Cancer Incidence and Mortality among Petroleum Industry Workers and Residents Living in Oil Producing Communities: A Systematic Review and Meta-Analysis

Petroleum extraction and refining are major sources of various occupational exposures and of air pollution and may therefore contribute to the global cancer burden. This systematic review and meta-analysis is aimed at evaluating the cancer risk in petroleum-exposed workers and in residents living near petroleum facilities. Relevant studies were identified and retrieved through PubMed and Web of Science databases. Summary effect size (ES) and 95% confidence intervals (CI) were analysed using random effect models, and heterogeneity across studies was assessed (I2). Overall, petroleum industry work was associated with an increased risk of mesothelioma (ES = 2.09, CI: 1.58–2.76), skin melanoma (ES = 1.34, CI: 1.06–1.70 multiple myeloma (ES =1.81, CI: 1.28–2.55), and cancers of the prostate (ES = 1.13, Cl: 1.05–1.22) and urinary bladder (ES = 1.25, CI: 1.09–1.43) and a decreased risk of cancers of the esophagus, stomach, colon, rectum, and pancreas. Offshore petroleum work was associated with an increased risk of lung cancer (ES = 1.20; 95% CI: 1.03–1.39) and leukemia (ES = 1.47; 95% CI: 1.12–1.92) in stratified analysis. Residential proximity to petroleum facilities was associated with childhood leukemia (ES = 1.90, CI: 1.34–2.70). Very few studies examined specific exposures among petroleum industry workers or residents living in oil producing communities. The present review warrants further studies on specific exposure levels and pathways among petroleum-exposed workers and residents living near petroleum facilities.


Introduction
The petroleum industry has been, and still is, an important pillar in many countries economy [1]. Petroleum (crude oil) is the origin of many complex mixtures such as petrol and diesel fuel and is, via chemical processes, used to produce plastics and other useful materials such as textiles [2], pesticides, cosmetics, paints, and insulating materials [3]. Workers in the petroleum industry are exposed to a variety of known or potentially harmful substances [4] and emissions from oil and gas extractions are among the major sources of air pollution in environments and communities where such facilities are situated and operate [5][6][7]. These potential harmful exposures include, among others, heavy metals and hydrocarbons such as benzene which is up to 4 g/L in crude oil [8]. Asbestos which is known to be the cause for mesothelioma and cancers of the lung, larynx, and ovary has been used extensively in petroleum industries [9]. Both benzene and asbestos are classified as Group 1 carcinogens by the International Agency for Research on Cancer (IARC) [4].
A recent systematic review and meta-analysis including studies published up to 2012 was based on data from 36 cohort studies assessed the meta-relative risk of 11 cancers among petroleum refinery workers [10]. Their results showed elevated meta-relative risks of mesothelioma, malignant skin melanoma, and acute lymphoid leukemia and no elevated risks of total leukemia, acute non-lymphocytic leukemia, chronic lymphocytic leukemia, chronic myeloid leukemia, multiple myeloma, non-Hodgkin's lymphoma (NHL), and lung and kidney cancers. The most recent review of residents near petroleum industry sites has also found elevated risks for leukemia (the review only focused on hematological malignancy) [11].
The aim of this present systematic review was to evaluate the epidemiological studies on cancer risks among petroleum workers and residents living near petroleum facilities through articles published between 1990 and 2019. Our review adds to the previous review [10] with four cohort, 14 case-control and two cross sectional studies on petroleum workers and with 13 additional cancer sites (i.e., cancers of the brain, urinary bladder, esophagus, stomach, colon, rectum, liver, pancreas, gall bladder, prostate, testis, breast and Hodgkin's lymphoma).

Eligibility Criteria
Our systematic review was conducted in accordance with the Preferred Reporting Items for Systemic Reviews and Meta-Analysis (PRISMA) check list of 2009 [12]. The present study included original articles in English published in peer reviewed journals from January 1990 to December 2019 (30 years period). The vast majority of relevant literature was published within these three decades, including early studies which were updated during that time. Very few studies were published only in the 1980s with working circumstances rather diluting the overall picture. The studies had to be cohort, case-control, or cross-sectional studies on cancer in petroleum workers or residents living close to operating facilities. Only studies that provided estimates of the Relative Risk (RR), such as Standard Mortality Ratios (SMR), Standard Incidence Ratios (SIR), Incidence Rate Ratio (IRR), Mortality Rate Ratio (MRR) or Odds Ratio (OR) with 95% confidence intervals (CI) were included. When multiple studies were identified from the same cohort/authors, we included them if they reported results for separate cancer sites or for different geographical locations, but if they studied the same cancer site in the same geographical location, the latest result with the longest follow-up or the largest study population was included.

Information Sources
Scientific papers were identified and retrieved through PubMed and Web of Science (WOS) databases, imported and automatically screened for duplicates in EndNote version X8.2, and later screened manually. The search strategies included a list of key words and Mesh terms (Supplementary Material). Additional relevant articles were identified through the exploration of lists of references. The initial search was performed in December 2019. Thereafter search alerts were placed on both databases using the search strategy to enable regular updates, with relevant articles until February 2021. We defined petroleum facilities as the upstream sector including exploration, development, and production of petroleum resources; the downstream sector including refining, marketing, and distribution of petroleum.

Assessment of Eligibility, Quality and Data Extraction
Eligibility screening of title and abstract were carried out independently by the first and second authors, and disagreements were resolved by the last author through discussions in line with the Cochrane handbook for systematic reviews [13]. Following removal of duplicates, the following data were extracted from the full-text articles: author and year of publication, country, continent, minimum employment time (petroleum industry workers), follow-up duration, exposed population, specific exposure, cancer site/type, number of cases and controls, as well as RR with their respective CIs. Information on study design (cohort, case-control and cross-sectional) was also extracted. All papers selected for inclusion were subjected to a rigorous appraisal for methodological quality using Joanna Briggs Institute critical appraisal (JBI) tools for cohort, case-control and cross-sectional studies [14]. The critical appraisal checklist has 11 criteria for cohort, 10 for case-control and 8 for cross-sectional studies, each question with "yes" score 1, "no" score 0 and "unclear" score 0.

Statistical Analyses
Random effects meta-analyses were conducted using the Stata ® , version 14.0. (Stata ® statistical software, College Station, TX, USA) "metan" command in order to (1) compute pooled effect size (ES) for mortality studies and incidence studies with 95% confidence intervals (CI) and (2) explore heterogeneity between studies, expressed as a percentage (I 2 ) [15]. Heterogeneity was considered as "no heterogeneity" when I 2 was 0%, "probably unimportant" when I 2 was 1% to 35%, "moderate" when I 2 was 36% to 55%, "substantial" when I 2 was 56% to 70% and "considerable" when I 2 was 71% to 100% [16]. Mortality and incidence studies were combined to one pooled ES for some digestive and hematological cancers ( Figure 2) because of similarity in incidence and mortality rates of the underlying cancer types and due to few studies. Likewise, the relative risks from incidence and mortality studies among residents were combined except for brain and skin melanoma. Potential publication bias was evaluated using the Egger's funnel plot and its associated p-value [17,18]. We also performed stratified analysis for the different categories of petroleum industries i.e., refinery, petroleum, petrochemical, refinery and petrochemical, and offshore. In all instances for this stratified analysis both mortality and incidence were combined. Meta regression (metareg in Stata ® ) was used to determine the impact of specific variables (country, continent, decade (1990-1999/2000-2009/2010-2019)) on the study ES. For heterogeneity and Egger's regression tests a p-value less than 0.05 were considered statistically significant.

Search Strategy Outcome
The search strategy for PubMed, WOS and other records yielded a total of 969 studies, whereof 273 papers were duplicates (see the PRISMA flow chart in Figure 1). Following duplicate removal, 569 non-eligible papers were removed based on titles and abstracts. One hundred and twenty-seven papers underwent full-text assessment for eligibility, and resulted in 57 eligible papers (41 cohort, 14 case-control and two cross-sectional studies) from which the data were extracted. Of those, 41 cohort and 10 case-control studies were included in the meta-analysis, while four case-control and two cross-sectional studies were not included in the meta-analysis because these studies examined specific cancer types or exposures for which there were only one study and therefore it was not possible to perform a meta-analysis; yet they are described and discussed separately within our review (Tables 1 and 2).

Study Characteristics
Tables 1 and 2 shows the list of 57 eligible papers with the majority of studies coming from North America (n = 23) and without a single study from Africa. In terms of publication year, most articles (n = 31) were published in the 2000s. Out of the 57 studies, 43 estimated cancer risk among workers, while 14 estimated cancer risk among residents.

Study Characteristics
Tables 1 and 2 shows the list of 57 eligible papers with the majority of studies coming from North America (n = 23) and without a single study from Africa. In terms of publication year, most articles (n = 31) were published in the 2000s. Out of the 57 studies, 43 estimated cancer risk among workers, while 14 estimated cancer risk among residents.

Assessment of Risk of Bias
The methodological quality of studies evaluated on petroleum workers resulted in an average score of 74% (Cohort studies) and 84% (Case control studies). The average score of papers on residents living near petroleum facilities was 41% (Cohort studies), 61% (Cross-sectional studies) and 70% (case-control studies). We did not exclude papers on residents based on their methodological quality due to very few papers being eligible for the final analysis in this group. The risk of bias or quality assessment grading for the different components of each study is shown in Tables S1-S5.

Review and Quantitative Analyses of Petroleum Workers
In this section we describe 36 papers with results from cohort studies among petroleum workers (summarized in Figure 2 (Figure 2). The meta-regression analysis did not reveal

Stratified/Subgroup Analysis
Among the five categories of petroleum industries evaluated, there was no substantial effect on the summary ES results of the main meta-analysis (

Review and Quantitative Analyses of Residents Living Near Petroleum Facilities
Herein, we describe five papers including results from cohort studies among residents living near petroleum plants by cancer site (summarized in Figure 3).

Solid Cancers
Lung cancer was reported in three studies [61,63,66], out of which one study conducted in Ecuador (Kelsh et   Brain cancer was investigated in two studies [61,63] from Ecuador which compared brain cancer risks in residents living near petroleum facilities with the national rate. Hurtig and Sebastian, reported RR 3.80 (95% CI: 0.24, 60.65) for females and RR 0.14 (95% CI: 0.01, 1.34) for males [61]. The study by Kelsh et al. reported an MRR of 0.31 (95% CI: 0.21, 0.47). The pooled ES was 0.39 (95% CI: 0.11, 1.43) based on 24 brain cancer deaths and two brain cancer cases, with a "moderate" heterogeneity across studies (Figures 3 and S2).

Review and Quantitative Analyses of Case-Control Studies of Workers and Residents Living Near Petroleum Facilities
Hereunder we describe 10 case-control studies among workers and residents living near petroleum facilities by cancer site (summarized in Figure 4).

Review and Quantitative Analyses of Case-Control Studies of Workers and Residents Living Near Petroleum Facilities
Hereunder we describe 10 case-control studies among workers and residents living near petroleum facilities by cancer site (summarized in Figure 4). In two case-control studies [56,58] on leukemia subtypes, we observed an elevated risk of AML with a pooled ES of 2.02 (95% CI: 0.66, 6.19) based on 70 cases and 656 controls in cumulative benzene exposure analysis on highest compared to lowest. There was a "moderate" but non-significant heterogeneity (I 2 = 42.6%, p = 0.19). The pooled ES in relation to maximum intensity was 2.17 (95% CI: 0.86, 5.48) with "probably unimportant" heterogeneity (I 2 = 23.3%, p = 0.25 ( Figure 4). Both studies (Rushton et al. = 1.39; 95% CI: 0.68,
For chronic myeloid leukemia (CML) Glass et al. [57] reported single increased ORs (cumulative benzene exposure = 2.20; 95% CI: 0.63, 7.68 and maximum benzene exposure intensity = 2.12; 95% CI: 0.77, 5.82) out of many cases in a multi-country nested case-control study including 28 cases and 122 matched controls, but the authors concluded that no convincing association was identified between CML and low exposure to benzene because there were no coherent patterns.

Solid Cancers
Kidney cancer in relation to cumulative benzene exposure was analysed in two nested case-control studies [55,60] with a pooled ES of 1.55 (95% CI: 0.84, 2.83, based on 130 cases and 487 controls) with no heterogeneity. This is similar to the pooled ES of the cohort studies where both mortality and incidence have OR above 1.00 Finkelstein [53] conducted a death certificate-based case-control study of 17 mesothelioma cases and 424 lung cancer cases compared to other blue-collar workers in order to investigate risks among refinery and petrochemical sector workers in Canada. Employment as a maintenance worker in the refinery and petrochemical company was associated with an OR of 24.5 (90% CI: 3.1, 102) for mesothelioma, and 1.73, 90% CI: 0.83, 3.6) for lung cancer.

Residents Living Near Petroleum Facilities Childhood Leukaemia and Other Hematopoietic Cancers
The two studies [70,74] yielded a significantly increased pooled ES of 1.90 (95% CI: 1. 34, 2.70) with no heterogeneity between studies, p = 1.00. Weng et al. conducted a casecontrol study including 405 childhood leukemia deaths showing an elevated risk of 1.90 (95% CI: 1.26, 2.87) for children living in the municipalities characterized by the highest proportion of the population employed in the petrochemical industry, when compared to the municipality with the lowest proportion. McKenzie et al. [74] conducted a registrybased case control study in rural Colorado in the USA, with study participants 0-24 years old and diagnosed with cancer between 2001 and 2013; of which were 87 children/young adults with ALL and 50 with NHL. The results showed that ALL cases 5-24 years old were more likely to live in the highest tertile of residential proximity to oil and gas development, with a significant test for trend (p-value = 0.035). No association was found for younger ALL cases or NHL cases.
Lyons et al. [68] compared the incidence of leukemia and lymphoma in young people (0-24 years) living in the surroundings (within 1.5 and 3 km) of a petrochemical plant with the expected numbers of cases calculated from the Welsh cancer registration rates. The authors concluded that the incidence of leukemia and lymphomas in young people was not significantly greater than expected.
Micheli et al. [73] investigated mortality from hematological malignancies (HM) among residents in relation to benzene pollution in three municipalities surrounding an Italian refinery, based on 177 HM-related deaths. The authors observed an excess risk of death in relation to proximity of residence to refinery for women and in a sub-group of study participants that were retired, homemakers, or unemployed, but not in men. The authors suggested that the positive association for females in this subgroup was due to them supposedly spending most of their time at home and having, therefore, the highest exposure due to residential proximity to the refinery [73].

Solid Cancers
Yu et al. [72] assessed the association between residential petrochemical exposure and brain tumor risk in a population-based case-control study including 143 brain tumor cases residing in an area of 657.1 km 2 hosting four petrochemical industrial complexes in Kaohsiung, China/Taiwan [72]. The results showed no association between residential exposure to petrochemicals and brain tumor risks (OR 1.00; 95% CI: 0.83, 1.12). On the contrary, a case-control study by [71], which included 340 brain cancer cases, found that those who lived in the municipalities characterized by the highest levels of petrochemical air pollution had a statistically significant higher risk of developing brain cancer ( Choi et al. [67] conducted a cross-sectional study of 63042 participants to evaluate prostate cancer risk in relation to oil spillage in Taean County. They reported elevated incidence rate ratio (IRR = 1.5; 95% CI: 0.8, 2.7) for the high exposed area compared to the low exposed area (IRR= 0.8; 95% CI: 0.5, 1.4).

Discussion
Our systematic review and meta-analyses included 43 publications on cancer risks among petroleum workers and 14 among residents living near petroleum industry sites. Among petroleum workers, six cancer types were associated with elevated risks, namely mesothelioma in both incidence and mortality studies, as well as multiple myeloma, skin melanoma, prostate cancer and urinary bladder cancer in incidence (but not mortality) studies. In stratified analysis, lung cancer and leukemia were associated with an elevated risk among offshore workers. Cancers of the liver, esophagus, colon, pancreas, stomach, and rectum showed inverse associations among petroleum workers, which were stronger in the mortality studies than in the incidence studies. We found significant elevated risks of childhood leukemia for residents living near petroleum facilities.
The meta-analysis of the nested case-control studies of hematopoietic cancer among petroleum workers in relation to benzene exposure showed some non-significant elevated risk estimates for AML, CLL and leukemia overall but no conclusive exposure-response relationship. Although the present meta-analysis for cohort studies did not suggest elevated risk for kidney cancer, a non-significant increased risk of kidney cancer was observed in the meta-analysis for case-control studies. Among residents living near petroleum facilities there were non-significant elevated results for brain cancer in adults in both cohort (mortality but not incidence) and case control studies, one positive study of urinary bladder cancer, and one positive study of prostate cancer.
The increased risks of mesothelioma, skin melanoma, and multiple myeloma observed among petroleum workers in our review confirm the findings of a previous review [75]. The vast majority of mesothelioma occurs due to asbestos exposure, with even relatively low exposure levels increasing risk [76,77]. Asbestos is widely used for its non-corrosive ability in combustion petroleum pipes in petroleum refineries [78], although there are no available data on the concentration of asbestos fibers in petroleum facilities. Workers who are engaged in maintenance in petroleum companies in particular experienced an increased risk of mesothelioma, probably due to more asbestos exposure [79].
Skin melanoma is also associated with multiple risk factors which workers can be exposed to including occupational chemical exposures (e.g., arsenic) [80] and ultraviolet radiation [81] in outdoor workers, especially the latter as the most likely explanation for the increased risk in petroleum industry workers. Furthermore, the increased risk of multiple myeloma among petroleum employees in the present meta-analysis is supported by additional reviews [10,75] and the epidemiological evidence linking multiple myeloma with hydrocarbon exposure [82][83][84].
However, the increased risk of prostate cancer we observed among petroleum workers was not observed in the review by Wong and Raabe [75], but has already been reported in individual studies, and our systematic review included studies that were published after Wong and Raabe review. The risk of prostate cancer may be accelerated by exposure to cadmium, which is a composite of crude oil classified as carcinogenic to humans [85]. As the increase in risk was observed only for incidence but not mortality, it is unclear how much a major determinant of prostate cancer incidence, namely PSA screening, plays a role, given the medical surveillance workforces normally receive.
The increased risk of incident urinary bladder cancer we observed among petroleum workers was not examined in the review of Schnatter et al. [10]. Instead, Wong and Raabe reported decreased urinary bladder mortality, which was confirmed by our own review. The observed increase in risk could occur as a result of exposure to chemicals such as benzidine which is known to cause bladder cancer [4]. However, this does not explain the reduced risk of urinary bladder cancer mortality. Further studies evaluating disease stage at diagnosis and access to care may help clarify this unexplained discrepancy.
In our review, skin melanoma, prostate cancer and urinary bladder cancer did not show any association in the mortality-based studies. All three cancers have a good prognosis especially in earlier stages and may be detected early through regular medical check-ups; this may explain why we observe an increased incidence of these cancers but not an increased risk of mortality [86][87][88].
The association in childhood leukemia we observed is in agreement with previous reviews [89,90]. Childhood leukemia may occur due to chromosomal translocation of a pre-leukemic clone, in which the genes are rearranged and may be triggered by exposure to environmental factors such as benzene [89]. In adults, a non-significant increased risk of leukemia was observed in the meta-analysis for cohort and case-control studies, but was significantly elevated in stratified analysis of offshore workers when all leukemia subtypes were combined.
The present meta-analysis showed that mortality was reduced for cancers of the lung, esophagus, stomach, colon, rectum, and pancreas, which are known to be associated with life-style factors such as tobacco, alcohol, and diet [91,92]. This suggests that workers in this industry smoke and drink less, eat more healthily and do more physical activity than the general population. Although we observed a significant elevation in stratified analysis among offshore category, as lung cancer is also known to be related to many occupational exposures, lung cancer did not show increased risk or mortality among workers or residents living in proximity to petroleum facilities in the main analysis, even if there was considerable heterogeneity across individual studies. These negative findings are consistent with previous reviews [10,75].
Strengths of this review include that many of the studies had long term follow up in cohort studies, and that both mortality and incidence studies among petroleum workers were clearly separated in the main analyses of our review when compared to previous reviews. Other strengths include stratified analysis by sub-sites of the petroleum industry based on the extracted information. Weaknesses include crude exposure assessment in some studies, e.g., ever employment in this industry, while the exposure variability may be considerable between jobs/tasks and thereby dilute any potential risk for highly exposed workers. Another caveat is the lack of analyses by duration or level of exposure in some of the studies. Insufficient exposure contrasts may have caused bias among petroleum workers, due to non-stratified classes of workers in some of the studies. Furthermore, the cohort studies reported results on multiple cancer sites, for which some analyses were severely under-powered for the detection of any potential risk. Finally, it is relatively common to see lower mortality in occupational cohorts because unhealthy people are usually excluded from employment, which makes the overall death rates lower than those of the general population [93].
The above mentioned limitations could be reduced by improving the exposure assessment, especially in areas with substantial occupational or residential exposures like low and middle income countries (LMICs) with high crude oil production capacity and artisanal refining activities. Unfortunately, there were no studies conducted on petroleum workers and residents living in oil producing communities in Africa. Meanwhile Angola and Nigeria are among the world's top oil producers. There have been frequent oil spills contaminating landscape and polluting soil and ground water, as well as constant emission of hydrocarbons and other toxic chemicals in the Niger Delta region of Nigeria. In the same region houses are built around petroleum oil wells with pipelines running across the various communities and crude oil is sometimes used for dermal care and ingested for childhood convulsion treatments among other illnesses [94]. The United Nations Environment Programme (UNEP) reported that benzene level was 900 times above WHO guidelines in some spots in Ogoniland, Niger Delta region of Nigeria, as a result of petroleum pollution [95]. Similarly, major oil pollution has also been reported in some other petroleum producing regions of the world like the Middle East and Russian Federation where the oil industry is one key [96,97]. Therefore, well designed epidemiological studies on workers and residents are warranted in these regions.

Conclusions
In conclusion, living close to petroleum facilities was associated with increased risk of childhood leukemia, while petroleum industry work was associated with an increased risk of mesothelioma, skin melanoma, multiple myeloma, and cancers of the prostate, urinary bladder, and a decreased risk of cancers of the esophagus, stomach, colon, rectum, and pancreas. Offshore petroleum work was associated with an increased risk of lung cancer and leukemia in stratified analysis. Many of the associations, however, appear to be due to factors other than those directly emerging from petroleum production. This applies to asbestos for mesothelioma, which is a rare cancer, so that the number of cases attributable to this risk were small compared to the working population. This also applies to ultraviolet radiation from outdoor work for skin melanoma, or from healthier lifestyles in petroleum workers for cancers of the colon, rectum, esophagus, stomach and pancreas. Prostate cancer incidence studies may have been affected by medical surveillance. For urinary bladder cancer, studies showed inconsistent results, which may not be surprising given that bladder cancer is positively associated with several workplace chemicals (some occurring in the petroleum industry) but also with smoking (apparently reduced in petroleum workers given the reduced risks of other smoking-related cancers), and therefore inclusion criteria of individual studies and their possibilities to adjust for confounders may have had a significant impact on the study results.
The overall evidence remains weak, particularly for the residential studies. Improved exposure assessment is needed in further studies to describe exposure pathways of petroleum and its closest derivatives, e.g., benzene, in order to identify the drivers of the observed cancer risks or identify association which may have been missed using the crude industry-based exposure approach. In particular, there is a need for targeted studies in under-researched areas of high petroleum production with presumably higher exposures. An international consortium guiding new generation studies in Africa, the Middle East and Asia, to harmonize study protocols and exposure assessments, may be the most promising way forward.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/ijerph18084343/s1, Figure S1: Forest and funnel plot of studies evaluating cancer incidence and mortality among petroleum, Figure S2: Forest and funnel plot of studies evaluating cancer incidence and mortality among residents of oil producing communities, Figure S3: Forest plot of nested case-control studies evaluating cancer among petroleum industry workers and case-control studies evaluating cancer risk among residents living in proximity to petroleum facilities, Tables S1-S5: JBI quality appraisal checklist for studies evaluating cancer incidence and mortality among petroleum industry workers and residents of oil producing communities. Funding: The work reported was undertaken during the tenure of a Postdoctoral Fellowship from the International Agency for Research on Cancer, partially supported by the European Commission FP7 Marie Curie Actions-People-Co-funding of regional, national and international programs (COFUND).

Informed Consent Statement: Not applicable.
Data Availability Statement: Data is contained within the article and supplementary material.