grade tumor in comparison to the Caucasians [15]. Secondly, evidence from previous studies revealed that Asian tends to have a higher PSA density but a lower level of PSA and prostate volume, in comparison to Caucasians (Tables 1 and 2) [16,17]. Therefore, it is essential to understand the baseline PSA profile and its associated factors in Asian populations. Notably, the baseline PSA level is a strong indicator for predicting subsequent clinically diagnosed prostate cancer, raising the possibility of risk stratification in prostate cancer screening [11].
We demonstrated that age and ethnicity were amongst the significant variables associated with baseline median PSA level. Besides the significant association of serum PSA concentration with age (p = 0.000), our results showed that Indians were more likely to have a lower median PSA than Malay (p = 0.007; univariable analysis) whilst a comparable PSA level was described amongst Malay and Chinese (p = 0.077; univariable analysis) after adjusting for age-group. The age-dependent PSA trend was consistently observed in studies involving either Caucasian [16] or Asian [17][18][19][20] across all age groups. For instance, Osterling et al [17] found a direct correlation between the age of the Japanese men and the PSA level (correlation coefficient, r = 0.33, p,0.001); the serum PSA level of the Japanese was lower than the whites even after adjusting for age (p,0.001). Interestingly, results from our study demonstrated that the median PSA level of our population, except Indian, was relatively higher compared to other Asian studies but similar to the Caucasian (Table 1). These discrepancies could be due to methodological differences between the current and four previous reports [16], [17], [19] and [20], although the fundamental differences in the biology of the PSA in various ethnicities remain under investigation. For example, subjects with history of prostate cancer, prostatectomy or other specified conditions that would interfere with voiding function were excluded from the previous studies [16,17,19,20]. Conversely, subjects who did not undergo 5a-reductase inhibitor treatment for known or newly diagnosed LUTS (IPSS.7) were included in the current study, suggesting a potential skew towards higher baseline PSA levels. Although the serum PSA level was measured with different PSA assays in the five studies, there was an excellent correlation of these total PSA assays based on the results from earlier studies [10,[19][20][21][22]. For instance, the serum PSA concentration measured using the IMx instrument in [17] demonstrated a high correlation with those reported with Tandem-R (r = 0.99) [22], ELSA-PSA (p,0.001) [19] and AxSYM assays (r = 0.99) [21]. Of note, the latter three assays were used in [16], [19] and current studies respectively. Similar trend was observed in the comparisons of assays performed in the Caucasian [16] and Korean [20] studies with a strong correlation of r = 0.98 and p = 0.0001 [20]. In this study, the r 2 values were .0.983 when comparing ARCHITECT with ADVIA Centaur and AxSYM total PSA assays [10]. It is noteworthy that the presence of interassay variability may be resulted from the different epitope specificity of the antibodies used [23].
Our finding revealed that the PSA level was significantly associated with prostate volume, in parallel with previous studies [24][25][26][27][28][29][30][31][32]. Being the most significant predictor of PSA levels with an odds ratio of 3.79 in the multivariable analysis, it is suggested that prostate volume could be used as a tool for estimating the degree of prostate enlargement accurately, and for therapeutic decisionmaking [33]. Notably, the prostatic cellular composition differs between Asians and Whites in which the prostates of Chinese have significantly more glandular lumens but less connective tissue and smooth muscle than Caucasian [34]. We also observe significant inverse association between PSA level and body weight in the multivariable analysis. It is hypothesized that the lower PSA level associated with obesity is due to PSA haemodilution, a condition in which the total amount of PSA in the blood (PSA mass) is unaffected by an increased blood volume in obese man [31,32]. Other factors including difficulties in performing a thorough DRE and larger prostates might lead to the detection bias and fewer early stage of prostate cancer diagnosed in obese individuals [35,36]. A positive trend was demonstrated between the presence of LUTS (IPSS score .7) and the baseline serum PSA concentration albeit not achieving statistical significance (p = 0.180; univariable analysis). Subjects with LUTS may have an elevated PSA level resulting from BPH-related prostatic enlargement or prostatic inflammation; nevertheless, there are several other causative factors of LUTS including the aging of the bladder muscle, a dynamic component of the smooth muscle tone of the prostate, metabolic factors and serum sex steroid [37][38][39][40]. Results from a recent Korean study demonstrated that IPSS did not improve the predictive value of PSA in spite of their significant correlation [30]. There was no clear association between DRE and the PSA level (Table 3). Nevertheless, the presence of nodule during DRE was significantly and positively associated with diagnosis of prostate cancer (data not shown).
Our study findings provide evidence of the PSA profiles and their associated factors within a multiethnic Asian population. An important clinical implication of this study is that ethnicity may also need to be taken into account when investigating serum PSA concentrations [41] for prostate cancer detection in a multi-ethnic Asian population, in addition to age and prostate volume.