An Up-to-date Meta-analysis of Coffee Consumption and Risk of Prostate Cancer Jia-Dong Xia1, Jie Chen2, Jian-Xin Xue1, Jie Yang1, Zeng-Jun Wang1* Purpose: Results of the association between coffee consumption (CC) and the risk of prostate cancer (PC) are still controversy. Based on published relevant studies, we conducted an up-to-date meta-analysis to investigate this issue. Materials and Methods: The protocol used in this article is in accordance with the PRISMA checklist. Eligible studies were screened and retrieved by using PUBMED and EMBASE as well as manual review of references up to July 2016. We calculated the pooled relative risk (RR) with 95% confidence interval (CI) with random effect models. The dose-response relationship was assessed by generalized least-squares trend estimation analysis. Results: Totally, we included twenty-eight studies (14 case-control and 14 cohort studies) on CC with 42399 PC patients for the final meta-analysis. No significant association of PC was found for high versus non/lowest CC, with RR = 1.07 (95% CI: 0.96-1.18). In subgroup meta-analysis by study design, there were no significant positive associations between CC and PC in case-control studies (RR = 1.19, 95% CI: 1.05-1.35) or in the cohort studies (RR = 0.97, 95% CI: 0.84-1.12). Additionally, RR with different quality of studies were respectively 1.15 (95% CI: 0.99-1.34) and 1.28 (95% CI: 1.03-1.58) for high and low quality in the case-control studies; while were respectively 1.02 (95% CI: 0.88-1.20) and 0.81 (95% CI: 0.57-1.14) in the cohort studies. When analyzed by ge- ographic area, we found no association between CC and PC, with RR = 1.06 (95% CI: 0.86-1.30) for 10 studies from Europe, 1.06 (95% CI: 0.94-1.20) for 13 studies conducted in America; 1.12 (95% CI: 0.70-1.79) for 4 studies from Asia. However, in subgroup analysis by subtype of the disease, there was a significant negative (beneficial) association in the localized PC (RR = 0.90, 95% CI: 0.84-0.97), but not for the advanced PC (RR = 0.90, 95% CI: 0.70-1.16). Additionally, RR = 0.99 (95% CI: 0.98-0.99) for an increment of one cup per day of coffee intake shows significant association with the localized PC. Conclusion: Our results indicate that CC has no harmful effect on PC. On the contrary, it has an effect on reducing the localized PC risk. Further prospective cohort studies of high quality are required to clarify this relationship. Keywords: prostate cancer; coffee consumption; dose-response; stage-specific; meta-analysis. INTRODUCTION Since the introduction of prostate specific antigen testing, the rate of men diagnosed with prostate can- cer (PC) has increased, which makes PC the most fre- quently diagnosed tumor and the second leading cause of death from cancer in men(1). In General, the incidence of PC in Western countries is approximately six-fold higher than that of non-Western countries. Some of this discrepancy may be caused by increased screening, but it has been hypothesized that differences in dietary in- take may also account for it, though much of the re- search has no explicit conclusions(2-4). Coffee is one of the most widely consumed beverages in the world. It is a complex chemical mixture that con- tains many compounds, which have been suggested to have potential genotoxic, mutagenic and anti-mutagen- ic activities in lower organisms(5). Coffee is also a main source of dietary methylxanthines, e.g. caffeine(6). It has been reported that caffeine has obvious effects on a variety of physiologic, cellular and molecular systems, which is fundamental in basic and clinical research(7). Since the 1980s, many epidemiologic studies have es- 1Department of Urology, The First Affiliated Hospital of Nanjing Medical University, Nanjing, China. 2Department of Obstetrics and Gynecology, Nanjing Drum Tower Hospital, Nanjing Medical University, Nanjing, China. *Correspondence: Department of Urology, The First Affiliated Hospital of Nanjing Medical University, 300 Guangzhou Road, Nanjing, China. E-mail: zengjunwang2002@sina.com. Received August 2016 & Accepted July 2017 timated the association between coffee consumption (CC) and PC risk with inconsistent results. So far, me- ta-analyses have been conducted on this issue, yet with opposite conclusions(8-10). However, most of them were methodologically defective—neither of them carried out meta-regressions to examine dose-response analy- sis, nor did they include all the published studies avail- able at the time of their compilations(11). Furthermore, some large prospective cohort studies with high quality have examined the association between CC and PC risk as well as stage-specific (localized or advanced). Addi- tionally, we used multiple subgroup analysis to assess the association between CC and PC, which is different from the previous meta-analysis, and we used general- ized least-squares trend estimation analysis to assess the dose-response relationship, which could complicate the interpretation of the pooled results. Therefore, the aim of the present study is to provide a quantitative assessment on this topic, we systematically performed a meta-analysis by summarizing all avail- able data of both case-control and cohort studies, be- sides, we also conducted the meta-analysis to see the REVIEW Vol 14 No 05 September-October 2017 4079 Coffee Consumption and Risk of Prostate Cancer-Xia et al. Table 1. Characteristics of studies included in the meta-analysis of coffee consumption and prostate cancer risk Authors Study Study Study Cases/ Coffee Adjusted OR/RR NOS Adjustments (publication year) Design Country period noncases consumption (95% CI) score All PCA Local PCA Advanced PCA Talamini et al. Case-control 1992 (32) Hospital based Italy 1986- 1990 271/685 1 Low 1.0 NA NA 6 Age, area of residence, education, and BMI. 2 Inter-mediate 1.12 (0.78-1.62) NA NA 3 High 1.34 (0.93-1.93) NA NA 0 cups/ 1.00 NA 1.00 Slattery et al. Case-control USA 1983- week 1993 (33) Population based 1986 362/685 1-20 cups 0.99 NA 1.39 5 Age /week (0.68-1.47) (0.67-2.87) >20 cups 1.09 NA 1.04 /week (0.75-1.60) (0.47-2.26) Gronberg et al. Case-control 1959- 406/1218 0 cups/day 1.00 NA NA Specific food items, smoking 1996 (34) Population based Sweden 1989 1-2 cups/day 1.77 NA NA habits and alcoholic (0.65-5.09) 7 consumption 3-5 cups/day 1.99 NA NA (0.78-5.46) 6-9 cups/day 1.91 NA NA (0.73-5.30) Key et al. Case-control 1997 (35) Population based England 1989- 328/328 0 cups/day 1.00 NA NA 7 Energy intake 1992 1 cups/day 0.92 (0.60-1.42) NA NA 2 cups/day 1.41 (0.89-2.21) NA NA ≥ 3 cups/day 0.94 (0.59-1.51) NA NA Jain et al. Case-control 1998 (36) Population based Canada 1989- 617/636 0 g/day 1.00 NA NA 6 Age and total energy intake 1993 0-500 g/day 0.84 (0.58-1.22) NA NA > 500 g/day 0.97 (0.65-1.44) NA NA Hsieh et al. Case-control Greece 1994- 1999 (37) Hospital based 1997 320/246 0 cups/day 1.00 NA NA 5 Age, height, BMI, < 1 cups/day 0.38 and years of schooling (0.15-0.99) NA NA 1-2 cups/day 0.72 (0.35-1.45) NA NA 2-3 cups/day 0.57 (0.29-1.12) NA NA > 3 cups/day 1.15 (0.53-2.47) NA NA Villeneuve et al. Case-control 1999 (19) Population based Canada 1994- 1623/1623 0 cups/day 1.0 NA NA 6 Age, province of residence, 1997 < 1 cups/day 0.8 race, years since quitting (0.6-1.1) NA NA smoking, cigarette pack-years, 1-4 cups/day 1.0 alcohol, grains (0.7-1.3) NA NA ≥4 cups/day 1.1 (0.8-1.5) NA NA Sharpe et al. Case-control 2002 (37) Population based Canada 1979- 399/476 0 1.00 NA NA Age, ethnicity, respondent 1985 drinks/day status, family income, BMI, 1-2 1.1 NA NA cumulative cigarette smoking, drinks/day (0.6-1.9) 5 alcohol consumption 3-4 1.1 NA NA drinks/day (0.6-1.9) ≥ 5 0.9 NA NA drinks/day (0.5-1.7) Chen et al. Case-control 2005 (38) Hospital based China 1996- 237/481 No 1.00 NA NA 6 Age and BMI 1998 Yes 1.88 (1.07-3.30) NA NA Gallus et al. Case-control Italy 1991- 219/431 1st Tertile 1.0 NA NA 5 Age, study center, 2007 (39) Hospital based 2002 2nd Tertile 1.3 (0.8-2.1) NA NA education, occupational 3rd Tertile 1.9 (1.2-3.0) NA NA physical activity at 30 –39 years, BMI, family history, and total energy intake Ganesh et al. Case-control 1999- No 1.0 NA NA 5 Age, religion and education 2011 (40) Hospital based India 2001 123/167 Yes 1.3 (0.6-2.7) NA NA Review 4080 Deneo-Pellegrini Case-control et al. Hospital based Uruguay 1996- Tertile I 1.0 NA NA 5 Age, residence, urban/rural 2012 (41) 2004 326/652 Tertile II 1.54 NA NA status, education, family (0.91-2.59) history of prostate cancer Tertile III 1.37 NA NA among first degree relatives, (0.82-2.29) BMI and total energy intake Geybels et al. Case-control USA 2002- ≤ 1/week 1.0 1.0 1.0 age,race, 2013 (22) Population-based 2005 892/863 2-6/week 1.22 1.25 1.01 first-degree family history (0.88-1.69) (0.89-1.76) (0.55-1.83) of prostate cancer, smoking 1 /day 1.13 1.07 1.27 status, and history of prostate (0.84-1.51) (0.97-1.47) (0.77-2.11) 6 cancer screening 2-3 /day 1.16 1.13 1.23 (0.90-1.50) (0.86-1.49) (0.78-1.61) ≥ 4/day 1.16 1.12 1.33 (0.82-1.63) (0.78-1.93) (0.74-2.38) Wilson et al. Case-control Sweden 2001- < 1 1.00 1.00 1.00 2013 (18) Population- based 2002 1499/1112 cup/day age, region, smoking, BMI, 1-<2 0.97 0.88 0.70 education, and intake of cups/day (0.62-1.52) (0.59-1.31) (0.40-1.23) calcium, zinc, and total energy 2-<4 0.98 0.98 0.83 cups/day (0.65-1.49) (0.71-1.35) (0.53-1.29) 7 4-5 1.06 1.01 1.02 cups/day (0.69-1.62) (0.72-1.42) (0.64-1.62) > 5 0.97 0.89 0.73 cups/day (0.60-1.57) (0.59-1.35) (0.41-1.30) Jacobsen et al. Cohort Norway 1967- 1986 (13) 1969 205/13664 ≤2 cups/day 1.00 NA NA Age, residence, 3-4 cups/day 0.83 NA NA cigarette smoking (0.59-1.15) 6 5-6 cups/day 0.78 NA NA (0.53-1.15) ≥7 cups/day 0.74 NA NA (0.47-1.25) Nomura et al. Cohort USA 1965- 108/7355 0 cups/day 1.00 NA NA 8 Age 1986 (14) 1968 1-2 cups/day 1.21 NA NA (1.02-1.43) 3-4 cups/day 1.06 NA NA (0.88-1.26) >5 cups/day 1.43 NA NA (1.20-1.69) Severson et al. Cohort USA 1965- 174/7999 ≤1 time 1.00 NA NA 8 Age 1989 (42) 1978 /week 2-4 times 0.96 NA NA /week (0.39-2.37) ≥5 times/ 0.92 NA NA week (0.59-1.44) Hsing et al. Cohort USA 1966- 1990 (43) 1986 149/17633 ≤3 cups/day 1.00 NA NA 6 Age 3-4 cups/day 0.8 NA NA (0.6-1.2) ≥5 cups/day 1.0 NA NA (0.6-1.6) Marchand et al. Cohort USA 1975- 1994 (44) 1980 198/20316 1 Quantile 1.0 NA NA 8 Age, ethnicity and income 2 Quantile 0.9 (0.6-1.4) NA NA 3 Quantile 1.2 (0.8-1.8) NA NA 4 Quantile 1.1 (0.7-1.7) NA NA Stensvold et al. Cohort Norway 1977- ≤2 cups/day 1.00 NA NA Age, cigarettes per day and 1994 (45) 1982 177/21735 3-4 cups/day 0.3 NA NA county of residence (0.13-1.10) 7 5-6 cups/day 0.6 NA NA (0.30-1.71) ≥7 cups/day 0.4 NA NA (0.23-1.44) Ellison et al. Cohort Canada 1970- 145/3400 0 ml/day 1.00 NA NA 7 Five-year age group and wine 2000 (46) 1993 0-250 ml/day 1.14 NA NA consumption (0.66-1.97) 250-500 1.42 NA NA ml/day (0.80-2.52) 500-700 1.35 NA NA ml/day (0.77-2.61) > 750 1.42 NA NA ml/day (0.77-2.61) Iso et al. Cohort Japan 1988- 2007 (47) 1997 161/43500 ≤1-2/month 1.00 NA NA 8 Age and area of study 1-4/week 0.96 NA NA (0.48-1.92) 1/day 1.19 NA NA (0.71-1.97) ≥2/day 1.13 (0.73-1.75) NA NA Coffee Consumption and Risk of Prostate Cancer-Xia et al. Vol 14 No 05 September-October 2017 4081 Nilsson et al. Cohort Sweden 1992- 653/32425 < 1 occ/day 1.00 NA NA 8 Age, BMI, smoking, education and 2010 (20) 2007 1-3 occ/day 0.92 NA NA recreational physical activity (0.70-1.21) ≥4 occ/day 1.03 NA NA (0.77-1.38) Wilson et al. Cohort USA 1986- 5035/47911 none 1.00 1.00 1.00 8 2011 (15) 2006 < 1 cup/day 0.94 1.01 0.81 (0.85-1.05) (0.88-1.15) (0.64-1.02) 1-3 cups/day 0.94 0.99 0.75 Race, height, BMI, vigorous (0.86-1.04) (0.87-1.12) (0.60-0.93) physical activity, smoking, 4-5 cups/day 0.93 1.02 0.73 diabetes, family history of (0.83-1.04) (0.88-1.18) (0.56-0.95) prostate cancer, multivitamin ≥6 cups/day 0.82 0.93 0.47 use, intake of processed (0.68-0.98) (0.74-1.16) (0.28-0.77) meat, tomato sauce, calcium, alpha linolenic acid, supple mental vitamin E, alcohol intake and history of PSA testing Shafique et al. Cohort England 1970- 318/6017 0 cups/day 1.00 NA NA Age at screening, cholesterol, 2012 (21) 2007 1-2 cups/day 0.84 8 systolic blood pressure, BMI, (0.60-1.21) NA NA alcohol intake, tea ≥3 cups/day 0.74 consumption, smoking status, (0.47-1.16) NA NA social class Discacciati et al. Cohort Sweden 1998- 3801/ 0 NA 1.13 0.96 2013 (23) 2010 44613 (0.93-1.27) (0.68-1.35) < 1cup/day NA 1.00 0.97 tea, alcohol, BMI, personal (0.86-1.16) (0.78-1.21) history of diabetes, family 1-3 cups NA 1.00 1.00 history of PCa, smoking /week status, physical activity education and total energy 4-5 cups/day NA 0.93 0.95 intake (0.83-1.03) (0.79-1.14) ≥ 6 cups/day NA 0.81 0.87 (0.69-0.96) (0.66-1.16) Bosire et al. Cohort USA 1995- 23335/ 0 cup/day 1.00 1.00 1.00 8 2013 (17) 2007 288391 < 1 cup /day 1.03 1.03 1.10 (0.98-1.08) (0.97-1.09) (0.95-1.28) age, race, height, BMI, 1 cup /day 1.00 1.01 0.97 physical activity, smoking, (0.95-1.06) (0.95-1.07) (0.83-1.14) history of diabetes, family 2-3 cups/day 1.00 1.01 0.98 history of prostate cancer , (0.96-1.05) (0.96-1.07) (0.86-1.12) PSA testing, intakes of toma 4-5cups/day 1.00 0.99 1.08 to sauce, alpha-linolenic acid, (0.94-1.06) (0.93-1.06) (0.92-1.27) and total energy intake ≥ 6 0.94 0.92 1.15 cups/day (0.87-1.02) (0.84-1.01) (0.92-1.43) Li et al. Cohort Japan 1995- 318/18853 0 cups/day 1.0 1.0 1.0 7 age, education level, BMI, 2013 (16) 2005 < 1 cup/day 0.81 0.89 1.26 time engaging in sports or (0.61-1.07) (0.48-1.65) (0.73-2.16) exercise, marital status, time 1-2 cups/day 0.73 1.16 0.73 status, family history of (0.53-1.01) (0.61-2.20) (0.38-1.39) cancer, consumption of ≥3 cups/day 0.63 0.54 0.90 spent walking, smoking (0.39-1.00) (0.18-1.66) (0.38-2.12) tea, job status, daily total energy intake, passive smoking, alcohol drinking, daily consumption of miso soup Coffee Consumption and Risk of Prostate Cancer-Xia et al. Abbreviations: OR/RR, odd ratio/rate ratio; C, confidence interval; NOS, Newcastle-Ottawa Scale; BMI, body mass index (kg/m2); occ, occasion; PSA, pros- tate specific antigen; NA, not available, PCA; prostate cancer Review 4082 relationship of CC with stage-specific prostate cancer incidence. MATERIAL AND METHODS Publication search We systematically reviewed the literature by electron- ically searching PUBMED and EMBASE up to July 2016. The search terms included the keywords “cof- fee”, “caffeine”, “diet”, combined with “prostate can- cer”, “prostate carcinoma”, “prostate neoplasm”. All of the references in the relevant articles were screened for any further articles that were not identified in the ini- tial search. Two reviewers (JX and JC) independently searched and extracted the data according to the defined inclusion and exclusion criteria. Inclusion and exclusion criteria Inclusion criteria were as follows:(1) Studies had a case-control or cohort design;(2) The outcome of inter- est was primary prostate cancer;(3) The exposure of in- terest was CC; and(4) Relative risk (RR) and their 95% confidence intervals (CI) could be extracted or calcu- lated from relevant articles. Exclusion criteria were as follows:(1) incomplete data availability;(2) duplicated or updated data;(3) non-inclusion of their own data, such as reviews, comments, editorials, letters and congress. Data extraction Two reviewers (JX and JC) independently extracted and recorded the following information: first author’s surname, year of publication, study design, study coun- try, follow-up period or study period, number of partic- ipants (cases or controls/subjects), the exposure to CC, the odds ratios (OR, from case-control studies) or rate ratios (RR, from cohort studies) estimated with 95% CI for each category of CC of all PC and stage-specific (localized or advanced), and variables adjusted for in the analysis . If 95% CI were not provided, but the num- bers of cases and controls (or person-time) in exposure categories were reported(12-14), these data were used to calculate the standard error of the crude RR, and then approximate CI for the reported adjusted RR. For sev- eral RRs from age-adjusted model to different multivar- iate models(15-23), we chose the RRs from multivariate models with the most complete adjustment for poten- tial confounders. Disagreements were resolved through consensus with a third reviewer (XJ). Quality assessment of included studies Two independent reviewers (JX and JC) systematically performed the methodological quality assessment of se- lected studies according to the Newcastle-Ottawa Scale (NOS)(24). The quality criteria assessed were as follows: the representative and applicability of study groups, comparability of the groups, evaluation of outcomes, and adequacy of follow-up. Since standard criteria have not been stated, we defined scores as ≥6 for case-control Table 2. Summary relative risk estimates and 95% for coffee consumption and prostate cancer risk. Study No. of studies No. of cases Relative risk (95% CI) P value Heterogeneity test Q P I2 (%) Highest vs. lowest All studies 28 42399 1.07 (0.96-1.18) 0.228 54.40 0.001 52.2 Study design Case-control studies 14 7622 1.19(1.05-1.35) 0.005 12.12 0.518 0.0 Cohort studies 14 34777 0.97(0.84-1.12) 0.668 34.10 0.001 64.8 Hospital based case-control studies 6 1496 1.50 (1.21-1.85) 0.000 2.72 0.743 0.0 Population based case-control studies 8 6126 1.06 (0.91-1.23) 0.445 2.54 0.924 0.0 Study geographic area Europe 10 4396 1.06 (0.86-1.30) 0.586 16.60 0.055 45.8 America 13 33363 1.06 (0.94-1.20) 0.361 28.04 0.005 57.2 Asia 4 839 1.12 (0.70-1.79) 0.635 9.03 0.029 66.8 Methodological quality of study Case-control study High quality 8 5873 1.15 (0.99-1.34) 0.060 6.58 0.474 0.0 Low quality 6 1749 1.28 (1.03-1.58) 0.026 4.96 0.421 0.0 Cohort study High quality 8 6647 1.02 (0.88-1.20) 0.781 24.30 0.001 71.2 Low quality 5 994 0.81 (0.57-1.14) 0.221 7.46 0.114 46.4 Stage-specific Localized PCA 6 26064 0.90 (0.84-0.97) 0.006 4.07 0.539 0.0 Case-control studies 2 1745 1.01 (0.77-1.33) 0.922 0.67 0.413 0.0 Cohort studies 4 24319 0.90 (0.83-0.97) 0.004 2.65 0.448 0.0 Advanced PCA 7 5304 0.90 (0.70-1.16) 0.399 12.83 0.046 53.2 Case-control studies 3 584 0.99 (0.69-1.44) 0.976 2.07 0.356 3.2 Cohort studies 4 4720 0.84 (0.58-1.21) 0.340 10.71 0.013 72.0 Increment of 1 cup/day All studies 19 36985 0.99 (0.98-1.00) 0.046 56.21 0.542 0.0 Study design Case-control studies 9 6446 1.01 (0.95-1.06) 0.825 29.97 0.269 3.9 Cohort studies 10 30539 0.99 (0.98-1.00) 0.012 23.12 0.810 0.0 Stage-specific Localized PCA 6 26064 0.99 (0.98-0.99) 0.003 15.4 0.800 0.0 Case-control studies 2 1745 1.01 (0.96-1.06) 0.680 2.76 0.907 0.0 Cohort studies 4 24319 0.99 (0.98-0.99) 0.002 11.86 0.539 0.0 Advanced PCA 7 5304 0.98 (0.94-1.02) 0.410 27.62 0.231 0.2 Case-control studies 3 584 1.02 (0.96-1.08) 0.539 6.74 0.664 0.0 Cohort studies 4 4720 0.97 (0.91-1.02) 0.263 19.20 0.117 2.6 Abbreviations: CI, confidence interval; PCA, prostate cancer Coffee Consumption and Risk of Prostate Cancer-Xia et al. Vol 14 No 05 September-October 2017 4083 studies and ≥8 for cohort studies being of high method- ological quality, otherwise being of low quality(8). Statistical analysis Study-specific log (rate ratio) for cohort studies and log (OR) for case-control studies were combined to compute a pooled RR and its 95% CI for the highest versus non/lowest category of coffee consumption from each study with the Dersimonian and Larid ran- dom effects models(25). The heterogeneity of effect size among studies was tested by Q statistics (P < .10 indi- cated the presence of heterogeneity), and inconsistency was quantified by I2 statistics (I2 > 50% is considered significant)(26,27). In situations with substantial hetero- geneity, the subgroup analysis was used to explore the sources of heterogeneity based on the characteristics of the studies (study design, geographic region, study quality, stage-specific), and a sensitivity analysis was performed to assess the stability of the results. Based on the method developed by Greenland and Longnecker(28,29), we applied generalized least-squares trend estimation analysis to examine dose-response re- lationship between different categories of coffee intake using the random-effects model. For all studies, the median cups of CC for each category were calculated as the average consumption by assigning the midpoint of upper and lower boundaries. If the upper bound was not provided, we assumed that the average consump- tion had the same amplitude of intake as the preceding category. This method requires that the distributions of case patients and control subjects (or person-time) and the risk estimates with their variance estimates for at least three quantitative exposure categories, so studies providing no cutoff or median of coffee intake in each category, or reporting only two categories of exposure, or lacking the number of cases and non-cases in each exposure category were excluded. For studies using units or milliliter other than cups for consumption, we roughly converted them into cups per day as a standard measure (1 time/occasion/drink=1 cup, 125 ml= 1 cup, 250 g =1 cup). Ultimately, we evaluated the possibility of publication bias through a funnel plot and with the Begg’s and Egger’s tests(30,31). A two-tailed P < .05 was considered statistically significant. All statistical analyses were performed with STATA (version 11.0; Stata Crop). RESULTS Study characteristics A total of 114 potentially eligible studies were initial- ly identified, most of which were excluded because the exposure or endpoint was not relevant to our analy- sis. The study identification and selection progression were summarized in Figure 1. Finally, we identified 28 eligible studies in our meta-analysis,(12-23,32-47) in- cluding 14 case-control studies and 14 cohort studies. The former included 7622 cases of PC and 9603 con- trols, while the latter involved 34777 cases of PC and 573812 participants. Particularly, one study only re- ported the stage-specific RR but not all the PC(23). Of these 28 studies, 13 were conducted in America (USA, Canada and Uruguay), 11 in Europe (Sweden, England, Greece, Italy and Norway) and 4 in Asia (China, In- dia and Japan). Among the case-control studies, 6 used hospital-based controls and 8 applied population-based controls. The RRs of most studies were adjusted for age or body mass index (BMI, kg/m2), which are the most likely confounder of relationship between coffee intake and PC. General characteristics in the studies included in this meta-analysis were shown in Table 1. High versus non/lowest coffee consumption Figure 2 and Table 2 present the multivariable-adjusted RRs in each study and the pooled RR of PC for the high- est versus non/lowest categories of coffee intake. The combined summary RR from all the studies was 1.07 (95% CI: 0.96-1.18, P = .228). In the subgroup analysis by study design, the summary RRs from case-control studies and cohort studies were respectively 1.19 (95% CI: 1.05-1.35, P = .005) and 0.97 (95% CI: 0.84-1.12, P = .668). When separating the hospital-based case-con- trol studies from the population-based case-control Figure 1. Flow diagram of the studies identified in the meta-anal- ysis Figure 2. Forest plot of case-control and cohort studies assess- ing the association between high coffee consumption (high ver- sus non/lowest) and prostate cancer risk. Horizontal lines indicate 95% confidence interval (CI); diamonds indicate summary relative risk estimate with its corresponding 95% CI. Coffee Consumption and Risk of Prostate Cancer-Xia et al. Review 4084 studies, we found an apparent difference between them ((hospital based RR: 1.50 (95% CI: 1.21-1.85, P < .001); population-based RR was 1.06 (95% CI: 0.91- 1.23, P = .445)). Based on different geographic regions, the summary RRs were 1.06 (95% CI: 0.86-1.30, P = .586) for the studies conducted in Europe, 1.06 (95% CI: 0.93-1.20, P = .361) for the studies performed in America, and 1.12 (95% CI: 0.70-1.79, P = .635) for the studies carried out in Asia. According to the quality of studies, the pooled RRs for high quality and low quality were respectively 1.15 (95% CI: 0.99-1.34, P = .060), 1.28 (95% CI: 1.03-1.58, P = .026) in the case-control studies; and respectively 1.02 (95% CI: 0.88-1.20, P = .781), 0.81 (95% CI: 0.57-1.14, P = .221) in the cohort studies. Based on the studies, which explored the rela- tionship of CC with stage-specific PC, the meta-anal- ysis showed that the pooled RRs were 0.90 (95% CI: 0.69-1.16, P = .399) in the advanced PC, but 0.90 (95% CI: 0.84-0.97, P = .006) in the localized PC. There was some evidence of heterogeneity among all the studies of CC overall (P = .001, I2 = 52.2%), and the heterogeneity mainly existed in the cohort studies (P = .001, I2 = 64.8%). As the heterogeneity was remarka- ble, we conducted a sensitivity analysis with any single study omitted in all the studies. The results showed that the pooled RRs and 95% CI changed little, which indi- cated that the meta-analysis results were stable (Figure 3). To explore the source of heterogeneity among the cohort studies, we did the subgroup analysis by charac- teristics of studies. When stratified by study geographic area and methodological quality of study, the hetero- geneity of the cohort studies reduced slightly but not significantly (Table 2). Dose-response meta-analysis We incorporated nineteen studies (nine case-con- trol studies(12,18,19,22,33-37) and ten cohort stud- ies(13,15-17,20,21,42,44,46,47) into the dose-response analysis of CC and risk of PC (Table 2), because other remaining studies reported only 2 quantitative exposure catego- ries (38,40), or did not provide cutoff of coffee intake in each category (32,39,41,44), or did not reveal the number of cases and non-cases in each exposure category(41,43). There was marginally statically significant departure from linearity (P = .049). The pooled RR for a one cup per day increment in CC was 0.99 (95% CI: 0.98- 1.00), which was evident for cohort studies (RR = 0.99, 95% CI: 0.98-1.00, P = .012) , but not significant in the case-control studies (RR = 1.01, 95% CI: 0.95-1.06, P = .825). When grouped by stage-specific prostate cancer, the pooled RR for studies conducted in localized PC was 0.98 (95% CI: 0.98-0.99, P = .003), but 0.98 (95% CI: 0.94-1.02, P = .410) in advanced PC. Publication bias No evidence of publication bias was found from either visualization of funnel plot, Begg’s test (P = .632), or Egger’s test (P = .229) (Figure 4). There was no signif- icant indication of publication bias for the sixteen stud- ies, which were included in the dose-response analysis (Begg’s P = .629; Egger’s P = .152). DISCUSSION Based on the published results from 14 case-control and 14 cohort studies, our meta-analysis assessed the poten- tial association between CC and PC. The overall pooled RR of PC for high versus non/lowest coffee consump- tion was 1.07 (95% CI: 0.96-1.18), which indicates that CC is not associated with an increased risk of PC. Strat- ified by study design, the meta-analysis showed CC in- creased the risk of PC in the case-control studies, but did not increase in the cohort studies. The discrepancy of the results between case-control and cohort studies may be explained with potential biases of case-control stud- ies, such as selection bias and recall bias. Additionally, it is worth noting that when subgrouped by the control characteristics or quality of case-control studies, there was not an increased risk of PC in population-based or high quality of case-control studies despite its pres- ence in hospital-based and low quality of case-control studies. Generally, population-based case-control stud- ies are considered more reliable because their subjects are more representative as controls than those of hos- Figure 3. The sensitivity analysis diagram for each study used to assess the relative risk estimates for coffee consumption and pros- tate cancer risk in the cohort studies. Figure 4. Publication bias in all the studies. Both visualization of funnel plot and Begg’s test (P = .632), or Egger’s test (P = .229) test indicated no publication bias in the studies included in the meta-analysis. Coffee Consumption and Risk of Prostate Cancer-Xia et al. Vol 14 No 05 September-October 2017 4085 pital-based case-control studies. As we know, the de- sign and methodology of studies could affect the effi- cacy outcome differently. Hence, these results above suggested that there is no causal relationship between coffee drinking and increased PC. On the contrary, the dose-response relationship analysis showed that there was an inverse dose-response relationship between a one cup per day increment and decreased risk of PC (P = .049), which was more significant in cohort stud- ies (P = .012). More interestingly, in the subgroup of stage-specific PC of both high versus non/lowest CC and dose-response meta-analysis, we found that CC could substantially reduce the localized, rather than ad- vanced prostate cancer incidence. Therefore, based on the results of our analyses, we could conclude that CC could not increase the incidence of PC, but reduce the risk of localized prostate cancer. Compared with the previous meta-analyses, our me- ta-analysis has some advantages. Firstly, it is well known that the inclusiveness of all relevant studies for the meta-analyses is very important. In Park et al. ’ meta-analysis (8), they totally included twelve stud- ies (eight case-control studies and four cohort studies) and found RR of 1.16 (95% CI: 1.01–1.33) for highest versus lowest coffee drinkers. In another one, it only contained five prospective cohort studies and shows an inverse association of PC risk with high coffee in- take (RR 0.79, 95% CI: 0.61–0.98)(9). Recently, there is another meat-analysis demonstrating a borderline significant inverse association between CC and PC risk based on cohort studies. Altogether, these results are inconsistent and confusing. We include all the pub- lished studies available as possible as we could, and the number of total cases included in the meta-analysis was more massive (14 case-control and 14 cohort stud- ies). Secondly, because cutoffs for the highest coffee categories varied from each study, the dose–response relationship analysis is especially important. The dose– response relationship is to describe the change in effect on an organism caused by differing levels of exposure (or doses) to a stressor after a certain exposure time, and it is critical for determining “safe” and “hazardous” levels and dosages for drugs, potential pollutants, and other substances to which humans or other organisms are exposed(48). We did a dose-response meta-analysis, which was not carried out specially in either of the pre- vious meta-analyses. Thirdly, as we know, publication bias is much of concern in a meta-analysis, and there was little evidence of publication bias in our meta-anal- ysis. Finally and the most importantly, to the best of our knowledge, in the meta-analysis, we first evaluate subgroups based on the stage of PC, finding that there is an inverse association with the incidence of localized rather than advanced PC. Coffee is produced by infusing ground, roasted coffee beans, with the most common forms being coffea ara- bica and coffea canephoria var. robusta(49). Coffee con- tains more than a thousand different chemicals. While it has caffeine and methylglyoxal with potentially carcinogenic effects, some other chemicals have been suggested to have potentially chemo-preventive effects, such as chlorogenic, caffeic acids, diterpenes cafestol and kahweol(7,50-52). Our analyses have found that cof- fee drinking could reduce the risk of localized but not advanced PC. This is an investing finding. It has been reported that other environmental agents, like chemi- cal, physical or microbial agents, could enhance or sup- press coffee on the carcinogenic effect, depending on the carcinogen it is used with, the type of host cell, and the stage of cell cycle in which it is introduced(53). It is hypothesized that coffee drinking may be associated with increased levels of sex hormone-binding globu- lin (SHBG) and total testosterone levels, which might play a role in PC(54). However, a recent randomized tri- al showed that consumption of caffeinated coffee had no evident effect on SHBG levels, but significantly in- creased total testosterone and decreased both total and free estradiol in men(55). At the same time, CC is also associated with reductions in the levels of inflamma- tion-related molecule, which have an important role in prostatic carcinogenesis(56). Furthermore, an animal study showed that caffeine treatment increased the per- centage of mitotic tumor cells undergoing lethal mi- tosis, which indicated oral administration of caffeine might be an effective strategy for the prevention of PC progression(57). Despite these advantages, there are still some limita- tions. Firstly, heterogeneity among studies may have been involved because of methodological differences among studies, including different methods of coffee preparation, misclassification of CC, differences in serving size and brew strength. Furthermore, the indi- vidual RR estimate included in our meta-analysis was adjusted for different covariates in the different studies. Nevertheless, the results did not change substantially after the sensitivity analysis. Secondly, unfortunately, because of the small number of studies investigating the relationship between CC and subtypes of PC, our meta-analysis could only evaluate subgroups based on tumor stage, but not on Gleason grade or prostate can- cer-specific mortality. Lastly, most of the studies in this meta-analysis were conducted in Europe, the United States, Canada and Japan; thus the data should be ex- trapolated to other populations with caution. CONCLUSIONS In summary, although data from low quality case-con- trol studies suggest that coffee is a risk factor for PC, there is no association between CC and increased PC based on the results of high quality of case-control studies and cohort studies and dose-response analysis. On the contrary, according to the stage-specific pros- tate cancer, subgroups analysis showed that CC could be a protective exposure that reduces the localized PC risk. However, prospective studies, focusing on more detailed results, including subtypes of coffee, taking a broad range of confounders into account, are required to clarify this relationship. ACKNOWLEDGMENTS This work was supported in part by the Project of the National Natural Science Foundation of China (Grant No. 81501245). CONFLICT OF INTEREST The authors have no conflict of interest. REFERENCES 1. Siegel R, Ward E, Brawley O, Jemal A. Cancer statistics, 2011: the impact of eliminating socioeconomic and racial disparities on Coffee Consumption and Risk of Prostate Cancer-Xia et al. Review 4086 premature cancer deaths. CA Cancer J Clin 2011; 61: 212-36. 2. Stacewicz-Sapuntzakis M, Borthakur G, Burns JL, Bowen PE. Correlations of dietary patterns with prostate health. Mol Nutr Food Res 2008; 52:114-30. 3. Muller DC, Severi G, Baglietto L, et al. Dietary patterns and prostate cancer risk. Cancer Epidemiol Biomarkers Prev 2009; 18:3126-9. 4. Niclis C, Díaz Mdel P, Eynard AR, Román MD, La Vecchia C. Dietary habits and prostate cancer prevention: a review of observational studies by focusing on South America. Nutr Cancer 2012; 64:23-33. 5. Nehlig A, Debry G. Potential genotoxic, mutagenic and antimutagenic effects of coffee: a review. Mutat Res 1994; 317:145-62. 6. Tarka SM Jr. The toxicology of cocoa and methylxanthines: a review of the literature. Crit Rev Toxicol 1982; 9:275-312. 7. Porta M, Vioque J, Ayude D, et al. Coffee drinking: the rationale for treating it as a potential effect modifier of carcinogenic exposures. Eur J Epidemiol 2003; 18: 289-98. 8. Park CH, Myung SK, Kim TY, et al. Coffee consumption and risk of prostate cancer: a meta-analysis of epidemiological studies. BJU Int 2010; 106:762-9. 9. Yu X, Bao Z, Zou J, Dong J. Coffee consumption and risk of cancers: a meta- analysis of cohort studies. BMC Cancer 2011; 11:96. 10. Zhong S, Chen W, Yu X, Chen Z, Hu Q, Zhao J. Coffee consumption and risk of prostate cancer: an up-to-date meta-analysis. Eur J Clin Nutr 2014; 68:330-7 11. Arab L. Should men drink more coffee to delay progression of prostate cancer? Nutr Cancer 2011; 63:1161-2. 12. Hsieh CC, Thanos A, Mitropoulos D, Deliveliotis C, Mantzoros CS, Trichopoulos D. Risk factors for prostate cancer: a case- control study in Greece. Int J Cancer 1999: 80: 699-703. 13. Jacobsen BK, Bjelke E, Kvåle G, Heuch I. Coffee drinking, mortality, and cancer incidence: results from a Norwegian prospective study. J Natl Cancer Inst 1986; 76: 823-31. 14. Nomura A, Heilbrun LK, Stemmermann GN. Prospective study of coffee consumption and the risk of cancer. J Natl Cancer Inst 1986; 76: 587-90. 15. Wilson KM, Kasperzyk JL, Rider JR, et al. Coffee consumption and prostate cancer risk and progression in the Health Professionals Follow-up Study. J Natl Cancer Inst 2011; 103: 876-84. 16. Li Q, Kakizaki M, Sugawara Y, et al. Coffee consumption and the risk of prostate cancer: the Ohsaki Cohort Study. Br J Cancer 2013; 108: 2381-9. 17. Bosire C, Stampfer MJ, Subar AF, Wilson KM, Park Y, Sinha R. Coffee consumption and the risk of overall and fatal prostate cancer in the NIH-AARP Diet and Health Study. Cancer Causes Control 2013; 24:1527-34. 18. Wilson KM, Bälter K, Möller E, et al. Coffee and risk of prostate cancer incidence and mortality in the Cancer of the Prostate in Sweden Study. Cancer Causes Control 2013; 24:1575-81. 19. Villeneuve PJ, Johnson KC, Kreiger N, Mao Y. Risk factors for prostate cancer: results from the Canadian National Enhanced Cancer Surveillance System. The Canadian Cancer Registries Epidemiology Research Group. Cancer Causes Control 1999; 10: 355-67. 20. Nilsson LM, Johansson I, Lenner P, Lindahl B, Van Guelpen B. Consumption of filtered and boiled coffee and the risk of incident cancer: a prospective cohort study. Cancer Causes Control 2010; 21: 1533-44. 21. Shafique K, McLoone P, Qureshi K, Leung H, Hart C, Morrison DS. Coffee consumption and prostate cancer risk: further evidence for inverse relationship. Nutr J 2012; 11: 42. 22. Geybels MS, Neuhouser ML, Stanford JL. Associations of tea and coffee consumption with prostate cancer risk. Cancer Causes Control 2013; 24: 941-8. 23. Discacciati A, Orsini N, Andersson SO, et al. Coffee consumption and risk of localized, advanced and fatal prostate cancer: a population-based prospective study. Ann Onco 2013; 24:1912-8. 24. Stang A .Critical evaluation of the Newcastle- Ottawa scale for the assessment of the quality of nonrandomized studies in meta-analyses. Eur J Epidemiol 2010; 25: 603-5. 25. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986; 7: 177-88. 26. Cochran WG. The combination of estimates from different experiments. Biometrics 1954; 10:101-29. 27. Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta- analyses. BMJ 2003; 327: 557-60. 28. Greenland S, Longnecker MP. Methods for trend estimation from summarized dose- response data, with applications to meta- analysis. Am J Epidemiol 1992; 135: 1301-9. 29. Orsini N, Li R, Wolk A, Khudyakov P, Spiegelman D. Meta-analysis for linear and nonlinear dose-response relations: examples, an evaluation of approximations, and software. Am J Epidemiol 2012; 175: 66-73. 30. Begg CB, Mazumd ar M. Operating characteristics of a rank correlation test for Coffee Consumption and Risk of Prostate Cancer-Xia et al. Vol 14 No 05 September-October 2017 4087 publication bias. Biometrics 1994; 50: 1088– 101. 31. Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test. BMJ 1997; 315: 629– 63. 32. Talamini R, Franceschi S, La Vecchia C, Serraino D, Barra S, Negri E. Diet and prostatic cancer: a case-control study in northern Italy. Nutr Cancer 1992; 18: 277-86. 33. Slattery ML, West DW. Smoking, alcohol, coffee, tea, caffeine, and theobromine: risk of prostate cancer in Utah (United States). Cancer Causes Control 1993; 4: 559-63. 34. Grönberg H, Damber L, Damber JE. Total food consumption and body mass index in relation to prostate cancer risk: a case-control study in Sweden with prospectively collected exposure data. J Urol 1996; 155: 969-74. 35. Key TJ, Silcocks PB, Davey GK, Appleby PN, Bishop DT. A case-control study of diet and prostate cancer. Br J Cancer 1997; 76: 678-87. 36. Jain MG, Hislop GT, Howe GR, Burch JD, Ghadirian P. Alcohol and other beverage use and prostate cancer risk among Canadian men. Int J Cancer 1998; 78: 707-11. 37. Sharpe CR, Siemiatycki J. Consumption of non-alcoholic beverages and prostate cancer risk. Eur J Cancer Prev 2002; 11: 497-501. 38. Chen YC, Chiang CI, Lin RS, Pu YS, Lai MK, Sung FC. Diet, vegetarian food and prostate carcinoma among men in Taiwan. Br J Cancer 2005; 93: 1057-61. 39. Gallus S, Foschi R, Talamini R, et al. Risk factors for prostate cancer in men aged less than 60 years: a case-control study from Italy. Urology 2007; 70: 1121-6. 40. Ganesh B, Saoba SL, Sarade MN, Pinjari SV. Risk factors for prostate cancer: An hospital- based case-control study from Mumbai, India. Indian J Urol 2011; 27: 345-50. 41. Deneo-Pellegrini H, Ronco AL, De Stefani E, et al. Food groups and risk of prostate cancer: a case-control study in Uruguay. Cancer Causes Control 2012; 23: 1031-8. 42. Severson RK, Nomura AM, Grove JS, Stemmermann GN. A prospective study of demographics, diet, and prostate cancer among men of Japanese ancestry in Hawaii. Cancer Res 1989; 49: 1857-60. 43. Hsing AW, McLaughlin JK, Schuman LM, et al. Diet, tobacco use, and fatal prostate cancer: results from the Lutheran Brotherhood Cohort Study. Cancer Res 1990; 50: 6836-40. 44. Le Marchand L, Kolonel LN, Wilkens LR, Myers BC, Hirohata T. Animal fat consumption and prostate cancer: a prospective study in Hawaii. Epidemiology 1994; 5: 276-82. 45. Stensvold I, Jacobsen BK. Coffee and cancer: a prospective study of 43,000 Norwegian men and women. Cancer Causes Control 1994; 5: 401-8. 46. Ellison LF. Tea and other beverage consumption and prostate cancer risk: a Canadian retrospective cohort study. Eur J Cancer Prev 2000; 9:125-30. 47. Iso H, Kubota Y, Japan Collaborative Cohort Study for Evaluation of Cancer. Nutrition and disease in the Japan Collaborative Cohort Study for Evaluation of Cancer (JACC). Asian Pac J Cancer Prev 2007; 8: 35-80. 48. Crump KS, Hoel DG, Langley CH, Peto R. Fundamental carcinogenic processes and their implications for low dose risk assessment. Cancer Res 1976; 36: 2973-9. 49. Grigg D. The worlds of tea and coffee: Patterns of consumption. Geojournal 2002; 57: 283-94. 50. Hashimoto T, He Z, Ma WY, et al. Caffeine inhibits cell proliferation by G0/G1 phase arrest in JB6 cells. Cancer Res 2004; 64: 3344- 9. 51. Lee WJ, Zhu BT. Inhibition of DNA methylation by caffeic acid and chlorogenic acid, two common catechol-containing coffee polyphenols. Carcinogenesis 2006; 27: 269- 77. 52. Cavin C, Holzhaeuser D, Scharf G, Constable A, Huber WW, Schilter B. Cafestol and kahweol, two coffee specific diterpenes with anticarcinogenic activity. Food Chem Toxicol 2002; 40: 1155-63. 53. Zhou Y, Tian C, Jia C. A dose-response meta- analysis of coffee consumption and bladder cancer. Prev Med 2002; 55:14-22. 54. Platz EA, Giovannucci E. The epidemiology of sex steroid hormones and their signaling and metabolic pathways in the etiology of prostate cancer. J Steroid Biochem Mol Biol 2004; 92: 237-253. 55. Wedick NM, Mantzoros CS, Ding EL, et al. The effects of caffeinated and decaffeinated coffee on sex hormone-binding globulin and endogenous sex hormone levels: a randomized controlled trial. Nutr J 2012; 11:86. 56. Sfanos KS, De Marzo AM. Prostate cancer and inflammation: the evidence. Histopatholog 2012; 60:199-215. 57. Crump KS, Hoel DG, Langley CH, Peto R. Fundamental carcinogenic processes and their implications for low dose risk assessment. Cancer Res 1976; 36: 2973-9. Coffee Consumption and Risk of Prostate Cancer-Xia et al. Review 4088