Is acne in adolescence associated with prostate cancer risk? Evidence from a meta-analysis

Introduction Previous studies regarding the relationship between acne and prostate cancer risk have reported inconsistent results. We performed the present meta-analysis of observational studies to summarize the evidence on this association. Methods A comprehensive literature search up to March 2018 was performed in PubMed, Scopus, Web of Science, and Chinese National Knowledge Infrastructure (CNKI) databases. Summary odds ratios (ORs) with 95% confidence intervals (CIs) were estimated with a random effects model. The Q statistic and the I2 index were used to evaluate the heterogeneity across the studies. Results Eight studies were ultimately included in this meta-analysis. In the overall analysis, no significant association was found between acne and prostate cancer risk (OR = 1.08, 95% CI 0.93–1.25). A significant heterogeneity was observed across studies (P = 0.006, I2 = 64.5%). In the subgroup analysis by study design, a significant association was observed in the cohort studies (OR = 1.51, 95% CI 1.19–1.93) but not in the case-control studies (OR = 0.98, 95% CI 0.86–1.12). Conclusions In summary, this meta-analysis did not find an association between acne in adolescence and prostate cancer risk. However, because there was some heterogeneity in the overall analysis and a significant association was observed in the meta-analysis of the cohort studies, further well-designed large prospective studies are warranted to confirm our results.


Introduction
Prostate cancer is the second most commonly diagnosed cancer in men worldwide, with 1.1 million new cases estimated to have occurred in 2012 [1]. Advanced age, African descent and a positive family history of prostate cancer are the only established risk factors for prostate cancer [2,3]. Therefore, emerging studies are currently evaluating new markers that may help identify men with a high risk of prostate cancer, in whom the benefits of screening may outweigh the potential side effects [4,5].
Recently, emerging studies have examined the association between acne and the risk of prostate cancer given that acne vulgaris is a proxy for androgen status [6]. In addition, Propionibacterium acnes (P. acnes), a skin bacterium closely associated with acne, is reported to be associated with prostatic inflammation and carcinogenesis [7][8][9]. Several observational studies have found a possible link between a history of acne and prostate cancer risk but with inconsistent results. A recent large prospective cohort study by Ugge et al. [10] indicated that acne was significantly associated with an increased risk for prostate cancer. The Glasgow Alumni Cohort Study suggested that individuals with acne in adolescence had a higher risk of adult prostate cancer mortality [11]. By contrast, the studies by Lightfoot et al. [12] and Cremers et al. [13] failed to establish a relationship between acne and prostate cancer risk. Given these controversial findings, we performed a systematic review and meta-analysis to summarize all relevant evidence. To the best of our knowledge, this is the first meta-analysis to assess the potential association between acne and prostate cancer risk.

Literature search
A comprehensive literature search was performed in PubMed, Scopus, Web of Science, and Chinese National Knowledge Infrastructure (CNKI) databases up to March 2018 with the following search keywords: (acne or propionibacterium) and (prostatic neoplasms or prostatic cancer or prostate neoplasms or prostate cancer). We first evaluated the potentially relevant articles by checking their titles and abstracts. Any studies possibly matching the eligibility criteria were further evaluated by reading the full texts. We also examined the cited references lists from retrieved articles and reviews to identify any additional relevant studies. This systematic review and meta-analysis was planned, performed, and reported in adherence to the reporting and publication requirements of the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) statement [14].

Study selection
Studies included in this meta-analysis met all of the following criteria: (i) the exposure of interest was acne; (ii) the outcome of interest was prostate cancer; (iii) the study had a cohort, nested case-control, or case-control design; and (iv) the study provided risk estimates with their corresponding 95% confidence intervals (CIs) or enough data to calculate them. There was no restriction on language or publication date.

Study quality assessment
The method quality of each study was assessed by two independent reviewers (X.Z. and Y.L.) with the Newcastle-Ottawa Scale (NOS, http://www.ohri.ca/programs/clinical_epidemiology/ oxford.asp). The NOS is an eight-item instrument that primarily assesses the following sections of each study: source of study population, study comparability, follow-up, and outcome of interest. We assigned NOS scores of <7 and �7 for low-and high-quality studies, respectively.

Information extraction
Data were extracted by two independent reviewers (X.Z. and Y.L.) with a predefined information collection form. Any discrepancies were resolved by group consensus. We recorded the following data for each study: first author's surname, the country in which the study was performed, publication year, study design, sample size, methods of exposure assessment, fully adjusted risk estimates with their 95% CIs, and adjusted confounders in the data analysis.

Statistical methods
As prostate cancer is a rare disease, the relative risk (RR) and the hazard ratio (HR) were assumed to be approximately the same as the odds ratio (OR). The ORs and their 95% CIs were used as the study outcome to assess the strength of the relationship between acne and prostate cancer risk. Summary ORs with 95% CIs were estimated with a random effects model using the method of DerSimonian-Laird [15], which incorporates both within-and betweenstudy variability. Subgroup analyses were performed according to study design, geographic region, study quality, number of cases, and publication year.
Heterogeneity across studies was evaluated using the Q statistic with a significance level set to P value < 0.10 [16]. I 2 score (95% CI) was further used to assess the degree of heterogeneity (I 2 <25%, mild heterogeneity; I 2 = 25-50%, moderate heterogeneity; I 2 >50%, large or extreme heterogeneity) [16]. Finally, a meta-regression and Galbraith plot analysis [17] were used to explore the possible sources of heterogeneity if any.
Sensitivity analysis was performed by repeating the pooled analysis after omission of each study in turn. A cumulative meta-analysis was performed by sorting each study by publication year. All of the statistical analyses were performed with Stata 10.0 (StataCorp, College Station, TX). A two-sided P value < 0.05 was considered statistically significant.

Publication search and study characteristics
The detailed process of our literature search and selection is shown in Fig 1. A total of 372 studies were identified primarily from electronic databases. After excluding 124 reviews/editorials and 235 studies that were obviously not relevant (e.g., animal or cell line studies and studies concerning other types of cancer), 13 studies were evaluated by reading the full texts. Four studies that did not report prostate cancer risk in relation to acne and one study that reported the exposure as concentration of P. acnes antibody were also excluded. Eight studies [10][11][12][13][18][19][20][21] were ultimately included in this meta-analysis to evaluate the association between acne and prostate cancer risk. These studies were published between 2003 and 2018 with a total of 10,145 cases. There were three cohort studies [10,11,20] and five case-control studies [12,13,18,19,21]. These studies were performed in the following geographic region: North America (n = 2), Europe (n = 4), and Oceania (n = 2). Information on acne was obtained by interview or selfadministered questionnaire. The scores for study quality assessed by the NOS ranged from 5 to 8. The main characteristics of each study included in our meta-analysis are shown in Table 1.

Overall analysis and subgroup analysis
The multivariable-adjusted ORs for each study and combination of all studies for acne versus no acne groups are shown in Fig 2. No significant association was found between acne and prostate cancer risk (OR = 1.08, 95% CI 0.93-1.25). Then, we performed subgroup analyses by study quality, study design, geographical region, number of cases, and publication year. In the subgroup analysis by study design, a significant association was observed in the cohort studies (OR = 1.51, 95% CI 1.19-1.93) but not in the case-control studies (OR = 0.98, 95% CI 0.86-1.12). No significant association was observed in any other subgroups ( Table 2).

Evaluation of heterogeneity
In this meta-analysis, the Q statistic and the I 2 index were used to evaluate the heterogeneity across the studies. Significant heterogeneity was observed among studies in the overall analysis (Fig 2, P = 0.006, I 2 = 64.5%). Using Galbraith plot analysis, the study by Ugge et al. [10] was identified as the major source of heterogeneity (S1 Fig). Furthermore, we assessed heterogeneity across studies with a meta-regression method and study design was identified as a potential source of heterogeneity in the overall meta-analysis ( Table 2, P for interaction = 0.030).

Sensitivity analysis
A sensitivity analysis was performed to assess the influence of each study on the overall estimate. Meta-analysis estimates were computed by omitting one study at a time. The overall estimate (95% CIs) ranged from 1.03 (95% CI 0.89-1.20) to 1.12 (95% CI 0.96-1.31) after the omission of the study by Ugge et al.(2018) [10] and the study by Giles et al.(2003) [18], respectively (Fig 3).

Cumulative meta-analysis
A cumulative meta-analysis was performed on the included studies sorted by publication year. As shown in S2 Fig, the association between acne and prostate cancer risk is presented in chronological order. The 95% CIs became narrower with cumulative sample size, indicating that the precision of risk estimates was progressively boosted by the continual addition of studies.

Discussion
The present study summarized the current evidence of the relationship between acne in adolescence and prostate cancer risk using a meta-analysis of observational studies, including five case-control studies and three cohort studies. To the best of our knowledge, this is the first meta-analysis evaluating the association between acne and prostate cancer risk. The results indicated that acne was not significantly associated with the risk of prostate cancer (OR = 1.08, 95% CI 0.93-1.25).
When we assessed heterogeneity across the studies with a meta-regression method, we found that study design was a potential source of heterogeneity in the overall meta-analysis. The summary OR (95% CI) for the cohort studies was 1.51 (1.19-1.93) without obvious heterogeneity across the studies. However, no significant association was observed in the casecontrol studies, and this meta-analysis of 8 studies was outweighed by the case-control studies, which would tend to make the summary risk estimate skew toward the null. However, as cohort studies are generally superior to case-control studies, further large prospective cohort studies are warranted because of the positive findings in the subgroup of cohort studies. This study had some strengths. A total of 10,145 prostate cancer cases were included in this meta-analysis; the large sample size provided reliable summary risk estimates. We extracted data from the most fully adjusted models in each study, which might minimize potential confounding factors. Various subgroup analyses, heterogeneity analyses, and sensitivity analyses were performed to evaluate the robustness of our findings.
The exact pathogenic mechanism of prostate cancer is still largely unknown. It has been hypothesized that chronic inflammation may contribute to prostate carcinogenesis [22]. P. acnes isolated from radical prostatectomy specimens have been reported to be positively related to the onset and extent of both acute and chronic prostate inflammation [23]. P. acnes were more common in men with prostate carcinoma than in the controls, and the cells infected with P. acnes had increased proliferation in vivo [8,24,25]. Therefore, P. acnes infection may be a contributing factor to the development of prostate cancer [26,27]. In theory, incorporating some simple factors, such as age, family history, and new reported markers (e.g., HPV-16 infection [28] and hypertension [29]), into prostate cancer risk calculators may help identify men with a relatively high risk of prostate cancer. Thus, if there is a significant association between acne and prostate cancer, we may be able to reduce the risk of prostate cancer by clearing P. acnes. However, in this study, we failed to detect a potential link between a history of acne and prostate cancer risk. This study had several limitations that should be acknowledged. First, substantial heterogeneity was detected across individual studies (P = 0.006, I 2 = 64.5%), which might distort the pooled risk estimates. This is a large meta-analysis, and there is a direct link between metaanalysis size and detected heterogeneity. Thus, obvious heterogeneity is the norm, and it is great if heterogeneity has been identified and can be incorporated in the model [30]. Second,  because only 8 studies were included in our meta-analysis, we were not able to perform publication bias tests or display a funnel plot. Third, the majority of the included studies adopted a case-control design, which may introduce selection bias and recall bias. Fourth, given the small number of studies included, the meta-regression and Cochran's Q test may have limited power [31]. Finally, there were differences in how acne was defined. Some studies used severe acne as exposure, while other studies used any acne as exposure, which may contribute to the heterogeneity that was observed.

Conclusions
In summary, this meta-analysis did not find an association between acne in adolescence and prostate cancer risk. Because of the obvious heterogeneity across the studies, especially heterogeneity by study design, and the significant association observed in a subgroup of cohort studies, further well-designed large prospective studies are warranted to confirm our preliminary findings.