Ischaemic Heart Disease and Occupational Exposures: A Longitudinal Linkage Study in the General and Māori Populations of New Zealand

Abstract Objectives This study assessed associations between occupational exposures and ischaemic heart disease (IHD) for males and females in the general and Māori populations (indigenous people of New Zealand). Methods Two surveys of the general adult [New Zealand Workforce Survey (NZWS); 2004–2006; n = 3003] and Māori population (Māori NZWS; 2009–2010; n = 2107), with information on occupational exposures, were linked with administrative health data and followed-up until December 2018. Cox proportional hazards regression (adjusted for age, deprivation, and smoking) was used to assess associations between organizational factors, stress, and dust, chemical and physical exposures, and IHD. Results Dust [hazard ratio (HR) 1.6, 95%CI 1.1–2.4], smoke or fumes (HR 1.5, 1.0–2.3), and oils and solvents (HR 1.5, 1.0–2.3) were associated with IHD in NZWS males. A high frequency of awkward or tiring hand positions was associated with IHD in both males and females of the NZWS (HRs 1.8, 1.1–2.8 and 2.4, 1.1–5.0, respectively). Repetitive tasks and working at very high speed were associated with IHD among NZWS females (HRs 3.4, 1.1–10.4 and 2.6, 1.2–5.5, respectively). Māori NZWS females working with vibrating tools and those exposed to a high frequency of loud noise were more likely to experience IHD (HRs 2.3, 1.1–4.8 and 2.1, 1.0–4.4, respectively). Exposure to multiple dust and chemical factors was associated with IHD in the NZWS males, as was exposure to multiple physical factors in males and females of the NZWS. Conclusions Exposures associated with an elevated IHD risk included dust, smoke or fumes, oils and solvents, awkward grip or hand movements, carrying out repetitive tasks, working at very high speed, loud noise, and working with tools that vibrate. Results were not consistently observed for males and females and between the general and Māori populations.


Introduction
Growing evidence suggests a role for psychosocial, organizational, and environmental workplace factors in cardiovascular disease (CVD) (Hwang and Hong, 2012). In particular, work-related psychosocial factors, such as stress, have been linked to CVD with relative risk estimates ranging from 1.5 to 1.9 (Sara et al., 2018), and there is suggestive evidence for modest associations with loud noise and shift work. A meta-analysis of a small number of prospective studies on noise estimated an overall relative risk for CVD of 1.34 (95% CI 1.15-1.56) (Skogstad et al., 2016). Organizational factors, such as shift work, have also been studied, with a 2018 meta-analysis suggesting an almost 20% higher risk of CVD for those involved in shift work (Torquati et al., 2018), and another recent meta-analysis demonstrating a dose-response relationship (Cheng et al., 2019). Physical exertion, sedentary behaviour, long working hours, and chemicals (such as pesticides) are among other occupational exposures associated with CVD. However, the extent of research into occupational risk factors for CVD and the consistency of findings has been variable (Kristensen, 1989a,b;Hwang and Hong, 2012).
Most research has been undertaken in males or male-dominated occupations, with effects on female cardiovascular health largely unexplored. Ethnic minorities have also rarely been studied, despite considerable differences in CVD burden and occupational exposure profiles between ethnic groups and males and females Denison et al., 2018;Ministry of Health, 2019). Therefore, our understanding of occupational risk factors may not equally apply to all demographic groups, with interventions potentially increasing (rather than decreasing) health inequities. In New Zealand (NZ), Māori (the indigenous population) have a considerably greater CVD burden compared to NZ Europeans (Ministry of Health, 2019), and they are also overrepresented in blue-collar occupations (Milne et al., 2019). The same is true for other indigenous populations (Liebler, 2018;Wiemers et al., 2018) but, to our knowledge, no previous occupational CVD studies have focused on indigenous workers.
In this study we used information from two NZ Workforce Surveys (NZWS), one conducted among Māori and one in the general population, to assess associations between occupational exposures and ischaemic heart disease (IHD), which makes up the greatest proportion of CVD cases, by linkage to routinely collected health records. To allow ethnic and sex-specific associations to be studied, analyses were stratified by survey and sex.

Methods
This is a longitudinal study using occupational history and lifestyle information from two previously conducted occupational surveys in the general and Māori population (see below). Incident IHD events were identified using routinely collected health data for a 7-14 year period from the date of the interview until 31 December 2018.

What's Important About This Paper?
A range of psychosocial, organizational, and environmental workplace factors have been linked to cardiovascular disease; however, ethnic minorities and females are often under-represented in this research. In this study, a number of exposures were associated with ischaemic heart disease (IHD), and multiple occupational exposures increased the risk of incident IHD. Associations were not consistent between males and females or for the Māori and general populations. These findings suggest that occupational risk factors for IHD are not equivalent across all populations and future research and interventions may not be generalizable across all populations.

Workforce surveys
The methods for the two New Zealand Workforce Surveys (general population NZWS (Eng et al., 2010) and Māori NZWS (Denison et al., 2018)) have been described in detail previously. Briefly, for each survey, a random proportionally stratified, systematic, and self-weighted sample of people aged 20-64 years were selected from the Māori and general electoral rolls. The general population NZWS was conducted from 2004 to 2006. Invitations to participate in a telephone interview were mailed up to three times and nonresponders were contacted by phone (if available), with 3003 participants (37%) completing the survey. The Māori NZWS was conducted from 2009 to 2010 using the same methodology and resulted in 2107 participants (29%) completing the survey (Denison et al., 2018). Potential for participation bias from low survey response was evaluated and considered to be small ('t Mannetje et al., 2011), i.e., while some groups were under-represented, the prevalence of key survey variables (both occupational exposure and health-related variables) were unchanged after standardizing to the demographic distribution of the source population, and similar between early and late responders. Two participants were included in both surveys and we, therefore, excluded their most recent interview (i.e., the Māori NZWS).
The questionnaire included questions about lifetime work history, current workplace exposures, and demographic and lifestyle factors. Ethics approval was granted by the Massey University Human Ethics Committee (NZWS-WGTN 03/133, Māori NZWS-MUHEC 08/28) and from the New Zealand Health and Disability Ethics Committee for the linkage of the two surveys (16/NTB/173).

Self-reported occupational exposures
Participants were asked whether the following exposures were present (yes/no) in their current or most recent workplace environment: dust; smoke or fumes; gas; oil and solvents; acids or alkalis; and pesticides. They were also asked about organizational factors, including working irregular hours (outside 7:00-20:30) and night shift (for at least 3 h between 00:00 and 5:00) in the previous month, as well as the number of hours worked per week (<35, 35-45, 46-54, and ≥55 h). Participants were asked how often their job involved exposure to the following physical factors: awkward and tiring positions; awkward grip or hand movements; lifting; standing; sitting; using tools that vibrate; loud noise; repetitive tasks; working at very high speed; and working to tight deadlines. This was measured on a scale from never to always (provided as a percentage of time or a point on the scale from never, ¼, ½, ¾, or all the time) and the median frequency of exposure (averaged across the cohorts), was calculated to determine a cut-point for 'low exposure' (i.e., <=median) and 'high exposure' (i.e., >median). Finally, participants rated how stressful they found their current job: none; mild; moderate; very; and extremely stressful, and this was categorized into: none/mild; moderate; and very/extremely.
The occupational exposures included in the analyses were selected based on previously reported associations with CVD (Hwang and Hong, 2012). We also asked about other exposures (e.g. boring work and working outside), but these were not included as they have not previously been shown to be associated with CVD.

Other CVD risk factors assessed via questionnaire
Participants provided a current or most recent job title with a description of job tasks, and each job was coded using the NZ Standard Classification of Occupations (NZSCO) 1999, which is a hierarchical skills-based classification with nine major occupational groups (Statistics New Zealand, 2001). Age at the interview was categorized as follows: 20-34, 35-44, 45-54, and ≥55 years. Socioeconomic status (SES) was assessed using the NZ Deprivation Index 2006, a census-based index with a relative deprivation score from 1 to 10, based on place of residence. The distribution of deprivation is presented in quintiles, but for subsequent analyses, it was dichotomized combining scores 1-8 (least deprived) and 9-10 (most deprived).
Smoking status at the time of the interview was analysed as never/ever and as pack-years, calculated from the number of cigarettes smoked per day divided by 20, multiplied by the number of years smoked. Results for both measures were very similar (not shown); therefore, only results for ever/never are presented. Body Mass Index (BMI) at the time of interview was calculated using self-reported height and weight grouped into four categories (<18.5, 18.5-24.9, 25-29.9, and ≥30) based on WHO guidelines (World Health Organization, 2000).

IHD identified from linked health data
IHD events from the deidentified survey information were linked to the Integrated Data Infrastructure (IDI), a longitudinal meta-dataset of de-identified data from government agencies, at the individual level (Milne et al., 2019). Before linkage, probabilistic matching was conducted based on the date of birth, sex, family name, and first two given names, to identify National Health Index numbers to enable linkage to Ministry of Health (MoH) datasets. From the surveys, 98% of respondents were successfully matched and could be linked to mortality, public hospital diagnoses, and pharmaceutical dispensing records.
The IHD definition included IHD deaths, hospital discharges, and procedures using International Classification of Disease (ICD) codes and ≥2 dispensing of anti-anginal drugs from the pharmaceutical claims dataset (see Supplementary Table S1, available at Annals of Occupational Hygiene online) and was based on a previously developed definition (Wells et al., 2017). Primary health care information was not used as it is not available in the IDI.

Follow-up of IHD events
Participants with prior IHD were excluded using the same IHD definition. For incident IHD, participants were followed from the date of interview: 2004-2006 for the general population and 2009-2011 for Māori. Participants that moved overseas or died from other causes were identified through immigration and mortality data in the IDI, respectively, and were censored from that time point. The date last observed for participants that were not lost to follow-up, not deceased, or did not have an IHD event, was 31 December 2018.

Statistical analyses
Analyses were stratified by survey and sex. Cox proportional hazards regression was used to calculate hazard ratios (HR) for associations between occupational exposures in participants' current or most recent job, and incident IHD. Models were adjusted for age groups, deprivation, and smoking status. For physical exposures, the three-level variable (none/low-/high-frequency exposure) was used. For analyses where the number of IHD cases was <6 and had to be suppressed due to IDI requirements (see below), both low-and high-level exposures were combined to create a dichotomous variable of exposed versus not exposed.
Associations for combined workplace exposures were assessed by combining dust/chemical exposure variables if they were significantly (p < 0.05) and positively associated with IHD in at least one cohort, by summing the number of exposures for each participant (0 (reference), 1, 2, or 3). We used the same approach for physical exposures resulting in a summed exposure variable ranging from 0 to 4. The focus for statistical analysis of ordinal exposure variables was on whether exposure indicated increased IHD risk.
In compliance with the IDI confidentiality requirements, all frequencies were rounded to the nearest multiple of three, and percentages were calculated from the rounded counts (hence total numbers of participants in each table vary slightly and do not add to exactly 100%). All statistical tests used the unrounded counts. All counts under six and the HRs from these are suppressed (marked 'S' in the tables).
As the study involved multiple comparisons, we compared the number of expected statistically significant (p < 0.05) findings (based on chance) with the number of actual observed statistically significant findings. We also assessed whether the difference in expected and observed significant findings was significantly (p < 0.05) different overall. To do this, we determined, via the binomial theorem, the probability of s 0 or more successes from a sequence of k Bernoulli trials given the probability of success for each test is p. This overall probability is: where k C s is the number of ways of choosing s items from k. Here p is set to 0.05. Evaluation of this sum is straightforward for any s and k and can proceed iteratively because the ratio of the (s+1)th to the sth term in the expansion is The procedure is a variation of the multiple comparison adjustment method of Šidák (Sidak, 1967), except that, rather than setting p 0 and solving for p, here p is set and the corresponding p 0 is determined.

Results
A total of 70 participants could not be linked to health data and for 15 a date last observed could not be determined. A further 213 participants (83 NZWS and 130 Māori NZWS) with an IHD event before the interview were excluded, resulting in 2875 participants of the NZWS survey and 1935 participants of the Māori NZWS survey for analyses. Mean follow-up for the NZWS and Māori NZWS were 12.1 and 7.5 years, respectively.

Incident IHD
There were 135 incident IHD cases in the NZWS and 93 in the Māori NZWS. As expected, IHD cases were overrepresented in males, the oldest age group, ever smokers, overweight/obese BMI categories, and high deprivation group (Table 1). IHD cases were also overrepresented in plant and machine operators and assemblers and elementary occupations.

Organizational factors, dust and chemicals, and stress
Dust, smoke or fumes, and oils and solvents (Table 2) were associated with IHD after adjusting for age, deprivation, and smoking in males of the NZWS [HR    There were no associations observed for other chemicals, organizational factors, or stress in either survey. The proportion of participants exposed to all three chemical exposures (dust; smoke or fumes; oils and solvents) was similar between surveys but considerably greater in males compared to females (NZWS males 12.2%; Māori NZWS males 12.7%; NZWS females 2.6%; Māori NZWS females 2.5%). In NZWS males, exposure to two dust and chemical factors was significantly associated with IHD [HR 2.3 (1.4-3.8)]; this was not found in the other groups (Table 2).

Physical exposures
High exposure to awkward grip or hand movements was associated with IHD in both males and females of the NZWS [ ]. An inverse association with IHD was found for exposure to awkward or tiring positions and loud noise in Māori NZWS males, statistically significant only for the low exposure categories (HRs 0.5 (0.2-1.0) and 0.4 (0.2-0.9), respectively). Among Māori NZWS females, there was a positive association with high exposure to loud noise [HR 2.3 (1.1-4.8)]; exposure to tools that vibrate was also positively associated [HR 2.1 (1.0-4.4)]. Sitting was inversely associated with IHD among female Māori NZWS (but not NZWS females), significant for the high exposure category [HR 0.3 (0.1-0.8)].
The proportion of participants exposed to all four physical exposures that were positively associated with IHD in at least one cohort (high frequency of awkward grip or hand movements; repetitive tasks; working at very high speed; loud noise) was higher among Māori NZWS compared to the NZWS (NZWS males 3.1%; Māori NZWS males 9.2%; NZWS females 2.4%; Māori NZWS females 6.1%). In NZWS males, exposure to three of these physical factors was associated with an almost four times greater IHD risk and in NZWS females, exposure to two physical factors was associated with a 4.6 times greater risk (Table 3). No associations were found for the Māori NZWS.
Physical exposure analyses were repeated using a dichotomous cut-off representing exposure occurring ≥25% of the time (rather than using a median cut-off), which showed very similar results (see Supplementary Table S2, available at Annals of Occupational Hygiene online).

Discussion
Several occupational exposures were associated with incident IHD. Oils and solvents, dust, and smoke or fumes were associated with an increased risk in the general population of males, and a high frequency of awkward grip or hand movements was associated with an increased risk in both males and females of the general population. High frequency of repetitive tasks and working at very high speed were positively associated with IHD among general population females. Among Māori NZWS females, working with tools that vibrate and exposure to loud noise were associated with an increased risk of IHD. No associations for physical factors were observed in Māori NZWS males or females, and none of the occupational exposures were associated with IHD in Māori NZWS males.
There are several explanations that may contribute to the differences in associations observed between Māori and the general population and males and females. First, statistical power was lower in the Māori cohort due to fewer participants and shorter follow-up, resulting in fewer incident IHD events, even though exposure prevalence of physical factors was higher for Māori NZWS males. Similarly, lower exposure prevalence and fewer IHD cases may have contributed to fewer significant associations observed for women. Reduced power was also an issue when comparing different exposure levels, with fewer study participants in the high exposure groups potentially resulting in non-significant findings for high exposure groups even if results for lower exposures were statistically significant.
Second, the nature, as well as circumstance of exposure may differ between groups. This may, for example, have contributed to the effect observed for sitting, which was inversely associated with IHD only in Māori NZWS females. Similarly, 'tools that vibrate' used by women can be different from those used by men, which may explain the difference in IHD risk observed between men and women reporting this exposure. More generally, measures of self-reported exposures are relatively crude, which may lead to exposure misclassification, particularly for factors such as 'working at very high speed' that, although frequently used as a measure of job intensity in working conditions surveys, are poorly defined and/or quantified. Also, perception and reporting of exposure may differ between Māori and non-Māori and males and females, and exposure misclassification may also differ between these groups, although our analyses, which were stratified by gender and ethnicity, would be less affected.
Thirdly, the distribution of occupations in Māori differs from that of the general population (Table 1), which Table 3. Associations between IHD and physical exposures. a Adjusted for age group, high deprivation, and smoking status.
*P < 0.05; **P < 0.01. Table 3. Continued impacts both exposure prevalence and the occupational composition of the reference group and the interpretation of HRs relative to each survey's reference group. Adjustment for 1-digit occupational group did not affect results (data not shown), although this adjustment alone may not be sufficient to address this issue, particularly in stratified analyses as reported here. Furthermore, occupational exposures were only available for the current or most recent job, and although the average number of jobs in males and females was similar (NZWS males 4.4; NZWS females 4.9; Māori NZWS males 4.5; Māori NZWS females 4.9), the average duration of employment of the last job was longer for males (NZWS males 9.5. years; NZWS females 6.0 years; Māori NZWS males 9.2 years; Māori NZWS females 6.9 years); exposure estimates may therefore not be entirely comparable for males and females. To partially address these issues, sensitivity analyses were conducted excluding those who were in the job for <5 years (582 NZWS males, 861 NZWS females, 336 Māori NZWS males, 558 Māori NZWS females). Although this resulted in wider confidence intervals and some findings losing statistical significance, particularly for NZWS females, results were largely unchanged (data not shown). Fourthly, this study included only a limited number of exposures that have previously been associated with IHD. Information on other relevant exposures [e.g., job insecurity, discrimination, electromagnetic fields, physical exertion, and environmental tobacco smoke (Wadsworth et al., 2007;Hwang and Hong, 2012;Virtanen et al., 2013)] was not collected in the original workforce surveys.
In addition to the aforementioned explanations, sexspecific differences in susceptibility and pathophysiology of CVD may play a role in the observed differences between males and females (Shufelt et al., 2018).
We observed an association with dust among general population males, which is consistent with previous studies on occupational particulate matter and CVD (Fang et al., 2010). Similarly, our finding that exposure to smoke or fumes was associated with IHD, is consistent with previous observations that smoke and fumes, such as carbon monoxide and combustion products, were associated with CVD (Kristensen, 1989a;Gustavsson et al., 2001). The evidence is less clear for solvent exposure; earlier work suggested organic solvents may be linked to CVD (Wilcosky and Simonsen, 1991), while a more recent study found no link between solvent exposure and CVD or IHD (Bulka et al., 2019).
In contrast to leisure-time physical activity, occupational physical exertion, and heavy lifting have been linked to increased IHD risk (Petersen et al., 2012;Li et al., 2013;Holtermann et al., 2018). In our study, heavy lifting was not significantly associated with IHD; high frequency of awkward grip or hand movements, on the other hand, was associated with IHD for both sexes, and repetitive tasks and working at very high speed was also associated with IHD for NZWS females. These specific exposures have not previously been studied in the context of IHD and may explain the increased IHD risk observed for occupational groups such as plant and machine operators and assemblers (Holmes et al., 2011).
Sedentary behaviour has been associated with CVD (Carter et al., 2017), but there is limited evidence for occupational sedentary behaviour (van Uffelen et al., 2010). In this study, sitting >50% of working time was associated with a reduced risk in female Māori NZWS only. The reasons are unclear, but it is possible that sitting times at work may not adequately represent sedentariness or that it is associated with other work-related risk/protective factors; for example, prolonged sitting may reflect a lack of physically demanding exposures.
Noise has repeatedly been associated with CVD (Skogstad et al., 2016). In this study, frequent exposure to noise was associated with >2 times the risk of IHD in Māori NZWS females, whereas noise exposure in Māori NZWS males was inversely associated. This supports previous literature indicating women may be more adversely affected by noise exposure (Dzhambov and Dimitrova, 2016). In Māori NZWS females, we also found an association with tools that vibrate, and whilst modest evidence of a relationship between vibration and CVD exists, there are limited studies among female workers (Krajnak, 2018).
Chronic stress, including occupational stress, has been associated with CVD (Sara et al., 2018), but this was not observed in this study. It is possible that our measure of stress was insufficiently nuanced, with previous studies using more refined measures, such as job strain and effort-reward imbalance (Sara et al., 2018).
Both shift work and long working hours have previously been linked to IHD, however, the overall evidence is inconsistent (Rivera et al., 2020). In this study, we did not observe significant associations with IHD for the number of hours worked, for working irregular hours, or for working night shifts, but the prevalence of these exposures was low.
Few studies considered the combined effect of occupational exposures, but some analysed exposure interactions (e.g., physical activity, noise, job strain, and shift work), and reported additive effects (Virkkunen et al., 2006;Eriksson et al., 2018). In this study, exposure to multiple occupational factors was associated with greater IHD risk, which may explain the elevated prevalence of IHD risk factors we have previously observed for some occupational groups (e.g. plant and machine operators and assemblers; elementary occupations (Barnes et al., 2020)) where exposure to multiple risk factors is common.
Limitations, in addition to those described above, include that we did not have access to private hospital or primary health care information, which may have resulted in an underestimation of incident IHD, although we did have access to community dispensing of anti-angina medications, which likely captured most IHD cases that did not result in public hospitalization. Community dispensing data had limited date range availability (2005 onward); however, most incident IHD events were identified through hospital admissions (73-80% across males and females of both surveys), which were available from 1988 onward. A related issue is that due to more limited access to tertiary hospitals in rural areas IHD diagnoses may be undercounted in these areas. However, the use of community dispensing of anti-angina medications in our IHD definition will likely have captured most cases that did not result in public hospitalization. In addition, the percentage of New Zealanders living in rural areas is relatively small [16.3% and 18.0% for the general and Māori population, respectively (Environmental Health Intelligence New Zealand, 2020)]. Any potential bias resulting from undercounting IHD cases in rural areas would therefore be minor.
There may be confounders that were not considered, such as diet, leisure-time physical activity, and alcohol consumption. However, analyses adjusting for BMI did not significantly alter results other than slightly change p-values (data not shown). We did not adjust analyses for high blood pressure, diabetes, elevated cholesterol as these may be on the causal pathway between (some) occupational exposures and IHD.
Finally, assessing multiple exposures stratified by sex and survey resulted in a large number of comparisons, so some statistically significant results may be due to chance. However, as shown in Supplementary Table  S3, available at Annals of Occupational Hygiene online, we found more significant results than would occur by chance. This was particularly the case for analyses involving males and females of the NZWS (Tables 2  and 3) and analyses described in Supplementary tables (Table S2, available at Annals of Occupational Hygiene online) for females in the Māori NZWS. For the remainder of the analyses, we still found more statistically significant findings than expected by chance, but the differences were less pronounced and findings of these analyses should therefore be interpreted with caution, particularly for those not previously reported in the literature, and for subgroup analyses with a small number of significant findings.
Although there are limitations of this study, there are also major strengths including the large proportion of females (>50%) and Māori (40%), the inclusion of a range of exposures, which were collected before IHD diagnosis, limiting recall bias, as well as the measure of IHD incidence not relying on self-reports.
In conclusion, associations with exposure to dust, smoke or fumes, oils and solvents, awkward grip or hand movements, carrying out repetitive tasks, working at very high speed, loud noise, and working with tools that vibrate and IHD were found, but results were often not consistent for males and females and between the general and Māori populations. These findings suggest that occupational risk factors for IHD may differ across populations.

Supplementary Data
Supplementary data are available at Annals of Work Exposures and Health online.