Association between allergic conditions and risk of prostate cancer: A Prisma-Compliant Systematic Review and Meta-Analysis

Association between allergic conditions and prostate cancer risk has been investigated for many years. However, the results from available evidence for the association are inconsistent. We conducted a meta-analysis to evaluate the relationship between allergic conditions (asthma, atopy, hay fever and “any allergy”) and risk of prostate cancer. The PubMed and Embase databases were searched to screen observational studies meeting our meta-analysis criteria. Study selection and data extraction from included studies were independently performed by two authors. Twenty studies were considered eligible involving 5 case-control studies and 15 cohort studies. The summary relative risk (RR) for developing prostate cancer risk was 1.04 (95%CI: 0.92–1.17) for asthma, and 1.25 (95%CI: 0.74–2.10) for atopy, 1.04 (95%CI: 0.99–1.09) for hay fever, 0.96 (95%CI: 0.86–1.06) for any allergy. In the Subgroup and sensitivity analysis, similar results were produced. Little evidence of publication bias was observed. The present meta-analysis of observational studies indicates that no indication of an association between allergic conditions and risk of prostate cancer was found, and the meta-analysis does not support neither the original hypothesis of an overall cancer protective effect of allergic conditions, nor that of an opposite effect in the development of prostate cancer.


Study characteristics.
The Tables 1 and 2 shows the descriptive data for all included studies. A total of 20 studies, comprising 5 case-control studies with 2,924 incident cases and 7,175 controls and 15 cohort studies including 16,526 cases and 1,681,562 pariticipants, contributed to the meta-analysis. These studies were published from 1985 to 2015. The number of prostate cancer patients ranged from 1 to 6,294 in the case-control studies and from 10 to 1,936 in the cohort studies. Ten studies were conducted in Europe 10,[12][13][14]17,23,24,[28][29][30] , six in the North America 11,16,18,19,22,31 , two in Australia 15,25 and two in Asia 21,26 . Eighteen studies reported findings for prostate cancer incidence [10][11][12][13][14][16][17][18][19][21][22][23][24][25][26]28,29,31 , whereas only two studies reported results for prostate cancer mortality 16,30 . We included a total of 5,757 prostate cancer deaths, 16,526 prostate cancer cases in the meta-analysis. The maximum numbers (1,102,247) of participants were from the Canada prospective study and minimum number (1,522) were from Australia prospective study in the cohort studies, while the maximum incident patients (1,936) were from the Canada Montreal PROtEuS study and minimum incident patients (10) were from Japan study in the case-control studies. The exposure categories that were measured were: 1) asthma; 2) hay fever (rhinitis); 3) atopy; and 4) any allergy (atopy and/or asthma, hay fever, or other allergic disease). For the assessment of allergic conditions, allergen-specific IgE measurement, skin prick testing, self-reported questionnaire, interviews by medical staff, and hospital discharge register were employed. For the assessment of prostate cancer, fourteen studies reported that cancers were histologically confirmed or were identified from national/regional cancer registeries, which we assumed verified the cancer pathologically. The remaining studies based their cancer diagnoses on different criteria; admission/discharge diagnoses; an automated general practice database; a linked national death register; review of hospital and nursing home recorder and death certificates. Six studies were designed to evaluate OR 10,[17][18][19]22,31 , five evaluated HR 15,25,26,29,30 , three evaluated RR 11,16,21,28 , while five studies compared observed cancer incidence rates in cohort of patients with asthma or hay fever/allergic rhinitis against expected numbers estimated from population-based cancer registries and their effect estimates were standardized incidence ratio (SIR) 13,14,23,24 or standardized mortality ratio(SMR) 12 .
Adjusted effect estimates could be determined for most cohort and case-control studies. Most risk estimates were adjusted for age (n = 12) 10   Continued index (n = 8) 10,[15][16][17]25,[28][29][30] . Some studies were also controlled for alcohol consumption (n = 5) 10,16,17,25,29 and race (n = 3) 16,19,28 , but few studies were adjusted for family history of prostate cancer (n = 2), total energy intake (n = 1) 25 , and intake of vegetable and red meat (n = 1) 16 . None of the studies were adjusted for exposure to heavy metals and androgen. The methodological quality of the included studies was generally good. The NOS scores ranged from five to seven ( Table 3). The median NOS score was 6.0.
Hay fever and risk of prostate cancer. The association of hay fever with prostate cancer was examined in 7 studies 10,11,16,18,23,28,31 . The six studies reported RR/OR 10,11,16,18,28,31 (Fig. 4), while SIR was reported in only one study, which reported a positive association between hay fever/allergic rhinitis and prostate cancer risk 23 . The combined relative risk was 1.04 (95%CI: 0.99-1.09), with low heterogeneity (P for heterogeneity = 0.428; I 2 = 0.0%). Subgroup and sensitivity analysis produced similar results (   asthma and mortality of prostate cancer 16 ). Both the Begg rank correlation test (P = 0.707) and the Egger linear regression test (P = 0.775) in the meta-analysis indicated no significant publication bias. Because number of included studies less than 10, we did not perform the funnel plot.
Atopy and risk of prostate cancer. The association of atopy with prostate cancer was investigated in 3 studies 10,15,29 . The combined relative risk was 1.25 (95%CI: 0.74-2.10), with notable heterogeneity (P for heterogeneity = 0.024; I 2 = 73.2%) (Fig. 5). In subgroup and sensitivity analysis, the results showed basically consistent with the overall analysis ( Table 4). The include studies achieved six or more stars and considered to be of high quality, so the result was consistent with overall analysis. When the analysis was restricted to studies adjusted for age, race, BMI, alcohol drinking and cigarette smoking, we found no association between asthma and prostate cancer. In a sensitivity analysis, similar results were observed, which ranged from 1.05 (95%CI: 0.62-1.77) with low heterogeneity (P for heterogeneity = 0.036, I 2 = 77.6%) (excluding the study by Talbot-Smith et al. 15 ) to 1.59 (95%CI: 0.93-2.71) with significant heterogeneity (P for heterogeneity = 0.193, I 2 = 41.0%) (excluding the study by Skaaby et al. 29 ). Both Egger's test (P = 1.000) and Bgger's test (P = 0.761) showed no publication bias. Because number of included studies less than 10, we did not perform the funnel plot.
Any allergy and risk of prostate cancer. The association of asthma with prostate cancer was investigated in 6 studies 10,11,16,19,22,31 . The combined relative risk was 0.96 (0.86-1.06), with low heterogeneity (P for heterogeneity = 0.287; I 2 = 19.3%). Subgroup and sensitivity analysis yielded similar result (Fig. 6). In subgroup and sensitivity analysis, the results showed basically consistent with the overall analysis (  16 ). The Begg rank correlation test (P = 0.133) and Egger linear regression test (P = 0.489) also indicated no evidence of publication bias. Because number of included studies less than 10, we did not perform the funnel plot.

Discussion
Allergic diseases (immune mediated conditions), encompassing hay fever, allergic asthma and atopy, are caused by inappropriate immunological response to antigens that do not elicit response in most individuals. Allergy-related carcinogenesis is a topic of interest but has generated considerable controversy. There has been a long-standing interest in determining whether individuals with allergic diseases have an altered risk of developing cancer. Several studies investigated the association between allergic diseases and specific cancers. Asthma, hay fever, and atopy have been associated with the risk of several specific cancers, such as pancreatic cancer, lymphomas, brain tumors, breast cancer and leukemia, although inconsistently 10,14,15,46,47 . Allergic diseases could theoretically both prevent and induce the development of several specific cancers. Two hypotheses that attempt to explain the possible mechanism between allergic diseases and cancer are immune surveillance and the antigenic stimulation theory 48 . Allergy might enhance the human immune system to recognize and eliminate cancer cells. In contrast, the antigenic stimulation hypothesis proposes that hyperactive immune conditions trigger chronic cellular inflammation, resulting in DNA mutation in dividing cells and inevitably leading to cancer development 49 . The determination of whether asthma, hay fever, atopy and any allergy are associated with prostate  Table 4. Results of subgroup analysis of asthma, hay fever, atopy, and any allergy.  20 . However, studies published subsequent to this report have reported inconsistent results [23][24][25][26][27][28][29][30][31] . In addition, the relation between allergic diseases and cancer risk remain unclear, and appear to be site-specific, we conducted an up-dated meta-analysis to summarize the current proof to evaluate the association between allergic conditions and prostate cancer risk.
The meta-analysis suggested that there is little observational support for the two theories in the development of prostate cancer. The effect of hyperreactive state and/or immune surveillance theories maybe mutually offset in the body.  The present meta-analysis exhibited several strengths, compared to the previous published meta-analysis. The first research highlight of this meta-analysis is its large sample size. The large number of total cases provided high statistical power to quantitatively evaluate the association between allergic conditions and prostate cancer risk. In addition, we expand the meta-analysis and the association between different geographic regions and study design, adjusted for covariates was also explored. Second, publication bias is a potential concern in any meta-analysis because small studies with null results do not get published. However, in our meta-analysis, we found little evidence of publication bias.
Nevertheless, there are some several limitations in the present meta-analysis. First, case-control studies have intrinsic limitations, such as selective bias and recall or memory bias. This limitation can partly explained the different results between case-control and cohort studies in the stratified analysis. Second, we cannot exclude the possibility that the observed null relationship between allergic disorders and prostate cancer risk is attributed to confounding factors. Majority of the studies were adjusted for potential confounding factors, but not all potential confounders were adjusted in every study. Although the allergic conditions and prostate cancer share common some potential confounders, such as cigarette smoking, alcohol drinking, old age, BMI and race, in analysis stratified by adjusting the smoking status, age, alcohol drinking BMI and race, similar results were obtained. Third, measurement error in assessment of allergic conditions are known to bias effect estimates, however, none of included studies in the meta-analysis made any corrections for measurement errors. An accurate assessment of allergic conditions remains a challenge, because these measures are based on different assessment methods, such as allergen-specific IgE measurement, skin prick testing, self-reported questionnaire, interviewed by medical staff, and hospital discharge register. Diagnosis of allergic conditions was largely based on a self-reported. Only a few studied were based on measurements that may be condisered to be more objective: serum LgE or skin prick tests. The increasing errors in measurements become inevitable. The imprecise measurement of allergic disorders might have attenuated the true associations. Fourth, potential sources of between-study heterogeneity, which is common in meta-analysis, should be explored although the low heterogeneity was found between hay fever or atopy and prostate cancer risk. Results from subgroup and sensitivity analysis indicated that geographic region, study design, quality of NOS may be potential sources of heterogeneity. Thirds, in subgroup analysis, there was a marginally positive association between hay fever and prostate cancer risk in studies that adjusted for age and alcohol drinking and conducted in the Europe, but a null association among studies adjusted for cigarette smoking and race. Due to the numerous comparisons and few included studies, this positive finding may have been a chance finding. It indicated that more relevant articles are needed to further explore this association. Fifth, analysis of the length of the induction period allows for characterising an exposure-outcome relationship and for falsifying the pathway assumed. The longer the supposed time sequence between the exposure and the occurrence of the outcome, the more crucial to analyze the empirical induction period. The average follow-up ranged from 5.05 years to 43 years among included cohort studies. Patients were followed up over five years in majority of the studies (92.8%). Therefore, the observation period in the included cohort studies covered a reasonable induction period. Overall, these limitations may affect our final conclusions.
In the meta-analysis, we found no indication of an association between allergic conditions (asthma, atopy, hay fever, or any allergy) and risk of prostate cancer, and there is little observational support for the immune surveillace theory or antigenic stimulation theory in the development of prostate cancer risk. However, these results should be carefully interpreted because of the significant heterogeneity among studies and potential confounders. Additional large-scale and high-quality prospective studies are needed to confirm the association between allergic conditions and risk of prostate cancer.

Methods
Literature search. A literature search was performed in March 15, 2016 without restriction to regions, publication types, or languages. The primary sources were the electronic databases of Pubmed and Embase databases. To identify eligible studies, the main search employed various combinations of Medical Subject Headings (MeSH) and non-MeSH terms "prostate carcinoma" OR "prostatic cancer" OR "prostate cancer" OR "prostatic carcinoma" combined with "asthma" OR "asthma*" OR "allergy" OR "allerg*" OR "atopy". The main search was completed independently by two investigators. Any discrepancy was solved by consultation of an investigator, not involved in the initial procedure. Moreover, the reference lists of all the studies and published systematic reviews, meta-analysis were also screened to identify other potentially eligible studies.
Study selection. To minimize the differences between studies, we imposed the following methodological restrictions for the inclusion criteria: 1) study design of interest was either cohort or case control study 2) the exposure of interest was allergic conditions (asthma, atopy, hay fever and "any allergy"); 3) the outcome was prostate cancer; 4) the study reported enough information to extract effect estimates and the corresponding 95% confidence intervals. In case of multiple publications, only the most recent or comprehensive one was considered eligible. Two authors (JKS and XHY) independently evaluated the eligibility of all retrieved studies and disagreements were resolved through discussion or consultation with a third author (LHG).
Data extraction. Data from the included studies were extracted and summarized independently by two of the authors using a pre-standardized data extraction form. Any disagreement was resolved by the senior author (ZWD). The following data were extracted from each study: first author, publication year, study design, country, sex, total number of cases and subjects for cohort studies, total number of cases and controls for case-control studies, assessment methods for allergic diseases, quantitative effect estimates (expressed as an odds ratio, hazard risk, relative risk, standardized mortality ratio, or standardized incidence ration) and 95%CI and variables adjusted for in the meta-analysis. The OR in one study 18 was not extracted, thus, we computed the crude risk estimates and corresponding CI. Two review authors (SJK and ZJG) independently extracted data from the selected studies. Any disagreements were resolved through discussion and consensus.
Quality assessment. Two review authors (SJK and ZJG) independently assessed the risk of bias using the Newcastle-Ottawa Scale (NOS), which consists of three factors: patient selection, comparability of the study groups, and assessment of outcome. A score of 0-9 (allocated as stars) was allocated to each study 50,51 . The studies achieving six or more stars were considered to be of high quality. Statistical analysis. Because the incidence of prostate cancer risk was low, the OR/HR was considered as approximation of RR, SMR was also considered as equivalent of SIR 52 . We computed the combined RR and SIR and corresponding 95%CI from the estimated reported in each study. The aggregated results and 95%CIs for effect sizes were calculated using inverse-variance weighted random-effects meta-analysis, which incorporated both within-study and between-study variability 53 . I 2 was used to assess heterogeneity across studies, with I 2 values of 0%, 25%, 50% and 75% representing no, low, moderate and high heterogeneity, respectively. Subgroup analysis was stratified by geographic region, study design, quality of NOS scale, body mass index (BMI) and whether the study adjusted for risk factors, including age, race, alcohol drinking and smoking. The sensitivity analysis was also conducted by removing one study at a time and the rest analyzed to determine whether an individual study affected the aggregate result or not. Small study bias, consistent with publication bias, was evaluated by statistical tests (Begg rank correlation test 54 and Egger's linear regression test 55 ), the visual examination of funnel plot when the number of included studies ≥ 10, and the results were considered to indicate publication bias when P < 0.10. All statistical analyses were conducted using Stata version 13.1 (Stata Corp., College Station, TX, USA).