Introduction

Cement is produced by quarrying and crushing calcareous materials (mainly limestone, chalk, and clay) and additives; the raw mixture is heated to high temperature (around 1440 °C) in a rotary kiln with use of different combustion sources (coke, fuel, gas, others). As a result working in a cement plant involves potential exposure to cement dust, particulate matters and gaseous pollutants, polycyclic aromatic compounds, silicates, aluminates, hexavalent chromium and other metals such as nickel, and cobalt, as well as, in the past and in some circumstances, asbestos because of protection from high-temperature processes (EPA 1994). Several of these substances are considered definite or possible carcinogens (Cogliano et al. 2011).

Some epidemiological studies investigated the association between employment in cement production and increased risk of overall and cancer-specific mortality, but a systematic review and meta-analysis of these data are lacking. A recent review addressed the issue of cancer risk from cement exposure (Cohen et al. 2014), without separating cement production workers from construction workers exposed to concrete dust and other agents not present in cement manufacture This review included a total of 26 studies, occupational cohort studies or case–control studies, and did not provide convincing evidence for increased risk of any cancers. Meta-relative risk of colorectal cancer incidence was elevated, without a corresponding increase in mortality.

We conducted a meta-analysis to clarify whether a hazard is present in this industry, and, if a hazard is detected, to quantify the risk in terms of mortality from all causes, cancer, cardiovascular, and respiratory diseases.

Methods

We followed established criteria for performing and reporting meta-analyses of observational studies according to the MOOSE (Stroup et al. 2000) and the PRISMA guidelines (Moher et al. 2009). Studies were identified through a systematic review of the scientific literature until February 2016 using the electronic PubMed and Scopus database. No restrictions were imposed. We used the search terms: (mortality OR cancer OR death) AND (cement workers). In addition, we manually searched for further articles the reference lists of recent reviews and other retrieved publications. The search was conducted independently by two of the authors and the results were compared and combined.

We included in the meta-analyses epidemiological observational studies based on original prospective cohort studies of cement workers that provided a measure of association, such as standardized mortality ratio (SMR), standardized incidence ratio (SIR) or proportionate mortality ratio (PMR) with the corresponding 95 % confidence intervals (CIs), for all-cause mortality, cardiovascular diseases, respiratory diseases, and all cancers, lung cancer and stomach cancer. We restricted the review to studies published as full-length papers in English. We also included case–control studies involving workers in cement production. The measure of association with case–control studies was the odds ratio (OR).

We reviewed all titles and abstracts to identify potentially relevant publications. For each relevant article, we abstracted details on study design, country, cohort size, duration of follow-up, outcome, the measure(s) of association and the corresponding CI, and the variables adjusted for in the analysis, if any. When we founded multiple publications based on the same study population, we included only the most recent or informative one. Selection of relevant publications and abstraction of data was conducted independently by two of the authors. We used Newcastle–Ottawa Scale (Wells et al. 2011) to score the quality of each study.

We performed separate meta-analyses for mortality from all causes, cardiovascular diseases, respiratory diseases and overall cancer, lung cancer, and stomach cancer deriving all summary risk estimates using random-effect models (Der Simonian and Laird 1986). For the two outcomes with the largest number of risk estimates (lung cancer and stomach cancer), we performed a cumulative meta-analysis, which provided summary risk estimates each year a new study was published. The meta-analyses were conducted using a random-effects model as describe above.

We evaluated the heterogeneity among studies using the χ 2 test (Greenland and Longnecker 1992) and quantified heterogeneity using the I 2 statistic which represents the percentage of the total variation across studies that is attributable to heterogeneity rather than chance (Higgins and Thompson 2002). We evaluated the presence of publication bias through visualization of funnel plots (Thornton and Lee 2000) and the test proposed by Egger et al. (1997). We conducted all analyses using STATA v. 13.

Results

The PubMed search resulted in 188 publications. At the end of the selection process, we retained twelve publications referring to eleven study populations (McDowall 1984; Amandus 1986; Jakobsson et al. 1990; Vestbo et al. 1991; Siemiatycki 1991; Jakobsson et al. 1993; Smailyte et al. 2004; Koh et al. 2011; Dab et al. 2011; Giordano et al. 2012; Rachiotis et al. 2012; Koh et al. 2013) that reported risk estimates for the outcomes of interest in relation to employment in cement production (Fig. 1).

Fig. 1
figure 1

Strategy for identification of relevant publications

Full size image

The key characteristics of the studies included in the review and meta-analysis are summarized in Table 1. The eight cohort studies included in our meta-analysis were conducted primarily in Europe (N = 6), with one study each performed in the USA and Korea. One proportionate mortality study was conducted in Greece. The two case–control studies were conducted in Canada and in Europe (Table 1).

Table 1 Characteristics and results of studies included in the meta-analysis
Full size table
Table 2 New Castle–Ottawa quality assessment scale
Full size table

Three studies included both male and female workers (Jakobsson et al. 1993; Koh et al. 2011; Dab et al. 2011), while the other studies were restricted to male workers. Results of all studies were adjusted for age and sex. Results of two studies were also adjusted for tobacco smoking (Vestbo et al. 1991; Siemiatycki 1991), and the authors of one study performed an ancillary descriptive study to assess the prevalence of tobacco smoking (Dab et al. 2011).

We assigned a quality score to each study, on the basis of Newcastle–Ottawa Scale (Wells et al. 2011), which ranged between 4 and 7. We decided not to exclude any of the studies from the analyses for low-quality score (Table 2).

Table 3 Results of the meta-analysis
Full size table

Cohort studies

McDowall (1984) studied 607 cement male workers employed in 1939 in England and Wales and followed up from 1948 to 1981; of these 607 men, 419 had died by December 31, 1981. Cement workers were identified through National Health Service Central Register (NHSCR). For each cement worker, information was extracted concerning job description, date of birth, dates of emigration; if relevant, date and cause of death were obtained by Office of Population Censuses and Surveys—causes of death were coded according to the 8th revision of the International Classification of Disease (ICD). Cement workers were divided in four occupational groups on the basis of their job titles: (1) laboratory workers, clerks, timekeepers, storekeepers, canteen staff, and other staff; (2) maintenance workers, craftsmen and machinery operators, and crane and lorry drivers; (3) laborers specified as packing or loading; and (4) other laborers and inadequately described occupations. Higher exposure of cement dust was supposed for workers in groups (2) and (3). Expected numbers of deaths were calculated on mortality rates for England and Wales. The overall mortality resulted slightly lower than expected (SMR 0.95; 95 % CI 0.86–1.05), as it was mortality from respiratory diseases (SMR 0.86; 95 % CI 0.66–1.11). Overall cancer mortality was close to expected (SMR 1.03; 95 % CI 0.83–1.26), and mortality for lung cancer was decreased (SMR 0.85; 95 % CI 0.56–1.23), while mortality from stomach cancer was increased (SMR 1.75; 95 % CI 1.10–2.65). Stomach cancer mortality was increased in three of the four occupational groups, maintenance workers and craftsmen and for laborers specified as packing or loading; this excess of mortality did not appear to be explained by differences in socioeconomic status. The study also showed some evidence of an increased risk from rectal cancer (SMR 1.86, 95 % CI 0.89–3.42), while mortality from other cancers was not altered.

Amandus (1986) investigated a white male cohort of 5292 cement and quarry workers, of whom 4231 were included in the analyses. Workers were selected for the analysis if employed for at least 4 years between 1950 and 1960 in one of six cement plants in the USA, with complete work histories and demographic information. Workers were followed for mortality until 1980. Plants were selected randomly by geographic region and were supposed to have started operation before 1950. Information about vital status of each cement worker was obtained from United States Social Security Administration, postmasters, and inquiries from friends and relatives. Death certificates were collected, and the underlying cause of death was coded according to the 8th revision of the ICD. The expected number of deaths was computed using age- and calendar year-specific death rates from the US white male population. Tenure was used as a surrogate estimate of exposure to cement dust. The study resulted in a SMR of stomach cancer equal to 1.35 (95 % CI 0.89–1.96) in the overall cohort; results by job title did not provide evidence of an increased risk in any particular subgroup of workers. No gradient was found with tenure group. No results were reported for other causes of death in this study.

Vestbo et al. (1991) studied a cohort of 546 male cement workers in Denmark employed for at least 1 year before 1974 and followed up for cancer incidence from 1974 to 1985. The referent population was a group of 858 randomly sampled men of similar age and area of residence, with job title listed as blue-collar worker, white-collar worker, or farmer. In 1974, all men gave information regarding lifelong occupational, smoking, and residential histories and underwent a lung function test. Information on incidence of cancer was obtained from the National Cancer Registry, cancer cases were coded according to the 7th revision of ICD, and expected cases were obtained from national rates. Information on deaths was obtained from the Danish National Board of Health. Results were reported for 1–20 years of employment and for more than 20 years of employment. No excess of incidence from all cancer was founded among cement workers [SIR 0.71 (95 % CI 0.47–1.08) for 1–20 years and 1.04 (95 % CI 0.75–1.44) more than 20 years of employment]. No increased risk of stomach cancer was noticed (1 observed vs. 2.8 expected deaths). The SIR for respiratory cancer, including laryngeal and lung cancer, was 1.14 (95 % CI 0.59–2.19) for 1–20 years of employment and 1.52 (95 % CI 0.90–2.57) for more than 20 years of employment. In an internal analysis restricted to 1124 cement workers and 737 other workers, employment in cement production was not associated with risk of respiratory cancer after adjustment for tobacco smoking.

Jakobsson et al. (1993) investigated mortality and cancer incidence in a cohort of 2400 male workers with at least 1 year of activity before 1952 in two Swedish cement plants. They were followed until 1986. For each worker, job histories were extracted from the plants records. There were 1205 men in plant A with two lost to follow-up (0.2 %), and 1195 men in plant B, of whom seven (0.6 %) were lost. Information on cause of death was obtained from Statistics Sweden that coded the death certificates according to ICD-8, and information on cancer incidence, coded according to ICD-7, was obtained from the National Swedish Cancer Registry, with histological or cytological examinations performed in 89 % of cases in plant A and 93 % of cases in plant B. Expected mortality and incidence were calculated using death rates for males in Sweden. For each worker, job histories were extracted from the plant records. Job titles were classified as no (white collar), low (e.g., workers in supply depots), medium (e.g., repair and maintenance workers), or high exposure (production and packing workers). Mortality from all causes (SMR 0.94; 95 % CI 0.86–1.03), all cancers (SMR 0.83; 95 % CI 0.68–1.02), lung cancer (SMR 0.53; 95 % CI 0.27–0.95), stomach cancer (SMR 0.85; 95 % CI 0.45–1.45), cardiovascular diseases (SMR 0.98; 95 % CI 0.86–1.11), and respiratory diseases (SMR 0.8; 95 % CI 0.55–1.30) resulted lower or equal to expected, and there was no trend with duration of employment. Among the other results, there was an increased incidence of colorectal cancer among workers with 15 or more years since the start of employment (SIR 1.61; 95 % CI 1.10–2.29), mainly due to the results for tumors in the right part of the colon (SIR 2.73; 95 % CI 1.41–4.77).

Smailyte et al. (2004) studied mortality and cancer incidence of 2498 cement workers employed for at least 1 year in a cement factory in Lithuania from 1956 to 2000 and followed up to December 2000 for cancer incidence. Those working in administration were not included. Information on each worker was obtained from the factory records including name, date of birth, departments, and the beginning and ending dates of all job periods. Information on vital status, emigration, and date of death was obtained from the local migration office and from the Bureau of Addresses of the Central Population Register. The cancer cases in the study population were identified from the Lithuanian Cancer Registry, and copies of death certificates were obtained from the Archive Department of the Local Municipality and the Registry Department of the Lithuanian National History Archive and coded at the Cancer Registry. The expected numbers of deaths and incidence cases were based on national incidence and mortality rates. Exposure data by different departments were available since 1975 as dust mass measurements performed four times per year, with highest concentration found in packing (15.0–22.2 mg/m3) and calcining department (37.6 mg/m3 in 1975, less than 20.0 in later years). Cumulative exposure estimates was calculated from the annual mean dust measurements. In this study, 35 workers were lost to follow-up. No excess was observed for overall mortality (SMR 1.0; 95 % CI 0.9–1.1 in men, 0.8; 95 % CI 0.6–1.0 in women), mortality from cardiovascular diseases (SMR 1.1; 95 % CI 0.9–1.2 in men; SMR 0.9; 95 % CI 0.6–1.2 in women), respiratory diseases (SMR 0.8; 95 % CI 0.5–1.3 in men; SMR1.0; 95 % CI 0.3–3.0 in women), stomach cancer (SMR 0.7; 95 % CI 0.4–1.4 in men; SMR 1.0; 95 % CI 0.3–3.1 in women), and colorectal cancer (SMR 1.1; 95 % CI 0.5–2.3 in men; SMR 0.4; 95 % CI 0.1–2.6 in women). In men and women combined, the SMR for all malignant neoplasms was 1.1 (95 % CI 1.0–1.4) and that from lung cancer was 1.4 (95 % CI 1.0–1.9). An increased risk of lung cancer among male workers was also reported in the incidence analysis (SIR 1.5; 95 % CI 1.1–2.1), and there was an association between lung cancer incidence and duration of employment and estimated cumulative dust exposure. No significant excess of risk from lung cancer incidence was found among women workers (SMR 1.4; 95 % 0.3–5.5; SIR 1.7; 95 % CI 0.6–5.4). Incidence data showed no significant excess of risk also for all malignant neoplasms (SIR 1.2; 95 % CI 1.0–1.4 for males and SIR 0.8; 95 % CI 0.6–1.1 for females). No individual smoking data were available in this study.

Koh et al.(2011) conducted a retrospective cohort study that investigated male workers in six Portland cement factories in South Korea, classified into five groups by job (quarry/production/maintenance/laboratory/office), who were followed up for cancer mortality and incidence from 1992 to 2007 and from 1997 to 2005, respectively. Lists of worker information were assembled for each factory and included employment date, retirement date, job type, and resident registration number; last job was used as a proxy for exposure. Dates and causes of death were obtained from Statistics Korea and coded according to the 10th revision of ICD. Cancer incident cases were identified from the Central Registry of Cancer in Korea and classified according to the US NIOSH occupational cancer classification. Korean male cancer mortality rates in 1994, 1999, 2004, and 2007, and Korean male cancer incidence rate in 2003 were used as reference rates. There were 5,146 subjects included in the mortality follow-up and 5596 in the cancer incidence follow-up. All-cancer mortality rates for all workers were lower than the expected values (SMR 0.83; 95 % CI 0.68–1.01), and no significant excess of cancer mortality was observed in any of five job groups. The SMR for lung cancer in the whole cohort was 1.05 (95 % CI 0.68–1.57); no significant excess was found in any of the five job groups. An elevated stomach cancer incidence was reported in production workers (SIR 1.56; 95 % CI 1.02–2.26), but not in other job groups (overall SIR 1.29; 95 % CI 0.97–1.69), and there was no corresponding increase in stomach cancer mortality either among production workers (SMR 0.52; 95 % CI 0.17–1.21) or in all subjects (SMR 0.83; 95 % CI 0.50–1.29). However, this study was carried out without quantitative exposure assessment; notably, in a successive publication, the authors selected the two factory among the six included in the original publication, because of the availability of worker job histories from the start of production and environmental monitoring records, used for calculating cumulative exposures and categorizing workers into high and low dust exposure groups (above and below 11.61 mg/m3-year, respectively) (Koh et al. 2013), examining the dose-related association using cumulative exposure. The resulting SIR for all cancers combined was increased (SIR 1.35, 95 % CI 1.01–1.78), but the SIR for stomach cancer was increased only in the group at high dust exposure (SIR 2.18, 95 % 1.19–3.65). A subgroup finding of this study was the increased rectal cancer risk among all workers (SIR 3.05; 95 % CI 1.32–3.65), with similar risk in the high- and the low-exposure groups, that suggests the excess of risk may not related to dust exposure. In a subsequent analysis of a subgroup of the same workers, Koh et al. (2013) examined the dose-related association using cumulative exposure, based on recent measurements, and found an increased stomach cancer risk. However, the small number of cancer cases and short observation period are limitations of this study.

Dab et al. (2011) studied a cohort of 9118 workers employed for at least 1 year in one of the four main cement manufacturing companies in France between 1990 and 2005. For each worker name, birth date and place, gender, hiring and end of employment dates, activity status on 2005 and job titles occupied since the hiring were extracted from the company administrative files. Information on vital status, and date and place of death was obtained from the National Institute of Statistics, and information on cause of death from the National Institute of Health and Medical Research and classified according to the 9th version of ICD for the period 1990–1999 and the 10th version after 1999. Job titles were classified into seven job categories: quarry, yard, production, shipping, maintenance, laboratory and administration. For an ancillary descriptive study, occupational physicians randomly selected 268 medical files of male cement workers active in 2005 to study smoking prevalence. These authors conducted a SMR analysis based on national mortality rates and explored the combined effect of job titles and duration of employment. There was a deficit in mortality from all causes (SMR 0.68; 95 % CI 0.61–0.74), all cancers (SMR 0.80; 95 % CI 0.69–0.92), and cardiovascular diseases (SMR 0.68, 95 % CI 0.53–0.85). Stomach cancer mortality (SMR 0.38; 95 % CI 0.08–1.26) was lower than expected, and mortality from lung cancer was close to expected (SMR 0.95; 95 % CI 0.68–1.21). Among the seven job category, all the SMRs for mortality from all causes were below unity. Workers in shipping, quarry, laboratory, and yard had a SMR in the range 0.87–0.81, without reaching the statistical significance in any category. No occupational exposures or individual exposure on tobacco and alcohol data were available to be used in this epidemiologic study. In this study, the prevalence of active smoking among male cement workers in activity was estimated to be 32.8 % (CI 30.0–35.7), 21.6 % (CI 15.8–27.3) for administrative employees, 36.5 % (CI 33.0–39.9) for workers involved in the production process and 28 % (CI 25.9–30.1) for laboratory workers being at an intermediary pattern.

Giordano et al. (2012) investigated a cohort of 748 male workers of a cement plant in Rome, Italy, employed during the period 1940–2006. The data were extracted from the individual employment records. Work history for each subject was reconstructed from the personnel registry of the factory. The vital status was ascertained through the population registries of the municipalities of residence. Causes of deaths, available only from 1969 (for the period 1956, when first death occurred, to 1969 it was only possible to calculate the overall SMR), were obtained from death certificate either from the local health units or from the municipal death registries, and they were classified according to 8th, 9th, and 10th revisions of ICD. For 9 cases, it was not possible to determine the cause of death. Expected deaths were calculated based on the mortality rates of the Latium region, as well as of the town where the plant is located (available only for 1982–2006). Authors classified at “high-exposure” unskilled workers, involved in cement production, with at least 10 years of employment, and at “low-exposure” all other workers. Overall mortality (SMR 0.87; 95 % CI 0.77–0.98) and mortality from all neoplasms (SMR 0.64; 95 % CI 0.48–0.82) were lower than expected (regional comparison). Mortality from cancer of the respiratory system was also reduced (SMR 0.56, 95 % CI 0.32–0.89), except for a subgroup of 39 subjects with previous employment in an asbestos–cement plant (SMR 2.86, 95 % 1.30–5.42). No excess mortality from other specific cancers or other non-neoplastic diseases was found. However, the SMR for non-malignant respiratory diseases was elevated (SMR 1.41, 95 % CI 0.95–2.03), but not associated with estimated exposure.

Proportionate mortality study

Rachiotis et al. (2012) conducted a proportionate mortality analysis of deaths recorded in the Greek Cement Workers Compensation Scheme, where all workers in cement industry were registered, covering the period 1969–1998. The information collected included sex, age at death, and starting and ending dates of all employments. Information on causes of death was obtained from death certificates. Causes of death were classified according to the 9th revision of ICD. The expected number of deaths was calculated based on 5-year age groups and calendar periods, on mortality rates of the Greek population obtained by the National Statistics Service. Three categories of exposure to cement dust were considered: high (cement production workers), medium (maintenance workers), and low (office workers). Data from 1157 death certificates fulfilled the inclusion criteria, of whom 632 among cement production workers, 336 among maintenance workers, and 189 among office workers. The PMR for lung cancer was increased among cement production workers (PMR 2.05 95 % CI 1.65–2.52). These workers presented also increased PMRs for all cancers and ischemic heart disease (PMR 1.37; 95 % CI 1.12–1.64 and PMR 1.3; 95 % CI 1.13–1.49, respectively). The PMR for lung cancer was also increased among maintenance workers (PMR 1.67; 95 % CI 1.15–2.34).

Case–control studies

Jakobsson et al. (1990) identified all male cases of cancers of the esophagus, stomach colon, and rectum that occurred in the period 1922–1988 in two parishes surrounding a cement plant in Sweden, using death records from these parishes and from the National and Regional Cancer Registries. All cement workers among cases and referents were identified from parish books; these cases were matched with the cement plant files or interviewed on their employment. The studies included a total of 60 cases, including 18 verified cases of esophageal or stomach cancer and 25 verified cases of colorectal cancer, and 17 uncertain cases. For each case, up to four deceased controls were identified from the population of the parishes, matched on age and year of death, for a total of 234 controls. The OR for the employment in the cement industry was 3.2 (95 % CI 1.1–9.4) for verified colorectal cancers, with some indication of a duration–response relationship. For 18 verified cases with an esophageal or stomach cancer, the OR among male workers of cement plant was 3.2 (95 % CI 1.0–11), without a relationship with duration of employment.

Siemiatycki (1991) conducted a multi-site population-based case–control study that investigated the cancer risk from occupational exposures. The study was conducted from 1979 to 1986 and included men aged between 35 and 70 years, resident in the Montreal, Canada, area and diagnosed with any one of 23 cancer types considered. Population controls were randomly selected from the electoral list for the Province of Quebec and matched to cancer cases by age and electoral district. There were 4576 eligible patients with cancer between 1979 and 1985, and 3730 of these (82 %) were successfully interviewed, together with 533 population controls. Level and frequency of exposure to 290 agents, including Portland cement dust, were assessed by a team of chemists and industrial hygienists who reviewed the detailed occupational histories collected through interview. The overall prevalence of exposure to cement dust was 7 %; the main occupations which entailed this exposure were construction laborers, stone masons, and carpenters. ORs for any exposure were 1.0 (95 % CI 0.7–1.3) for lung cancer and 1.0 (95 % CI 0.6–1.6) for stomach cancer (95 % CI was calculated based on the 90 % CI reported in the publication). Corresponding ORs for substantial exposure, defined as exposure at high level for at least 5 years, were 1.3 (95 % CI 0.9–2.0) and 1.1 (95 % CI 0.6–2.0) for stomach cancer. Among the other cancer sites for which results were reported, there the OR of rectal cancer was substantial exposure was 1.6 (95 % CI 0.8–3.1), while results for the remaining cancer sites (cancers of the esophagus, colon, pancreas, prostate, bladder, and kidney, non-Hodgkin lymphoma, and malignant melanoma) were unremarkable. Although this study addressed exposure to Portland cement dust, a large proportion of workers classified as exposed to this agent were employed in the construction industry, thus limiting the comparability of results with the other studies included in the review.

Meta-analysis

The results of the meta-analysis are listed in Table 3. The meta-analysis of studies on all-cause mortality resulted in a RR of 0.89 (95 % CI 0.76–1.01) for mortality from all causes (Table 3; Fig. 2a), with high heterogeneity between studies. The meta-analysis summary RR for risk from all neoplasms did not show an excess (RR of all studies 0.94, 95 % CI 0.80–1.08; RR of studies based on mortality 0.93, 95 % CI 0.77–1.08; RR of studies based on incidence 1.05, 95 % CI 0.96–1.14; Table 3; Fig. 2b), with high heterogeneity between studies in the overall analysis and in that based on cancer mortality.

Fig. 2
figure 2figure 2

a Forest plot of results of meta-analysis on risk of all-cause mortality. b Forest plot of results of meta-analysis on risk of all cancers. c Forest plot of results of meta-analysis on risk of lung cancer. d Forest plot of results of meta-analysis on risk of stomach cancer

Full size image

An increase in mortality from lung cancer was showed in two studies (Smailyte et al. 2004, Rachiotis et al. 2012), and an excess of incidence of lung cancer was observed in two studies (Vestbo et al. 1991; Smailyte et al. 2004), but the risk from this disease was lower than expected in the other cohorts; the meta-RR was 1.07 (95 % CI 0.79–1.35; Table 3, Fig. 2c). In the sensitivity analysis, the RR of studies based on mortality was 1.03 (95 % CI 0.68–1.39), while that of studies based on cancer incidence data showed a nonsignificant excess risk (RR 1.24, 95 % CI 0.98–1.50). Heterogeneity of results was high in studies based on mortality data.

Two cohort studies (McDowall 1984; Amandus 1986) and one case–control study (Jakobsson et al. 1993) showed an association with mortality from stomach cancer, while in Korean cohort (Koh et al. 2011) in front of an excess of incidence of stomach cancer did not correspond an increase in mortality from this disease. This association was not confirmed in the other studies examined and meta-analysis for all selected studies resulted in a RR of 0.93 (CI 95 % 0.70–1.17; Table 3; Fig. 2d), with limited inter-study heterogeneity. The sensitivity analysis based on type of outcome revealed no association for both studies based on mortality and incidence data.

The meta-analyses on cardiovascular and respiratory disease mortality resulted in RRs equal to 0.99 (95 % CI 0.85–1.13) and 0.92 (95 % CI 0.72–1.11), respectively (Table 3), with high heterogeneity for the former outcome.

The cumulative meta-analysis for lung cancer did not provide any suggestion of a trend in the summary risk estimates as the evidence accumulated over time (results not shown in detail). The cumulative meta-analysis of stomach cancer suggested that positive findings of early studies were not confirmed by subsequent results (Fig. 3).

Fig. 3
figure 3

Cumulative meta-analysis—stomach cancer

Full size image

With the exception of lung cancer, there was no evidence of publication bias, although the relatively small number of studies limited the power of the tests (Table 3).

Discussion

In this systematic review, we identified 11 studies that specifically investigated mortality or cancer incidence in relation to cement dust exposure among cement workers. Overall, the meta-analysis did not provide evidence for increased risk of either overall mortality or overall cancer incidence among cement production workers.

With respect to lung cancer mortality, two studies (Smailyte et al. 2004; Rachiotis et al. 2012) showed an increase in mortality. Smailyte et al. (2004) observed this excess of risk of both mortality and incidence, with some correlation with time since first exposure and the cumulative exposure to dust. Unfortunately in this study, data on individual smoking habits were not available, and the presence of an increased risk of bladder cancer suggests that smoking could have been a confounder. The study of Rachiotis et al. (2012) is based on proportionate mortality methodology, in which the mortality for each cause of death is dependent upon the mortality for the other causes of death. In addition, Vestbo et al. (1991) reported an excess of respiratory cancer incidence, but confounding by asbestos exposure from a different source was a plausible explanation of this result since there was an asbestos cement industry located close to the cement factory and in an analysis excluding all men with documented exposure to asbestos during employment in this asbestos cement factory revealed no increased risk of overall or respiratory cancer. The remaining studies did not show a significant excess of risk for this disease; the meta-analysis of all studies of lung cancer did not reveal an association, but that restricted to incidence data resulted in a summary RR, also for the subgroup of job category, equal to 1.24 and lower confidence limit equal to 0.98. Given the high fatality of this disease, it is unclear how the heterogeneity between mortality- and incidence-based studies—which in any case was not statistically significant—could be explained.

With respect to stomach cancer, McDowall reported an increased mortality, while Koh et al. (2011) reported an increased incidence without a corresponding mortality excess, possibly due to factors such as access to early diagnostic testing. However, these studies were carried out without quantitative exposure assessment, which is critical to relate exposures to health outcomes. Koh et al. (2013) performed a subgroup analysis based on cumulative exposure and found a dose–response relationship. However, the small number of cancer cases and short observation period are the major limitations of this study. Smailyte et al. and Amandus also observed a positive trend with estimated dose, but their results were not statistically significant, and there was no association with duration of employment; Jakobsson et al. (1990), in a case–control study, reported an increased risk of esophageal or stomach cancer restricted to short-time workers and thus unlikely to be related to cement dust exposure. Other studies, however, did not confirm this association, and the meta-analysis resulted in no association. The notion of lack of association between employment in cement manufacturing and risk of stomach cancer is reinforced by the results of the cumulative meta-analysis. These findings stress the need for replication of strong positive associations identified in the first available studies, which may represent false-positive results (Boffetta et al. 2008). Notably, no data on confounding factors were obtained in most of those studies such as smoking, dietary habits, or H. pylori infection.

The hypothesis of an association between cement dust exposure and risk of stomach cancer should be further investigated in studies with proper adjustment for confounders.

Some evidence of increased mortality or incidence for other specific cancers has been reported in some of the studies included in the systematic review. Jakobsson et al. (1990) observed an excess of incidence of colorectal cancer, with association with blue-collar employment, which did not result in a significant excess of mortality. In a subgroup analysis of the Korean study population (Koh et al. 2013), an increased risk of rectal cancer was reported, which, however, was not present in the main analysis of this population (Koh et al. 2011). No trend with exposure was detected in the subgroup analysis, which suggests that the increased rectal cancer risk may be due to random fluctuation. No support for an increased risk of colon or rectal cancer can be derived from the other studies. Increased mortality or incidence for other cancers has been reported sporadically [e.g., bladder cancer in the study from Lithuania (Smailyte et al. 2004)], without a consistent pattern. In the absence of replication, these results can be attributed to random fluctuation or residual confounding.

Results on cardiovascular disease mortality were limited, but they do not suggest an increased risk among cement workers. Mortality from non-malignant respiratory diseases was lower than expected in all studies except in the one by Giordano et al. (2012), in which, however, there was no association with estimated exposure. In occupational cohort studies, a reduced mortality from cardiovascular and non-malignant respiratory diseases may be due to a healthy worker effect; but the data available to address this form of bias in the case of studies of cement production workers are very limited.

Our systematic review and meta-analysis present several limitations. The number of studies and events was relatively small, reducing the precision of risk estimates. Differences in exposure circumstances, lack of quantitative exposure data, limited information on job titles or tasks, and use of job title as proxy for exposure may explain some of the inconsistencies in reported results. An additional limitation is the lack of data—in most studies—on possible confounders, primarily tobacco smoking and other occupational factors, but also dietary habits, including alcohol drinking. However, the lack of adjustment for the potential effect of these confounding factors would have likely biased the risk estimates away from the null and would not have likely produced negative results such as those observed in our meta-analyses. On the other hand, non-differential misclassification of cause of death (i.e., occurring both in the cohorts under study and in the reference population) might have occurred, leading to bias toward the null.

Moreover, the studies included in the review and meta-analysis investigated workers employed in cement production during more than 50 years, and during this period the industry has undergone a remarkable technological evolution. It is plausible that current exposure levels to potentially toxic agents, if any, are lower than those prevalent in the studies reviewed here.

The findings of our meta-analysis support the conclusion that cement dust exposure is not associated with an increased risk of cancer, cardiovascular diseases, and non-malignant respiratory diseases. It remains some degree of uncertainty regarding this general conclusion, particularly for some specific cancer sites, based on the fact that the results of the published studies are not consistently negative, and further investigations are needed to confirm or disprove these findings.