Cancer Epidemiology 38 (2014) 329–338
Contents lists available at ScienceDirect
Cancer Epidemiology The International Journal of Cancer Epidemiology, Detection, and Prevention journal homepage: www.cancerepidemiology.net
Sexually transmitted infections and prostate cancer risk: A systematic review and meta-analysis Saverio Caini a,*, Sara Gandini b, Maria Dudas c, Viviane Bremer d, Ettore Severi e, Alin Gherasim f a
Unit of Molecular and Nutritional Epidemiology, Institute for Cancer Research and Prevention (ISPO), Italy Division of Epidemiology and Biostatistics, European Institute of Oncology, Milan, Italy Department of Infectious Disease Epidemiology, National Center for Epidemiology, Budapest, Hungary d Division for HIV/AIDS, STI and Blood-Borne Infections, Department for Infectious Disease Epidemiology, Robert Koch Institute, Berlin, Germany e South East Regional Epidemiology Unit, Health Protection Agency, London, UK f Swedish Institute for Communicable Disease Control, Solna, Sweden b c
A R T I C L E I N F O
A B S T R A C T
Article history: Received 3 April 2014 Received in revised form 4 June 2014 Accepted 10 June 2014 Available online 28 June 2014
Prostate cancer (PC) is the second most incident cancer and the sixth cause of death by cancer in men worldwide. Despite extensive research efforts, no modifiable risk factors have been consistently identified for PC risk. A number of studies have focused on possible relationships between sexually transmitted infections (STIs) and PC. We performed a meta-analysis to explore the association between infection caused by Neisseria gonorrheae, Treponema pallidum, Chlamydia trachomatis, Trichomonas vaginalis, Ureaplasma urealyticum, Mycoplasma hominis, Herpes Simplex Virus types 1 and 2, Human Herpes Virus 8 and Cytomegalovirus, and PC. We conducted a comprehensive, systematic bibliographic search of medical literature to identify relevant studies. We calculated summary relative risk (SRR) and 95% confidence intervals (CI) for the association between each STI and PC through random effect models. Subgroup, metaregression and sensitivity analyses were carried out to detect between-study heterogeneity and bias. We included 47 studies published between 1971 and 2011. Men who reported having ever had any STI in lifetime had an increased PC (SRR 1.49, 95% CI 1.19–1.92). We found a significantly increased PC risk in men having had gonorrhoea (SRR 1.20, 95% CI 1.05–1.37). No other single STI was significantly associated with PC. Due to high incidence of both STIs and PC worldwide, prevention of STIs may help preventing a considerable number of PC cases. ß 2014 Elsevier Ltd. All rights reserved.
Keywords: Sexually transmitted diseases Gonorrhoea Prostate cancer Review Meta-analysis
1. Introduction Prostate cancer (PC) is the second most incident cancer in men worldwide. In 2008 there were 903,500 estimated new cases and 258,400 estimated deaths worldwide, resulting as the sixth cause of death by cancer in men [1]. Despite extensive research effort, no modifiable risk factors have been consistently identified for PC, except perhaps smoking [2], obesity [3] and a sedentary lifestyle [4]. Sexually transmitted infections (STIs) represent a major public health problem worldwide. Human Papilloma Virus (HPV) and Herpes Simplex Virus (HSV) are common STIs worldwide [5,6], the
* Corresponding author at: Institute for Cancer Research and Prevention (ISPO), Unit of Molecular and Nutritional Epidemiology, Via delle Oblate 2, 50141 Florence, Italy. Tel.: +39 3485828406. E-mail addresses: [email protected], [email protected] (S. Caini). http://dx.doi.org/10.1016/j.canep.2014.06.002 1877-7821/ß 2014 Elsevier Ltd. All rights reserved.
former being also involved in the aetiology of cancer of cervix uteri and other anatomical sites [7,8]. Bacterial STIs are also emerging or re-emerging worldwide [9]. Cases of Chlamydia trachomatis have been increasingly reported during past 20 years [10–13]. Although this trend is partly attributable to extended screening efforts and more sensitive tests, several cases still go undiagnosed due to underreporting or asymptomatic disease, especially among women. Gonorrhoea is especially increasing in those countries with previously low incidence rates [10–15], despite being still at near-historic lows; more worryingly, resistance to 3rd generation cephalosporins is being increasingly reported [16]. Syphilis incidence is overall stable in developed countries, but on rise among middle-aged men and men who have sex with men [10,11,13,15]. Due to high proportions of asymptomatic cases, little is known on epidemiology of other STIs, such as infections with Trichomonas vaginalis or Human Herpes Virus 8 (HHV-8, also known as Kaposi’s sarcoma-associated herpes virus).
330
S. Caini et al. / Cancer Epidemiology 38 (2014) 329–338
The first claims of an aetiological role of STIs in the development of PC date back to the 1950s [17], and several mechanisms were subsequently proposed to explain this association. For gonorrhoea and other bacterial infections, the process was investigated that leads to PC through the phases of prostate inflammation and prostate atrophy, whereas for viral infections the emphasis was placed on the transforming properties of viruses, in particular herpes viruses [18]. It has also been hypothesized that multiple episodes of STIs and infections of longer duration (as in the case of infections not or inefficiently treated) may represent a greater risk for PC development, due to a higher cumulative risk of prostate involvement [18]. A recently published meta-analysis showed a weak association between HPV-16 and PC and no association for HPV-18 [19]. For other STIs, the most recent meta-analysis, published in 2005, found significantly increased PC risk in men with a history of gonorrhoea (1.35, 95% CI 1.05–1.83) and any STI (1.48, 95% CI 1.26–1.73), but not syphilis (1.42, 95% CI 0.76–2.64) [20]. Since 2001, year of publication of the most recent paper included in the latter metaanalysis, several epidemiological studies have been conducted to explore the association between STIs and PC risk. We performed a meta-analysis to explore the association between gonorrhoea, syphilis, and other STIs (other than HPV) and PC risk. 2. Materials and methods
When two or more papers published results originating from the same study sample, as it may happen with cohorts analyzed at different points in time, we only considered the most recently published results. 2.4. Data extraction Two authors (SC and ES) independently extracted the following information from each paper included in the meta-analysis: year of publication, country, study design, source of cohort members (for cohort, case-cohort and nested case-control studies) or of cases and controls (for case-control studies), type of matching (if any), mean/median age, ethnicity, inclusion of controls with familial history of PC, inclusion of controls suffering from cancers other than PC and/or benign prostatic hyperplasia (BPH), blind assessment of exposure status, type of diagnosis for exposure, and adjustments that were made for statistical analyses. Any discrepancies in study selection and data extraction (emerged by comparing the two database that were prepared) were handled by discussion between the authors; if no agreement could be reached, another co-author (SG) would decide. For case-control studies with more than one control group we considered, for each association of interest, the odds ratio originating from the control group that was judged, on an a priori basis, to yield least biased estimates.
2.1. Definition of outcome and exposures 2.5. Statistical analysis We considered as outcome histologically confirmed prostate cancer, diagnosed through clinical examination or after consulting cancer registries. We defined exposure as either self-reported history, clinical or serological diagnosis of infection caused by any of the following agents: Neisseria gonorrheae, Treponema pallidum, C. trachomatis, T. vaginalis, Ureaplasma urealyticum, Mycoplasma hominis, HSV types 1 and 2, HHV-8 and Cytomegalovirus (CMV). We also considered for meta-analysis the association between reported history of ‘‘any STI’’ (either specified or not) and PC risk. 2.2. Data sources and search strategy We identified eligible studies to be included in the metaanalysis by reviewing published reports listed in the following databases: PUBMED, Ovid Medline, EMBASE, and ISI Web of Science – Science Citation Index Expanded. To search for published papers, we used all possible combinations of MESH terms defining the outcome (‘‘prostatic cancer’’ OR ‘‘prostate cancer’’) AND each of the aforementioned infective agents (see above) and STI caused by them (i.e., ‘‘gonorrhoea’’, ‘‘syphilis’’, ‘‘trachoma’’, ‘‘trichomoniasis’’, etc.). Additional studies were collected by searching through references lists of retrieved articles and previously published meta-analyses and reviews. A full copy was obtained of papers that were considered of interest after reading the abstract. 2.3. Criteria for including studies Retrieved papers published up to August 31, 2013 were included in the meta-analysis if they met the following inclusion criteria: human observational studies; a cohort, case-control (including matched and nested case-control) or case-cohort design; reporting or providing sufficient information for estimating a measure of relative risk (RR) (incidence rate ratio, risk ratio, odds ratio, hazard ratio, standardized incidence ratio) with 95% confidence intervals (95% CI) or another measure of statistical uncertainty (standard errors, variance, or exact p-value of the significance of the estimates). No language or time restrictions were applied.
For any different pair of outcome and exposure, the most adjusted measure of association and the corresponding confidence intervals were transformed into log relative risk and corresponding variance, with the formula proposed by Greenland [21]. When only the p-value was provided as measure of uncertainty, we calculated a ‘test-based’ estimate of variance [21]. When estimates were not available from the paper but only crude data were provided (for example, as a 2 by 2 table), we calculated crude, exact odds ratios and 95% CI. Summary relative risks (SRR) and 95% CI for the association between each STI and PC were obtained by pooling the studyspecific estimates by random effects models, with maximum likelihood estimates and 95% CI based on t-distribution [22], to be conservative. SRR were only calculated when five or more estimates were available. Heterogeneity across studies was evaluated using the I2 parameter, which represents the percentage of total variation across studies that is attributable to heterogeneity rather than to chance. Meta-regressions and sub-group analysis were carried out to investigate the influence of variables assumed to potentially confound or modify the association between STI and prostate cancer, such as country, publication year, study design, study setting (hospital- versus population based), method of exposure assessment, and percentage of non-Caucasian people in the study sample. Sensitivity analysis was carried out to verify the effect of single studies on the stability of the summary estimates. To verify whether publication bias might affect the validity of the estimates, funnel plots were investigated considering regression of ln(RR) on the sample size, weighted by the inverse of the pooled variance [23]. All analyses were performed with SAS software version 8.02 (SAS Institute Inc., Cary, NC, USA) and STATA software version 11 (Stata Inc., College Station, TX, USA). 3. Results Overall, 109 papers were obtained and examined for inclusion (Fig. 1). Eleven papers were excluded because they were reviews and/or meta-analyses. Forty-eight papers were excluded because
S. Caini et al. / Cancer Epidemiology 38 (2014) 329–338
331
Fig. 1. Flow-chart of the selection process for the studies included in the meta-analysis for the association between sexually transmitted infections and prostate cancer risk.
they did not present nor allowed to calculate an estimate for any association of interest; in three cases, authors stated that the association between any STI (two papers) and any of HSV-1, HSV-2 or CMV (one paper) with prostate cancer was not significant, but no measure of RR was provided nor could be obtained. Three papers were excluded because more recent results obtained from the same study sample were reported in other studies. Finally, 47 papers met inclusion criteria for the present metaanalysis (Table 1) [24–70], of which 34 have a case-control design, 10 were nested case-control studies, and 3 were cohort studies. Eighteen studies were population-based, twenty hospital-based, two enrolled cases from both general population and hospital, and in seven papers (all having a cohort or nested case-control design) PC cases arose within special-exposure cohorts. The papers included were published between 1971 and 2011. The papers by Hoffman [49] and Jenkins [60] presented results originating from respectively two and three different case-control studies, while the papers by Wynder [25], Ross [37] and Huang [63] provided separate results for white and black people, which were included separately in the meta-analysis. In five studies with a case-control design, PC cases were compared with two groups of controls: we chose to include estimates for the comparison of PC cases to hospital controls instead of patients suffering from BPH [24,39,40], to blood donors instead of patients with cancers not related to HHV-8 [49], and to neighbours instead of hospital controls [36]. In 16 papers [24–32,35,37,44,46,60,61,70] with a case-control or matched case-control design, no measure of relative risk was provided: a crude, exact OR and 95% CI were therefore calculated
from the raw distribution of exposure among cases and controls. In two papers [33,34], authors provided only a measure of relative risk and the exact value of the chi-square statistics, from which the 95% CI were calculated according to Greenland’s formulas [21]. Patients who reported having had gonorrhoea had a significantly increased risk of developing prostate cancer (SRR 1.20, 95% CI 1.05–1.37) (Table 2 and Fig. 2). This estimate was based on more than 10,000 PC patients from 23 separate studies. Three papers [47,55,57] also reported the risk of PC associated with an increasing number of lifetime episodes of gonorrhoea: the risk for the highest versus the lowest category, as defined in the respective paper, was 1.0 (95% CI 0.7–1.2), 3.3 (95% CI 1.4–7.8) and 1.8 (95% CI 1.1–3.1), respectively. The SRR for the association with PC risk was not significantly different from null value (Table 2) for syphilis (Fig. 2), HSV and HHV-8 (Fig. 3), and C. trachomatis (Fig. 4). Due to a too small number of estimates, we did not calculate any SRR for CMV (four estimates ranging between 0.62 [95% CI 0.32–1.14] and 1.14 [95% CI 0.41–3.15]), T. vaginalis (three estimates ranging between 0.97 [95% CI 0.70–1.34] and 1.43 [95% CI 1.00–2.03]), U. urealyticum and M. hominis (one estimate each: 2.06 [95% CI 1.08–4.28] for the former, 1.34 [95% CI 0.74–2.55] for the latter). Men who reported having ever had ‘‘any STI’’ had a significantly increased PC risk (SRR 1.49, 95% CI 1.19–1.92), based on 6313 PC cases from 18 studies (Table 2 and Fig. 4). A detailed definition of ‘‘any STI’’ was given in three papers only (Table 3): these definitions differ from each other and could include infective agents/STIs that were not considered in the present meta-analysis, like HPV [63,69] and chancroid [55].
S. Caini et al. / Cancer Epidemiology 38 (2014) 329–338
332
Table 1 Main characteristics of published studies included in the meta-analysis of sexually transmitted infections and prostate cancer risk. Source of casesb
Source of controlsb
% Black among cases
% Black among controls
BPH patients as controls
Variables used for adjustment
39
Hosp
Hosp
Unknown
Unknown
No
252
348
Hosp
Hosp
0.0%
0.0%
Unknown
CC
48
52
Hosp
Hosp
100.0%
100.0%
Unknown
Krain, 1973
CC
136
136
Pop
Pop
42.6%
42.6%
Unknown
[27]
Krain, 1974
CC
221
221
Hosp
Hosp
54.8%
54.8%
No
[28]
Heshmat, 1975
MCC
75
75
Hosp
Hosp
Unknown
Unknown
No
[29]
Herbert, 1976
CC
28
29
Hosp
BPH
Unknown
Unknown
Yes
[30]
Baker, 1981
CC
81
224
Hosp
BPH
Unknown
Unknown
Yes
[31]
Lu¨leci, 1981
CC
16
35
Hosp
Hosp
Unknown
Unknown
Yes
[32]
Ross, 1983
MCC
110
110
Pop
Pop
Unknown
Unknown
Unknown
[33] [34] [35]
Lees, 1985 Mishina, 1985 Checkoway, 1987
CC MCC CC
83 100 40
166 100 64
Hosp Hosp Hosp
BPH, hosp Pop Hosp
Unknown 0.0% 45.0%
Unknown 0.0% 26.6%
Yes No Yes
[36] [37]
Mandel, 1987 Ross, 1987e
MCC CC
250 142
478c 142
Pop Pop
Pop Pop
0.0% 0.0%
0.0% 0.0%
No Unknown
[37]
Ross, 1987e
CC
142
142
Pop
Pop
100.0%
100.0%
Unknown
[38] [39] [40] [41] [42] [43] [44]
Honda, 1988 Oishi, 1989 Oishi, 1990 Hiatt, 1994 Michalek, 1994 Ewings, 1996 Ilic´, 1996
CC CC CC NCC Cohort CC CC
216 100 100 238 25 159 101
216 100 100 238 10,262 325 202
Pop Hosp Hosp Pop Other Pop Hosp
Pop Hosp Hosp Pop Other Pop Hosp
0.0% 0.0% 0.0% 28.2% Unknown Unknown 0.0%
0.0% 0.0% 0.0% 28.2% Unknown Unknown 0.0%
Unknown No No Yes Unknown Yes No
[45] [46]
Dillner, 1998 Hsieh, 1999
NCC CC
165 320
290 246
Pop Hosp
Pop Hosp
Unknown Unknown
Unknown Unknown
Unknown Yes
[47] [48]
Hayes, 2000 Rosenblatt, 2001
CC CC
981 753
1315 703
Pop Pop
Pop Pop
48.8% 6.1%
45.2% 2.6%
Unknown Yes
[49] [49] [50]
Hoffman, 2004e Hoffman, 2004e Lightfoot, 2004
CC CC CC
138 100 760
140 177 1632
Pop Hosp Pop
Pop Other Pop
97.0% Unknown 3.1%
97.0% Unknown 3.0%
Yes Unknown Yes
[51]
Sanderson, 2004
CC
416
429
Pop
Pop
40.8%
42.4%
Unknown
[52] [53]
Anttila, 2005 Ferna´ndez, 2005
NCC CC
738 273
2271 254
Pop Hosp
Pop Hosp
Unknown 24.1%
Unknown 20.2%
Unknown Unknown
[54] [55]
Korodi, 2005 Patel, 2005
NCC CC
163 700
288 604
Pop Pop
Pop Pop
Unknown 50.4%
Unknown 42.6%
Unknown Unknown
[56] [57]
Pelucchi, 2006 Sarma, 2006
CC CC
280 129
689 703
Hosp Pop
Hosp Pop
Unknown 100.0%
Unknown 100.0%
Unknown Unknown
[58]
Sutcliffe, 2006
Cohort
2263
36,033
Other
Other
Unknown
Unknown
Unknown
[59]
Sutcliffe, 2006
NCC
691
691
Other
Other
Unknown
Unknown
Unknown
None (OR calculated from raw data) None (OR calculated from raw data) None (OR calculated from raw data) None (OR calculated from raw data) None (OR calculated from raw data) None (OR calculated from raw data) None (OR calculated from raw data) None (OR calculated from raw data) None (OR calculated from raw data) None (OR calculated from raw data) Age Age None (OR calculated from raw data) Age, ethnicity, SES, other None (OR calculated from raw data) None (OR calculated from raw data) Age, other Age, other Age, other Age, ethnicity, other Age Age None (OR calculated from raw data) Age, other None (OR calculated from raw data) Age, ethnicity, other Age, ethnicity, anthropometry, SES, smoking, alcohol, family history, other Age Age Age, anthropometry, SES, smoking, vasectomy, family history, other Age, ethnicity, anthropometry, SES, smoking, alcohol, family history, other Age, other Age, ethnicity, SES, family history, other Age, other Age, ethnicity, anthropometry, SES, smoking, alcohol, family history, other Age, SES, other Age, ethnicity, SES, smoking, family history, other Age, ethnicity, anthropometry, smoking, alcohol, vasectomy, family history, other Age, ethnicity, anthropometry, smoking, alcohol, vasectomy, family history, other
Ref.
Author, year
Study design
No. of PC cases
[24]
Steele, 1971
CC
39
[25]
Wynder, 1971e
CC
[25]
Wynder, 1971e
[26]
No. of controlsa
S. Caini et al. / Cancer Epidemiology 38 (2014) 329–338
333
Table 1 (Continued ) Source of casesb
Source of controlsb
% Black among cases
% Black among controls
BPH patients as controls
Variables used for adjustment
98
Hosp
BPH
0.0%
0.0%
Yes
10
34
Hosp
BPH
100.0%
100.0%
Yes
CC
199
155
Pop
Pop
47.7%
48.4%
Yes
Pourmand, 2007
CC
130
75
Hosp
Hosp
0.0%
0.0%
Unknown
[62]
Sutcliffe, 2007
NCC
691
691
Other
Other
0.9%
0.4%
Yes
[63]
Huang, 2008e
NCC
765
915
Pop
Pop
0.0%
0.0%
Yes
[63]
Huang, 2008e
NCC
103
368
Pop
Pop
100.0%
100.0%
Yes
[64] [65] [66]
Dennis, 2009 Dimitropoulou, 2009 Stark, 2009
NCC CC NCC
267 431 673
267 409 673
Other Hosp Other
Other Hosp Other
25.1% 4.4% Unknown
25.1% 1.0% Unknown
Unknown Unknown Unknown
[67]
Sutcliffe, 2009
NCC
616
616
Other
Other
6.2%
10.2%
no
[68]
Cheng, 2010
Cohort
1658
68,675
Pop
Pop
Unknown
Unknown
Yes
[69]
Ahmadi, 2011
CC
194
317
Hosp
Hosp
Unknown
Unknown
No
[70]
Hrbacek, 2011d
CC
329
105
Hosp
BPH
Unknown
Unknown
Yes
None (OR calculated from raw data) None (OR calculated from raw data) None (OR calculated from raw data) None (OR calculated from raw data) Age, ethnicity, tumour stage, family history, other Age, ethnicity, SES, smoking, family history, other Age, ethnicity, SES, smoking, family history, other Age, ethnicity, other Age, ethnicity Age, anthropometry, tumour stage, smoking, other Age, ethnicity, family history Age, ethnicity, anthropometry, SES, smoking, family history Age, anthropometry, smoking, alcohol, vasectomy, family history, other None (OR calculated from raw data)
Ref.
Author, year
Study design
No. of PC cases
[60]
Jenkins, 2007f
CC
41
[60]
f
Jenkins, 2007
CC
[60]
Jenkins, 2007f
[61]
No. of controlsa
PC, prostate cancer; BPH, patients with benign prostate hyperplasia; CC, case-control study; MCC, matched case-control study; NCC, nested case-control study; SES, socioeconomic status. a Cohort size for cohort studies. b Source of cohort members for cohort studies. c Number of controls for CMV and HSV was 240 instead of 478. d Presented a measure of relative risk for Mycoplasma hominis and Ureaplasma urealyticum as well. e Two separate estimates were reported, separately for white and black men (Refs. [25,37,63]), or originating from two different case-control studies [49]. f Three separate estimates were reported, originating from three different case-control studies.
Seven papers [48,54,55,57,63,64,68] reported a measure of relative risk for the association between ‘‘increasing number of any STI’’ and PC risk (Table 3). This could refer to either the number of different STIs (four papers) or the number of single episodes of STIs (two papers) a patient had during his lifetime; in one paper, no clear definition was given. As above, STIs included in the definition differ among papers (Table 3). The calculated SRR for any increase by one different STI/further episode was 1.03 (95% CI 0.90–1.18). The diagnosis of gonorrhoea and syphilis relied on self-reported history of disease in most cases, while for the other STIs/infective agents laboratory methods were used predominantly or exclu-
sively (Table 4). Laboratory methods used for diagnosing the same STI may differ among studies as well (Table 4). The study of between-study heterogeneity for the association of each STI considered and PC risk did not yield any significant result for any of the variables taken into consideration. The SRR for the association between syphilis and PC risk was slightly modified and approached statistical significance (SRR 1.22, 95% CI 0.97–1.54), with the I2 statistics dropping to 0, after excluding the two studies (together accounting for 1.2% of overall PC cases) where the diagnosis was not self-reported but either based on medical records [33] or laboratory test [42].
Table 2 Summary relative risk (SRR) and 95% confidence intervals (95% CI) for the association between sexually transmitted infections (STI) and prostate cancer (PCa). STI
No. of estimates
No. PC cases
SRR
95% CI
I2 (%)
Gonorrhoea Syphilis Chlamydia trachomatis HSV-2 HSV (any) HSV (all)a HHV-8 Any STI
23 17 9 8 5 14 9 18
10,379 8693 5469 1752 3490 5571 2210 6313
1.20 1.27 0.93 1.07 0.87 0.95 1.08 1.49
1.05–1.37 0.85–1.89 0.74–1.18 0.83–1.38 0.65–1.17 0.82–1.11 0.55–2.13 1.19–1.92
37 59 46 100 8 4 77 68
HSV, Herpes Simplex Virus; HHV, Human Herpes Virus. I2: proportion of total variation in study estimates that is due to heterogeneity. a Includes one estimate for HSV-1, eight estimates for HSV-2, and five estimates for HSV not specified whether -1 or -2.
S. Caini et al. / Cancer Epidemiology 38 (2014) 329–338
334
Fig. 2. Forest plot of gonorrhoea (a) and syphilis (b) and prostate cancer risk. SRR: summary relative risk.
Finally, we found a borderline statistically significant publication bias (p = 0.07) for the association between gonorrhoea and PC risk. 4. Discussion This meta-analysis revealed that men ever infected with gonorrhoea have a 20% increased risk of developing PC, based on over 10,000 PC cases from 23 independent studies. This is consistent with Taylor et al. [20] that obtained a slightly higher summary odds ratio with overlapping confidence intervals (SRR 1.35, 95% CI 1.05–1.83), based on about 2500 PC cases from 11 published studies (all included in our meta-analysis as well). The slight decrease in risk may have arisen by chance or be due to a higher proportion, in recent studies, of non-lethal PC diagnosed following a PSA testing. This finding does not vary by any of the variables considered for investigation of between-study heterogeneity, including study design, method of exposure assessment, ethnicity, country and publication year; however, its strength is somewhat weakened by the borderline significant publication bias we found. Besides gonorrhoea, no other single STI evaluated in at least five studies was significantly associated with an increased risk of PC. However, results on syphilis and infection with T. vaginalis deserve consideration. Concerning syphilis, all but two papers included in the present meta-analysis adopted the self-reported history as diagnostic criterion. Self-reported history is a reproducible, specific but not very sensitive diagnostic criterion for syphilis [71]. The SRR becomes borderline significant after exclusion of the two studies where the diagnosis is not based on self-reported diagnosis but on either medical records [33] or laboratory test [42]. A self-reported diagnosis of syphilis is usually a positive answer to
the question ‘‘Have you ever been told by a doctor that you had syphilis?’’ Thus, the former paper [33] is not conceptually different from those based on self-reported history of disease. In fact, this paper reports a statistically significant association between syphilis and PC risk (OR 2.9, p-value
Contents lists available at ScienceDirect
Cancer Epidemiology The International Journal of Cancer Epidemiology, Detection, and Prevention journal homepage: www.cancerepidemiology.net
Sexually transmitted infections and prostate cancer risk: A systematic review and meta-analysis Saverio Caini a,*, Sara Gandini b, Maria Dudas c, Viviane Bremer d, Ettore Severi e, Alin Gherasim f a
Unit of Molecular and Nutritional Epidemiology, Institute for Cancer Research and Prevention (ISPO), Italy Division of Epidemiology and Biostatistics, European Institute of Oncology, Milan, Italy Department of Infectious Disease Epidemiology, National Center for Epidemiology, Budapest, Hungary d Division for HIV/AIDS, STI and Blood-Borne Infections, Department for Infectious Disease Epidemiology, Robert Koch Institute, Berlin, Germany e South East Regional Epidemiology Unit, Health Protection Agency, London, UK f Swedish Institute for Communicable Disease Control, Solna, Sweden b c
A R T I C L E I N F O
A B S T R A C T
Article history: Received 3 April 2014 Received in revised form 4 June 2014 Accepted 10 June 2014 Available online 28 June 2014
Prostate cancer (PC) is the second most incident cancer and the sixth cause of death by cancer in men worldwide. Despite extensive research efforts, no modifiable risk factors have been consistently identified for PC risk. A number of studies have focused on possible relationships between sexually transmitted infections (STIs) and PC. We performed a meta-analysis to explore the association between infection caused by Neisseria gonorrheae, Treponema pallidum, Chlamydia trachomatis, Trichomonas vaginalis, Ureaplasma urealyticum, Mycoplasma hominis, Herpes Simplex Virus types 1 and 2, Human Herpes Virus 8 and Cytomegalovirus, and PC. We conducted a comprehensive, systematic bibliographic search of medical literature to identify relevant studies. We calculated summary relative risk (SRR) and 95% confidence intervals (CI) for the association between each STI and PC through random effect models. Subgroup, metaregression and sensitivity analyses were carried out to detect between-study heterogeneity and bias. We included 47 studies published between 1971 and 2011. Men who reported having ever had any STI in lifetime had an increased PC (SRR 1.49, 95% CI 1.19–1.92). We found a significantly increased PC risk in men having had gonorrhoea (SRR 1.20, 95% CI 1.05–1.37). No other single STI was significantly associated with PC. Due to high incidence of both STIs and PC worldwide, prevention of STIs may help preventing a considerable number of PC cases. ß 2014 Elsevier Ltd. All rights reserved.
Keywords: Sexually transmitted diseases Gonorrhoea Prostate cancer Review Meta-analysis
1. Introduction Prostate cancer (PC) is the second most incident cancer in men worldwide. In 2008 there were 903,500 estimated new cases and 258,400 estimated deaths worldwide, resulting as the sixth cause of death by cancer in men [1]. Despite extensive research effort, no modifiable risk factors have been consistently identified for PC, except perhaps smoking [2], obesity [3] and a sedentary lifestyle [4]. Sexually transmitted infections (STIs) represent a major public health problem worldwide. Human Papilloma Virus (HPV) and Herpes Simplex Virus (HSV) are common STIs worldwide [5,6], the
* Corresponding author at: Institute for Cancer Research and Prevention (ISPO), Unit of Molecular and Nutritional Epidemiology, Via delle Oblate 2, 50141 Florence, Italy. Tel.: +39 3485828406. E-mail addresses: [email protected], [email protected] (S. Caini). http://dx.doi.org/10.1016/j.canep.2014.06.002 1877-7821/ß 2014 Elsevier Ltd. All rights reserved.
former being also involved in the aetiology of cancer of cervix uteri and other anatomical sites [7,8]. Bacterial STIs are also emerging or re-emerging worldwide [9]. Cases of Chlamydia trachomatis have been increasingly reported during past 20 years [10–13]. Although this trend is partly attributable to extended screening efforts and more sensitive tests, several cases still go undiagnosed due to underreporting or asymptomatic disease, especially among women. Gonorrhoea is especially increasing in those countries with previously low incidence rates [10–15], despite being still at near-historic lows; more worryingly, resistance to 3rd generation cephalosporins is being increasingly reported [16]. Syphilis incidence is overall stable in developed countries, but on rise among middle-aged men and men who have sex with men [10,11,13,15]. Due to high proportions of asymptomatic cases, little is known on epidemiology of other STIs, such as infections with Trichomonas vaginalis or Human Herpes Virus 8 (HHV-8, also known as Kaposi’s sarcoma-associated herpes virus).
330
S. Caini et al. / Cancer Epidemiology 38 (2014) 329–338
The first claims of an aetiological role of STIs in the development of PC date back to the 1950s [17], and several mechanisms were subsequently proposed to explain this association. For gonorrhoea and other bacterial infections, the process was investigated that leads to PC through the phases of prostate inflammation and prostate atrophy, whereas for viral infections the emphasis was placed on the transforming properties of viruses, in particular herpes viruses [18]. It has also been hypothesized that multiple episodes of STIs and infections of longer duration (as in the case of infections not or inefficiently treated) may represent a greater risk for PC development, due to a higher cumulative risk of prostate involvement [18]. A recently published meta-analysis showed a weak association between HPV-16 and PC and no association for HPV-18 [19]. For other STIs, the most recent meta-analysis, published in 2005, found significantly increased PC risk in men with a history of gonorrhoea (1.35, 95% CI 1.05–1.83) and any STI (1.48, 95% CI 1.26–1.73), but not syphilis (1.42, 95% CI 0.76–2.64) [20]. Since 2001, year of publication of the most recent paper included in the latter metaanalysis, several epidemiological studies have been conducted to explore the association between STIs and PC risk. We performed a meta-analysis to explore the association between gonorrhoea, syphilis, and other STIs (other than HPV) and PC risk. 2. Materials and methods
When two or more papers published results originating from the same study sample, as it may happen with cohorts analyzed at different points in time, we only considered the most recently published results. 2.4. Data extraction Two authors (SC and ES) independently extracted the following information from each paper included in the meta-analysis: year of publication, country, study design, source of cohort members (for cohort, case-cohort and nested case-control studies) or of cases and controls (for case-control studies), type of matching (if any), mean/median age, ethnicity, inclusion of controls with familial history of PC, inclusion of controls suffering from cancers other than PC and/or benign prostatic hyperplasia (BPH), blind assessment of exposure status, type of diagnosis for exposure, and adjustments that were made for statistical analyses. Any discrepancies in study selection and data extraction (emerged by comparing the two database that were prepared) were handled by discussion between the authors; if no agreement could be reached, another co-author (SG) would decide. For case-control studies with more than one control group we considered, for each association of interest, the odds ratio originating from the control group that was judged, on an a priori basis, to yield least biased estimates.
2.1. Definition of outcome and exposures 2.5. Statistical analysis We considered as outcome histologically confirmed prostate cancer, diagnosed through clinical examination or after consulting cancer registries. We defined exposure as either self-reported history, clinical or serological diagnosis of infection caused by any of the following agents: Neisseria gonorrheae, Treponema pallidum, C. trachomatis, T. vaginalis, Ureaplasma urealyticum, Mycoplasma hominis, HSV types 1 and 2, HHV-8 and Cytomegalovirus (CMV). We also considered for meta-analysis the association between reported history of ‘‘any STI’’ (either specified or not) and PC risk. 2.2. Data sources and search strategy We identified eligible studies to be included in the metaanalysis by reviewing published reports listed in the following databases: PUBMED, Ovid Medline, EMBASE, and ISI Web of Science – Science Citation Index Expanded. To search for published papers, we used all possible combinations of MESH terms defining the outcome (‘‘prostatic cancer’’ OR ‘‘prostate cancer’’) AND each of the aforementioned infective agents (see above) and STI caused by them (i.e., ‘‘gonorrhoea’’, ‘‘syphilis’’, ‘‘trachoma’’, ‘‘trichomoniasis’’, etc.). Additional studies were collected by searching through references lists of retrieved articles and previously published meta-analyses and reviews. A full copy was obtained of papers that were considered of interest after reading the abstract. 2.3. Criteria for including studies Retrieved papers published up to August 31, 2013 were included in the meta-analysis if they met the following inclusion criteria: human observational studies; a cohort, case-control (including matched and nested case-control) or case-cohort design; reporting or providing sufficient information for estimating a measure of relative risk (RR) (incidence rate ratio, risk ratio, odds ratio, hazard ratio, standardized incidence ratio) with 95% confidence intervals (95% CI) or another measure of statistical uncertainty (standard errors, variance, or exact p-value of the significance of the estimates). No language or time restrictions were applied.
For any different pair of outcome and exposure, the most adjusted measure of association and the corresponding confidence intervals were transformed into log relative risk and corresponding variance, with the formula proposed by Greenland [21]. When only the p-value was provided as measure of uncertainty, we calculated a ‘test-based’ estimate of variance [21]. When estimates were not available from the paper but only crude data were provided (for example, as a 2 by 2 table), we calculated crude, exact odds ratios and 95% CI. Summary relative risks (SRR) and 95% CI for the association between each STI and PC were obtained by pooling the studyspecific estimates by random effects models, with maximum likelihood estimates and 95% CI based on t-distribution [22], to be conservative. SRR were only calculated when five or more estimates were available. Heterogeneity across studies was evaluated using the I2 parameter, which represents the percentage of total variation across studies that is attributable to heterogeneity rather than to chance. Meta-regressions and sub-group analysis were carried out to investigate the influence of variables assumed to potentially confound or modify the association between STI and prostate cancer, such as country, publication year, study design, study setting (hospital- versus population based), method of exposure assessment, and percentage of non-Caucasian people in the study sample. Sensitivity analysis was carried out to verify the effect of single studies on the stability of the summary estimates. To verify whether publication bias might affect the validity of the estimates, funnel plots were investigated considering regression of ln(RR) on the sample size, weighted by the inverse of the pooled variance [23]. All analyses were performed with SAS software version 8.02 (SAS Institute Inc., Cary, NC, USA) and STATA software version 11 (Stata Inc., College Station, TX, USA). 3. Results Overall, 109 papers were obtained and examined for inclusion (Fig. 1). Eleven papers were excluded because they were reviews and/or meta-analyses. Forty-eight papers were excluded because
S. Caini et al. / Cancer Epidemiology 38 (2014) 329–338
331
Fig. 1. Flow-chart of the selection process for the studies included in the meta-analysis for the association between sexually transmitted infections and prostate cancer risk.
they did not present nor allowed to calculate an estimate for any association of interest; in three cases, authors stated that the association between any STI (two papers) and any of HSV-1, HSV-2 or CMV (one paper) with prostate cancer was not significant, but no measure of RR was provided nor could be obtained. Three papers were excluded because more recent results obtained from the same study sample were reported in other studies. Finally, 47 papers met inclusion criteria for the present metaanalysis (Table 1) [24–70], of which 34 have a case-control design, 10 were nested case-control studies, and 3 were cohort studies. Eighteen studies were population-based, twenty hospital-based, two enrolled cases from both general population and hospital, and in seven papers (all having a cohort or nested case-control design) PC cases arose within special-exposure cohorts. The papers included were published between 1971 and 2011. The papers by Hoffman [49] and Jenkins [60] presented results originating from respectively two and three different case-control studies, while the papers by Wynder [25], Ross [37] and Huang [63] provided separate results for white and black people, which were included separately in the meta-analysis. In five studies with a case-control design, PC cases were compared with two groups of controls: we chose to include estimates for the comparison of PC cases to hospital controls instead of patients suffering from BPH [24,39,40], to blood donors instead of patients with cancers not related to HHV-8 [49], and to neighbours instead of hospital controls [36]. In 16 papers [24–32,35,37,44,46,60,61,70] with a case-control or matched case-control design, no measure of relative risk was provided: a crude, exact OR and 95% CI were therefore calculated
from the raw distribution of exposure among cases and controls. In two papers [33,34], authors provided only a measure of relative risk and the exact value of the chi-square statistics, from which the 95% CI were calculated according to Greenland’s formulas [21]. Patients who reported having had gonorrhoea had a significantly increased risk of developing prostate cancer (SRR 1.20, 95% CI 1.05–1.37) (Table 2 and Fig. 2). This estimate was based on more than 10,000 PC patients from 23 separate studies. Three papers [47,55,57] also reported the risk of PC associated with an increasing number of lifetime episodes of gonorrhoea: the risk for the highest versus the lowest category, as defined in the respective paper, was 1.0 (95% CI 0.7–1.2), 3.3 (95% CI 1.4–7.8) and 1.8 (95% CI 1.1–3.1), respectively. The SRR for the association with PC risk was not significantly different from null value (Table 2) for syphilis (Fig. 2), HSV and HHV-8 (Fig. 3), and C. trachomatis (Fig. 4). Due to a too small number of estimates, we did not calculate any SRR for CMV (four estimates ranging between 0.62 [95% CI 0.32–1.14] and 1.14 [95% CI 0.41–3.15]), T. vaginalis (three estimates ranging between 0.97 [95% CI 0.70–1.34] and 1.43 [95% CI 1.00–2.03]), U. urealyticum and M. hominis (one estimate each: 2.06 [95% CI 1.08–4.28] for the former, 1.34 [95% CI 0.74–2.55] for the latter). Men who reported having ever had ‘‘any STI’’ had a significantly increased PC risk (SRR 1.49, 95% CI 1.19–1.92), based on 6313 PC cases from 18 studies (Table 2 and Fig. 4). A detailed definition of ‘‘any STI’’ was given in three papers only (Table 3): these definitions differ from each other and could include infective agents/STIs that were not considered in the present meta-analysis, like HPV [63,69] and chancroid [55].
S. Caini et al. / Cancer Epidemiology 38 (2014) 329–338
332
Table 1 Main characteristics of published studies included in the meta-analysis of sexually transmitted infections and prostate cancer risk. Source of casesb
Source of controlsb
% Black among cases
% Black among controls
BPH patients as controls
Variables used for adjustment
39
Hosp
Hosp
Unknown
Unknown
No
252
348
Hosp
Hosp
0.0%
0.0%
Unknown
CC
48
52
Hosp
Hosp
100.0%
100.0%
Unknown
Krain, 1973
CC
136
136
Pop
Pop
42.6%
42.6%
Unknown
[27]
Krain, 1974
CC
221
221
Hosp
Hosp
54.8%
54.8%
No
[28]
Heshmat, 1975
MCC
75
75
Hosp
Hosp
Unknown
Unknown
No
[29]
Herbert, 1976
CC
28
29
Hosp
BPH
Unknown
Unknown
Yes
[30]
Baker, 1981
CC
81
224
Hosp
BPH
Unknown
Unknown
Yes
[31]
Lu¨leci, 1981
CC
16
35
Hosp
Hosp
Unknown
Unknown
Yes
[32]
Ross, 1983
MCC
110
110
Pop
Pop
Unknown
Unknown
Unknown
[33] [34] [35]
Lees, 1985 Mishina, 1985 Checkoway, 1987
CC MCC CC
83 100 40
166 100 64
Hosp Hosp Hosp
BPH, hosp Pop Hosp
Unknown 0.0% 45.0%
Unknown 0.0% 26.6%
Yes No Yes
[36] [37]
Mandel, 1987 Ross, 1987e
MCC CC
250 142
478c 142
Pop Pop
Pop Pop
0.0% 0.0%
0.0% 0.0%
No Unknown
[37]
Ross, 1987e
CC
142
142
Pop
Pop
100.0%
100.0%
Unknown
[38] [39] [40] [41] [42] [43] [44]
Honda, 1988 Oishi, 1989 Oishi, 1990 Hiatt, 1994 Michalek, 1994 Ewings, 1996 Ilic´, 1996
CC CC CC NCC Cohort CC CC
216 100 100 238 25 159 101
216 100 100 238 10,262 325 202
Pop Hosp Hosp Pop Other Pop Hosp
Pop Hosp Hosp Pop Other Pop Hosp
0.0% 0.0% 0.0% 28.2% Unknown Unknown 0.0%
0.0% 0.0% 0.0% 28.2% Unknown Unknown 0.0%
Unknown No No Yes Unknown Yes No
[45] [46]
Dillner, 1998 Hsieh, 1999
NCC CC
165 320
290 246
Pop Hosp
Pop Hosp
Unknown Unknown
Unknown Unknown
Unknown Yes
[47] [48]
Hayes, 2000 Rosenblatt, 2001
CC CC
981 753
1315 703
Pop Pop
Pop Pop
48.8% 6.1%
45.2% 2.6%
Unknown Yes
[49] [49] [50]
Hoffman, 2004e Hoffman, 2004e Lightfoot, 2004
CC CC CC
138 100 760
140 177 1632
Pop Hosp Pop
Pop Other Pop
97.0% Unknown 3.1%
97.0% Unknown 3.0%
Yes Unknown Yes
[51]
Sanderson, 2004
CC
416
429
Pop
Pop
40.8%
42.4%
Unknown
[52] [53]
Anttila, 2005 Ferna´ndez, 2005
NCC CC
738 273
2271 254
Pop Hosp
Pop Hosp
Unknown 24.1%
Unknown 20.2%
Unknown Unknown
[54] [55]
Korodi, 2005 Patel, 2005
NCC CC
163 700
288 604
Pop Pop
Pop Pop
Unknown 50.4%
Unknown 42.6%
Unknown Unknown
[56] [57]
Pelucchi, 2006 Sarma, 2006
CC CC
280 129
689 703
Hosp Pop
Hosp Pop
Unknown 100.0%
Unknown 100.0%
Unknown Unknown
[58]
Sutcliffe, 2006
Cohort
2263
36,033
Other
Other
Unknown
Unknown
Unknown
[59]
Sutcliffe, 2006
NCC
691
691
Other
Other
Unknown
Unknown
Unknown
None (OR calculated from raw data) None (OR calculated from raw data) None (OR calculated from raw data) None (OR calculated from raw data) None (OR calculated from raw data) None (OR calculated from raw data) None (OR calculated from raw data) None (OR calculated from raw data) None (OR calculated from raw data) None (OR calculated from raw data) Age Age None (OR calculated from raw data) Age, ethnicity, SES, other None (OR calculated from raw data) None (OR calculated from raw data) Age, other Age, other Age, other Age, ethnicity, other Age Age None (OR calculated from raw data) Age, other None (OR calculated from raw data) Age, ethnicity, other Age, ethnicity, anthropometry, SES, smoking, alcohol, family history, other Age Age Age, anthropometry, SES, smoking, vasectomy, family history, other Age, ethnicity, anthropometry, SES, smoking, alcohol, family history, other Age, other Age, ethnicity, SES, family history, other Age, other Age, ethnicity, anthropometry, SES, smoking, alcohol, family history, other Age, SES, other Age, ethnicity, SES, smoking, family history, other Age, ethnicity, anthropometry, smoking, alcohol, vasectomy, family history, other Age, ethnicity, anthropometry, smoking, alcohol, vasectomy, family history, other
Ref.
Author, year
Study design
No. of PC cases
[24]
Steele, 1971
CC
39
[25]
Wynder, 1971e
CC
[25]
Wynder, 1971e
[26]
No. of controlsa
S. Caini et al. / Cancer Epidemiology 38 (2014) 329–338
333
Table 1 (Continued ) Source of casesb
Source of controlsb
% Black among cases
% Black among controls
BPH patients as controls
Variables used for adjustment
98
Hosp
BPH
0.0%
0.0%
Yes
10
34
Hosp
BPH
100.0%
100.0%
Yes
CC
199
155
Pop
Pop
47.7%
48.4%
Yes
Pourmand, 2007
CC
130
75
Hosp
Hosp
0.0%
0.0%
Unknown
[62]
Sutcliffe, 2007
NCC
691
691
Other
Other
0.9%
0.4%
Yes
[63]
Huang, 2008e
NCC
765
915
Pop
Pop
0.0%
0.0%
Yes
[63]
Huang, 2008e
NCC
103
368
Pop
Pop
100.0%
100.0%
Yes
[64] [65] [66]
Dennis, 2009 Dimitropoulou, 2009 Stark, 2009
NCC CC NCC
267 431 673
267 409 673
Other Hosp Other
Other Hosp Other
25.1% 4.4% Unknown
25.1% 1.0% Unknown
Unknown Unknown Unknown
[67]
Sutcliffe, 2009
NCC
616
616
Other
Other
6.2%
10.2%
no
[68]
Cheng, 2010
Cohort
1658
68,675
Pop
Pop
Unknown
Unknown
Yes
[69]
Ahmadi, 2011
CC
194
317
Hosp
Hosp
Unknown
Unknown
No
[70]
Hrbacek, 2011d
CC
329
105
Hosp
BPH
Unknown
Unknown
Yes
None (OR calculated from raw data) None (OR calculated from raw data) None (OR calculated from raw data) None (OR calculated from raw data) Age, ethnicity, tumour stage, family history, other Age, ethnicity, SES, smoking, family history, other Age, ethnicity, SES, smoking, family history, other Age, ethnicity, other Age, ethnicity Age, anthropometry, tumour stage, smoking, other Age, ethnicity, family history Age, ethnicity, anthropometry, SES, smoking, family history Age, anthropometry, smoking, alcohol, vasectomy, family history, other None (OR calculated from raw data)
Ref.
Author, year
Study design
No. of PC cases
[60]
Jenkins, 2007f
CC
41
[60]
f
Jenkins, 2007
CC
[60]
Jenkins, 2007f
[61]
No. of controlsa
PC, prostate cancer; BPH, patients with benign prostate hyperplasia; CC, case-control study; MCC, matched case-control study; NCC, nested case-control study; SES, socioeconomic status. a Cohort size for cohort studies. b Source of cohort members for cohort studies. c Number of controls for CMV and HSV was 240 instead of 478. d Presented a measure of relative risk for Mycoplasma hominis and Ureaplasma urealyticum as well. e Two separate estimates were reported, separately for white and black men (Refs. [25,37,63]), or originating from two different case-control studies [49]. f Three separate estimates were reported, originating from three different case-control studies.
Seven papers [48,54,55,57,63,64,68] reported a measure of relative risk for the association between ‘‘increasing number of any STI’’ and PC risk (Table 3). This could refer to either the number of different STIs (four papers) or the number of single episodes of STIs (two papers) a patient had during his lifetime; in one paper, no clear definition was given. As above, STIs included in the definition differ among papers (Table 3). The calculated SRR for any increase by one different STI/further episode was 1.03 (95% CI 0.90–1.18). The diagnosis of gonorrhoea and syphilis relied on self-reported history of disease in most cases, while for the other STIs/infective agents laboratory methods were used predominantly or exclu-
sively (Table 4). Laboratory methods used for diagnosing the same STI may differ among studies as well (Table 4). The study of between-study heterogeneity for the association of each STI considered and PC risk did not yield any significant result for any of the variables taken into consideration. The SRR for the association between syphilis and PC risk was slightly modified and approached statistical significance (SRR 1.22, 95% CI 0.97–1.54), with the I2 statistics dropping to 0, after excluding the two studies (together accounting for 1.2% of overall PC cases) where the diagnosis was not self-reported but either based on medical records [33] or laboratory test [42].
Table 2 Summary relative risk (SRR) and 95% confidence intervals (95% CI) for the association between sexually transmitted infections (STI) and prostate cancer (PCa). STI
No. of estimates
No. PC cases
SRR
95% CI
I2 (%)
Gonorrhoea Syphilis Chlamydia trachomatis HSV-2 HSV (any) HSV (all)a HHV-8 Any STI
23 17 9 8 5 14 9 18
10,379 8693 5469 1752 3490 5571 2210 6313
1.20 1.27 0.93 1.07 0.87 0.95 1.08 1.49
1.05–1.37 0.85–1.89 0.74–1.18 0.83–1.38 0.65–1.17 0.82–1.11 0.55–2.13 1.19–1.92
37 59 46 100 8 4 77 68
HSV, Herpes Simplex Virus; HHV, Human Herpes Virus. I2: proportion of total variation in study estimates that is due to heterogeneity. a Includes one estimate for HSV-1, eight estimates for HSV-2, and five estimates for HSV not specified whether -1 or -2.
S. Caini et al. / Cancer Epidemiology 38 (2014) 329–338
334
Fig. 2. Forest plot of gonorrhoea (a) and syphilis (b) and prostate cancer risk. SRR: summary relative risk.
Finally, we found a borderline statistically significant publication bias (p = 0.07) for the association between gonorrhoea and PC risk. 4. Discussion This meta-analysis revealed that men ever infected with gonorrhoea have a 20% increased risk of developing PC, based on over 10,000 PC cases from 23 independent studies. This is consistent with Taylor et al. [20] that obtained a slightly higher summary odds ratio with overlapping confidence intervals (SRR 1.35, 95% CI 1.05–1.83), based on about 2500 PC cases from 11 published studies (all included in our meta-analysis as well). The slight decrease in risk may have arisen by chance or be due to a higher proportion, in recent studies, of non-lethal PC diagnosed following a PSA testing. This finding does not vary by any of the variables considered for investigation of between-study heterogeneity, including study design, method of exposure assessment, ethnicity, country and publication year; however, its strength is somewhat weakened by the borderline significant publication bias we found. Besides gonorrhoea, no other single STI evaluated in at least five studies was significantly associated with an increased risk of PC. However, results on syphilis and infection with T. vaginalis deserve consideration. Concerning syphilis, all but two papers included in the present meta-analysis adopted the self-reported history as diagnostic criterion. Self-reported history is a reproducible, specific but not very sensitive diagnostic criterion for syphilis [71]. The SRR becomes borderline significant after exclusion of the two studies where the diagnosis is not based on self-reported diagnosis but on either medical records [33] or laboratory test [42]. A self-reported diagnosis of syphilis is usually a positive answer to
the question ‘‘Have you ever been told by a doctor that you had syphilis?’’ Thus, the former paper [33] is not conceptually different from those based on self-reported history of disease. In fact, this paper reports a statistically significant association between syphilis and PC risk (OR 2.9, p-value
