The Mineral Resources Health Assessment Program (MRHAP) cohort was enumerated by the University of Minnesota in 1983 with the support of the Iron Range Resources and Rehabilitation Board (IRRRB) and the cooperation of the seven mines then in operation and included 68,737 individuals identified in employee records from the mining industry in northeastern Minnesota before 1983. It included workers employed in either the hematite and/or taconite industry. The MRHAP cohort was followed for vital status through 2010 (Allen et al., 2014). Vital status was ascertained using the Social Security Administration, the National Death Index, and the Minnesota Department of Health and state death certificates outside of Minnesota. All deaths were coded to the International Classification of Disease (ICD) codes in effect at the time of death.

Cases of mesothelioma were identified through the Minnesota Cancer Surveillance System (MCSS) and/or death certificates for out of state mesothelioma cases. In addition to the original 80 mesothelioma cases in the Lambert et al. (2016) study, 24 additional cases were identified with the MCSS through 2016. Mesothelioma in MCSS was histologically confirmed and coded using ICD-O-3 histology codes 9050–9053 and cases from death certificates were identified as ICD 10^th Revision code C45. Versions prior to the 10^th revision of ICD did not have a code specifically for mesothelioma. Four controls were selected for each case using an incidence density sampling approach from risk sets of cohort members of similar age (years of birth±2 years). Controls for the 80 cases in the Lambert et al. (2016) study were alive and without a diagnosis of mesothelioma on the date of diagnosis or death of the case. The controls for the newly added 24 cases were alive as of the last full vital status ascertainment (2010). All cases and controls had to have clear evidence of employment in the mining industry based on review of work history records obtained when MRHAP cohort was enumerated. The final study population in the current study included 104 cases and 410 controls. The study was approved by the University of Minnesota Human Subjects Research Committee (Reference number 0709S16941).

The detailed exposure assessment work for this study can be found elsewhere [19–21]. In brief, seven taconite mines of the Mesabi Iron Range were associated with this study. For confidentiality purposes, these mines were coded as Mine A to Mine G. Mines A and G were located on the eastern side (Zone 2 and Zone 4) of the range, and Mines B to F were on the western side (Zone 1). The opening year of these mines varied, ranging from 1955 to 1977 [22]. As of 2010, six mines were still in operation and one closed in 2001. The previous TWHS investigations, grouped taconite worker job titles into 28 similar exposure groups (SEG) within eight departments [23] and created a job exposure matrix (JEM) for EMP in these SEGs based on current and historical NIOSH 7400 phase-contrast microscopy (PCM) monitoring.

Four JEMs were used in this updated study. Details about how they were formed can be found elsewhere [19,20]. In brief, all JEMs were for EMP exposures, but each represented a unique EMP dimension range (see Table 1). The NIOSH EMP JEM provided yearly-averaged NIOSH EMP concentration estimates in the workplace for each mine-department combination and calendar year. Concentration estimates were created based upon 2036 historical NIOSH 7400 measurements and the historical reconstruction strategy described in Shao et al., 2019. The NIOSH EMP definition is used for regulatory purposes and is the same criteria used in our previous mesothelioma study. The JEMs for the other three EMP definitions were created by applying the department-level conversion factors to the NIOSH EMP JEM. These conversion factors were estimated based on probability density functions derived from actual dimensional information of 11,147 individual EMPs, which were associated with 1268 personal EMP samples analyzed using the ISO 10312 or 13794 transmission electron microscopy (TEM) methods, collected in the Minnesota taconite mines. The Chatfield EMPs refer to EMPs that meet the dimensional criteria proposed by Chatfield in 2008 [24]. He observed the dimensional difference between pre-determined asbestiform amphibole fibers and non-asbestiform amphibole fibers, and found that asbestiform ones were more likely to be thinner and longer. The Suzuki EMPs were defined here as the EMPs that were less than or equal to 5μm in length and 0.25μm in width. Suzuki et al. reported in 2005 that the majority of the asbestiform fibers in the human lung and mesothelial tissues examined were within this range. The Cleavage Fragments here were defined as any EMPs with less than 20:1 aspect ratio [25,26].

Work history information for cases and controls, including all job titles and dates of employment, was abstracted from available mining company work records from the 1950s through the end of 1982. All job titles were standardized into 28 SEGs of 8 departments. Some study subjects worked in the earlier hematite industry. Hematite as a “direct shipping ore” did not require the processing and concentrating techniques of taconite and no historical EMP samples were available during the hematite production period. Therefore, EMP exposure levels for the hematite mining period were set to zero. For each study subject, we estimated cumulative exposure values based on each of four EMP JEMs: “NIOSH EMP”, “Chatfield EMP”, “Suzuki EMP”, “Cleavage Fragments.” Cumulative exposure was calculated as the sum of the products of time/location-specific exposure concentrations and the time each individual spent in each exposure category, i.e., in units of (EMP/cc)×years. Commercial asbestos was likely used in some operations of the processing, but no historical measurements were available for commercial asbestos exposure in these operations. The study team and industrial hygienists from the taconite companies assessed the likelihood and frequency of exposure to commercial asbestos within each SEG and assigned a commercial asbestos score (low, medium, or high) accordingly. Ultimately, the number of years worked in an SEG with a high commercial asbestos score was used as a surrogate metric for historical commercial asbestos exposures. A sensitivity analysis was conducted to evaluate how the choice of a commercial asbestos surrogate metric would affect the estimation of mesothelioma risk.

Descriptive analyses were performed to assess the relationship between demographic and occupational factors and case/control status. Conditional logistic regression was used to estimate the effect occupational exposure to EMP and risk of mesothelioma for a unit change in exposure variables (1 taconite year or 1 cumulative exposure value EMP/cc×year). The main exposure was either years of employment in taconite mining or cumulative exposure to EMP (either NIOSH, Chatfield, Suzuki, or Cleavage fragments). When taconite mining was the main exposure, only years in the taconite industry counted toward the number of years. Years in hematite mining were not included in the taconite years exposure. Taconite-year equals to 0 for the hematite only workers. Separate models were fit for each exposure definition (NIOSH, Chatfield, Suzuki, Cleavage) and used cumulative exposure value in EMP/cc×year as the primary exposure. Models that included two EMP exposure terms were used to explore the effects of the Chatfield, Suzuki, Cleavage exposure definitions while controlling for the NIOSH EMP exposure. The conditional logistic regression approach accounted for the person-time matching of cases and controls within risk sets and results are reported as rate ratios and 95% confidence intervals. The final models incorporated the following two additional confounders: number of years employed in hematite mining, and duration of work in SEGs with a high commercial asbestos score. All statistical analyses were conducted using SAS (version 9.4, SAS Institute, Cary NC).

A total of 104 cases and 410 controls were included in the study (Table 2). All cases and 94% of controls were male. A greater proportion of cases (71 out of 104) were taconite workers compared to the controls (236 out of 410). While the duration of hematite employment was similar for both groups, cases had a higher median number of years working with taconite. Among those who had ever worked in taconite mines, a larger percentage of cases were associated with Zone 2, and they also had a higher median number of years in this zone compared to the controls. The cumulative exposure values of all four EMP types for cases were higher than for controls. Median years spent in SEGs with a high commercial asbestos score were the same between the controls and the cases.

The risk of mesothelioma was associated with the number of years of employment in the taconite mining industry (RR = 1.02, 95% CI 1.00 to 1.05) (Table 3). However, the Model dividing workers into categories based on the median length of employment of cases did not reveal an increased risk associated with longer tenure. The mesothelioma risk increased with duration of employment in both Zone 1 and Zone 2, but was not associated with duration of employment in Zone 4. The associations observed with employment duration were consistent with findings from the previous study.

Cumulative exposure to NIOSH EMP (EMP/cc×years) was associated with an increased risk of mesothelioma (RR=1.20, 95% CI 0.99 to 1.46) (see Table 4). Cumulative exposure to the other three EMPs were all positively associated with the risk of mesothelioma: RR = 2.85 (95% CI of 0.97 to 8.40) for the Chatfield EMP, 1.04 (95% CI of 1.00 to 1.09) for the Suzuki EMP, and 1.03 (95% CI of 1.00 to 1.06) for the Cleavage Fragments. Additionally, a sensitivity analysis was conducted by switching the default commercial asbestos metric from the number of years in SEGs with a high asbestos score to the total number of years in SEGs with a medium or high asbestos score. This change resulted in stronger estimation results: RR = 1.42 (95% CI of 1.06 to 1.90) for the NIOSH EMP, 3.70 (95% CI of 1.11 to 12.27) for the Chatfield EMP, 1.05 (95% CI of 1.00 to 1.10) for the Suzuki EMP, and 1.04 (95% CI of 1.01 to 1.08) for the Cleavage Fragments.

In our study population, all EMP definitions were positively correlated to varying degrees, (Figure 1). The Pearson correlation coefficients (r) between NIOSH EMP and Chatfield EMP, Suzuki EMP, and Cleavage Fragments were 0.53, 0.66, and 0.84, respectively. The correlation between Chatfield EMP and. Suzuki EMP was r=0.88. The degree of this correlation and the relatively small number of cases made separating the independent effects of the non-NIOSH EMPs challenging. In models adjusting for NIOSH EMP exposure, the estimates for the Chatfield EMP, Suzuki EMP, and Cleavage Fragments remained positive but were attenuated and less precise (RR = 2.19 (95% CI of 0.69 to 6.96), 1.03 (95% CI of 0.98 to 1.09), and 1.02 (95% CI of 0.98 to 1.07) respectively).

In this study, we were not able to update the vital status of the entire MRHAP cohort. All new mesothelioma cases were cases identified through MCSS system in Minnesota. It is possible that we missed out-of-state cases this time. Also, some new controls who were alive in 2010 may have died prior to the diagnosis of mesothelioma of the matched case. This would affect the final results if the deceased controls of similar age had an exposure experience that was not representative of the MRHAP cohort. However, no additional exposure was ascertained for either cases or controls after 1982.

Exposure reconstruction in this investigation was based on available work history information. This information was available through the end of 1982. It’s likely that some study subjects were still working after this point in time. If that were the case, then the life-time exposure values estimated for these subjects could be lower than their actual exposures, but the extent to which this might differ between cases and controls was not known. To partially address this limitation, we ran models with 0, 20 and 30-year latency, thus cases and controls had similar measures of missing data. The associations found in the exposure lagged models were similar to those for non-lagged models. Additionally, there is the potential for a healthy worker survivor bias, however the lack of complete work histories for cohort members employed after 1982 limits the potential for analyses that effectively adjust for such bias.

Uncertainties in historical exposure reconstruction [19,20] may also influence the efficacy of our epidemiological analysis: (1) Our NIOSH EMP JEM was developed using data from about 2000 historical NIOSH 7400 PCM measurements with the earliest measurement dating back to 1978, over two decades after the inception of the Minnesota taconite industry. (2) Historical exposure to other EMP metrics were entirely unknown from historical PCM measurements. To estimate these exposures from NIOSH EMP JEM, we examined up to 11,147 individual TEM EMPs collected from six study mines active in 2010, and applied a bivariant normal distribution assumption to characterize the overall size distribution of EMPs.