Effects of indium exposure on respiratory symptoms: a retrospective cohort study in Japanese workers using health checkup data

Study design and setting

This was a retrospective cohort study that used data obtained during a health check by the Chugoku Occupational Health Association between January 1, 2013 and June 30, 2015. In Japan, the Industrial Safety and Health Act requires employers to ensure that workers receive health checkups at least annually. Because this study used anonymized health checkup data, details about the companies and individuals were not available.

Subjects

The subjects were workers in indium-handling companies from Japan who had undergone a health checkup at the Chugoku Occupational Health Association during the study period. Subjects who had a history of respiratory disease or who were receiving treatment for the same, those with deficient outcomes, and those with outcomes from the start point were excluded.

The data of subjects who met all and none of the inclusion and exclusion criteria, respectively, were used in the final analysis.

In this study, new occurrence of respiratory symptoms was defined as outcomes, and thus observations in workers who already had outcome were excluded. Therefore, only the population at risk was included in the analysis.

Description of the dataset

The characteristics of the obtained data are shown in Fig. 1. The case where worker i received m health checkups has been illustrated as an example. The health checkups were nested among workers as per the data structure (Fig. 1). In the analysis, the period from the (k-1)th health checkup to the k th health checkup was treated as one observation. Exposure variables and covariates were obtained at the (k-1)th health checkup, and outcome variables were obtained at the kth health checkup. At k = 1, the observation start date was set to April 1 (first health checkup was from April to September) or October 1 (first health checkup was from October to March). This was done as in Japan, employment and relocation often starts on these days. Therefore, changes in employment and relocation related to the exposure status often occur from these days. In addition, the observations for all subjects were censored on the last health checkup before June 30, 2015.

Data collection

Statistical analysis

Observations with missing values for one or more of the variables required for analysis were excluded. All statistical analysis were performed using the Stata (Stata Corporation, version 15.1, College Station, TX, USA) software package. All p values were two-sided, and those less than 0.05 were considered to be statistically significant.

Ethical issues

This study was approved by the Ethics Committee of the Okayama University Graduate School of Medicine, Dentistry and Pharmaceutical Sciences and Okayama University Hospital (approval number K1607-008).

Because of the retrospective nature of the study, the ethics committee waived the need for obtaining informed consent from patients. The web page provides adequate information on the purpose and methods of the study for the benefit of the potential subjects. The web page also mentions how to decline participation and states that subjects are free to decline participation for any reason.

Descriptive statistics

The data of 3,161 workers from 22 companies were analyzed. The health checkups (i.e., observations) were performed for a total of 7,758 times. The worker selection and observation flowchart are summarized in Fig. 2. In cases where all health checks nested for a particular worker were excluded, the worker was also excluded (Fig. 2).

Initially, 1,389 observations and 526 workers were excluded owing to the presence of a history of treatment for respiratory disease; 231 observations with missing outcomes were then excluded along with 74 workers. These exclusions were necessary as 105 observations had an outcome at the start of the observation (k-1th health checkup). Finally, data of 6,033 observations from 2,561 workers of 22 companies were analyzed. The total person-years were 2,562.8 years, and a total of 162 outcome events occurred.

The demographic characteristics of the exposed and less-exposed workers at the first health checkup are shown in Table 1. At the first health checkup, 412 and 2,149 workers were exposed and less exposed, respectively. There were no large differences between the groups in terms of sex, age, and smoking status. However, there were large differences between groups with current and past work experience in a dusty environment. In the exposed group, 49 (11.9%) workers were currently working in a dusty environment, whereas 40 (1.9%) were in the less-exposed group. Among those with past working experience in a dusty environment, 51 (12.4%) and 13 (0.6%) were in the exposed and less-exposed groups, respectively (Table 1).

Table 2 shows the number of protective equipment wearers, the number of experienced workers at high exposure, chest radiograph findings, and S-In and KL-6 levels in the exposed group at the first health checkup. In total, 432 workers had been exposed at least once during the observation period. Overall, 414 (95.8%) workers always wore protective equipment, and none had experienced heavy exposures. The chest radiographs showed no abnormal findings in 400 (92.6%) workers. The S-In and KL-6 values were low, and their means and standard deviations were 0.15 ± 026 µg/L and 231.59  ± 88.48 U/mL, respectively; the standard values (less than 3 µg/L for S-In and less than 500 U/mL for KL-6) were exceeded in nine and one workers, respectively. Among them, the standard values of both serum markers had been exceeded in one worker (Table 2).

Figure 3 shows the cumulative incidence of the first complaint of respiratory symptoms; this was high in the exposed group. However, the log-rank test did not demonstrate statistical significance (p = 0.052).

Inferential statistics

The upper half of Table 3 shows the results of the Cox proportional hazard model considering only the first event during the observation period. The hazard ratios of the exposed group, with reference to the less-exposed group, were 1.42 (95% CI [0.99–2.04]) and 1.50 (95% CI [1.04–2.17]) in the crude and adjusted models, respectively.

The results of the PWP model are shown in the lower half of Table 3. The hazard ratios in the crude and adjusted models were 1.65 (95% CI [1.14–2.39]) and 1.61 (95% CI [1.04–2.50]), respectively.

Sensitivity analysis

The results of sensitivity analysis are shown in Table 4. For complaints of sleep disorders, the hazard ratios in the crude and adjusted models were 1.11 (95% CI [0.78–1.58]) and 1.31 (95% CI [0.86–1.98]), respectively. There were no statistically significant differences in the hazard ratios between the exposed and the less-exposed groups.

Strengths and limitations

The strengths of this research are as follows. First, the impact of random errors was extremely small owing to the large sample size. Second, this study employed the PWP model; as this model incorporates the probability of multiple failures into the statistical model, it is more realistic, and is likely to provide a more accurate estimate. Third, the indium medical examination was used to assess indium exposure. As this examination is mandated by Japanese law, misclassification of exposure was minimized.

However, there are also certain limitations. First, misclassifications may have resulted owing to the subjective outcomes. The exposed group was more likely to respond affirmatively to items “with symptoms” on the questionnaire. In cases where this tendency was strong, the HR was likely to be higher than the true value. However, sensitivity analysis did not demonstrate significant HRs. Therefore, the high HR of subjective respiratory symptoms may not be solely attributable to the influence of the response tendency of the exposed group. Second, the details of the exposure were unknown. In this study, the indium concentrations and nature of work were unknown, and they may have been large variations in exposure. This may have distorted the HR towards a value of one. However, this influence was likely to be minimal owing to the levels of significance. Third, exposures and outcomes were only evaluated during the health checkup, and data acquisition was delayed as the true timing of the exposures and outcomes was earlier than that of the health checkup. However, because this delay was similar for exposures and outcomes, it was likely to have no or minimal impact on the analysis. Fourth, the exposure group reported a higher proportion of dust work environment, which may confound the relationship between exposure and may have led to overestimations of the HRs. Meanwhile, because this confounding factor was adjusted using a statistical model, overestimation is minimal. Fifth, in this study, more sensitive biological markers such as S-In could not be compared between the exposed and the less-exposed groups. Therefore, it was not possible to conduct a more detailed analysis. However, S-In and KL-6 were at low levels even in the exposed group, and thus the effect was considered to be small.

Generalization of the study results

Because only companies and workers from one region of Japan were included in this study, the scope of the results may be limited by the geographical characteristics. Therefore, these results should be generalized only after careful consideration.