Lymphohematopoietic cancer mortality among Korean semiconductor manufacturing workers

Female operators working in the semiconductor industry in South Korea experienced excess cancer mortality due to malignant neoplasms in the lymphoid, hematopoietic, and related tissues, especially leukemia. Female operators showed a 2.59-fold increased risk of mortality due to malignant neoplasms in the lymphoid, hematopoietic, and related tissues compared with the general population and 2.92-fold increased risk of mortality due to leukemia.

To the best of our knowledge, this study is the first to report a significantly higher SMR for malignant neoplasms in the lymphoid, hematopoietic, and related tissues. Since the invention of the integrated circuit in 1959, Silicon Valley in California emerged as a high-tech industrial area through the 1960–1970s, when semiconductor enterprises such as Fairchild, Intel, and Advanced Micro Devices started their businesses. In the late 1970s, occupational health risks in the semiconductor industry began receiving attention. In 1983, Joseph Ladou raised concern over the high rate of illness and percentage of systemic poisoning among workers in the semiconductor manufacturing industry in California [22]. The Semiconductor Industry Study done in 1981 concluded that many toxic materials and a variety of solvents, acids, and metals, such as arsenic, are used in the semiconductor manufacturing process [23].

A key event that triggered health concerns in the semiconductor industry was a series of lawsuits filed by employees of the International Business Machine (IBM) in the 1990s. These employees claimed that their cancer was linked to chemical exposures in IBM’s semiconductor and disk drive plants. Thus, epidemiological studies were conducted in the late 1990s and early 2000s examining the health risks associated with work in the semiconductor industry. Clapp et al. investigated cancer mortality among employees who worked in the IBM Endicott plant between 1969 and 2001. The study reported that the proportional cancer mortality ratio of all lymphatic and hematopoietic tissues was 1.23 times higher than that of the standard population in male workers [24]. Bender et al. investigated the employees of IBM’s semiconductor manufacturing facilities. They reported that the standardized incidence ratios (SIRs) for all cancer types (SIR = 0.81, 95% CI = 0.77–0.85), NHL (SIR = 0.94, 95% CI = 0.49–0.98), and leukemia (SIR = 0.70, 95% CI = 0.49–0.98) were not significantly higher. Other cohort studies reported significantly lower SMRs for all cancer types. Using the data of two large US semiconductor companies between 1968 and 2002, with 37,225 fabrication workers and 62,856 non-fabrication workers, compared to the general population, the SMR for all cancers was significantly low in fabrication workers (SMR = 0.74, 95% CI = 0.66–0.83) and non-fabrication workers (SMR = 0.72, 95% CI = 0.66–0.79). Similarly, the SMRs for all leukemia and aleukemia were also not significantly higher among fabrication workers (SMR = 0.75, 95% CI = 0.34–1.23) and non-fabrication workers (SMR = 0.82, 95% CI = 0.52–1.22) [15]. Despite the high level of public concern and the resentment of victims, it had been nearly impossible to suggest a causal relationship between the hundreds of cases of cancer and the work environment, especially for hematopoietic malignant disorders.

A 1985 study on cancer risk in semiconductor wafer processing workers was reported for the first time in the U.K. and has been reported across a series of cohort studies in the West Midlands. A significantly increased risk of some types of cancer was reported [25,26,27]. Using the standardized registration ratios (SRRs), the cancer incidence is calculated as the ratio of observed to expected number of cases based on the cancer incidence rates in England and Wales from 1971 to 2000. Results showed a significant excess morbidity for cancer of the rectum (SRR = 1.99, 95% CI = 1.20–3.10) and malignant melanoma (SRR = 2.17, 95% CI = 1.12–3.79). McElvenny et al. reported the cancer incidence and mortality rates among current and former workers at a Scottish semiconductor manufacturing facility, including 2,126 male workers and 2,262 female workers. In this cohort study, a significantly increased SMR for all types of malignant neoplasms or SRR for all cancer types were not found [12]. The SRR for malignant neoplasms of the respiratory and intrathoracic organs was significantly increased in all female workers (SRR = 2.73, 95% CI = 1.36–4.88) and female workers in fabrication areas (SRR = 3.17, 95% CI = 1.45–6.02). The mortality and incidence of hematopoietic cancers did not increase in male or female workers. Follow-up studies reported that there was no significant increase in the incidence and mortality rates of lung cancer [28, 29]. The studies reported in the UK could not identify the relationship between semiconductor processes and diseases. Furthermore, there was no significantly increased incidence or mortality for lymphohematopoietic cancers.

This occupational health issue emerged in Asian countries, including Taiwan, Singapore, and Korea, in the 2000s, when it had already been a social issue in the 1980s in the US and the 1990s in the UK [30]. The Occupational Safety and Health Research Institute of South Korea conducted a retrospective cohort study of 113,443 workers from 11 semiconductor companies between 1998 and 2008 [7]. The SMRs for all-cause mortality were 0.25 (95% CI = 0.21–0.29) in male workers and 0.66 (95% CI = 0.55–0.80) in female workers, and the SMR for malignant neoplasms in the lymphoid, hematopoietic, and related tissues in female workers was 1.56 (95% CI = 0.78–2.78). The SIRs for all cancer types were 0.86 (95% CI = 0.74–0.98) in male workers and 0.88 (95% CI = 0.74–1.03) in female workers. Among female workers, the SIR for malignant neoplasms in the lymphoid, hematopoietic, and related tissues was 1.54 (95% CI = 0.98–2.31), with a significantly higher SIR for NHL of 2.31 (95% CI = 1.23–3.95).

Our study reported a significant increase in lymphohematopoietic cancers among semiconductor workers, especially female operators. Several decades after this problem was identified across continents and countries, epidemiological evidence has been found supporting that the industry could increase the risk of lymphohematopoietic cancer. The semiconductor industry, historically, has faced challenges in occupational health due to the use of hazardous materials and chemicals [22]. Without definitive scientific proof of harm, the precautionary principle dictates that companies should take proactive measures to ensure worker safety. These measures may include implementing rigorous safety protocols, providing adequate protective gear, and carrying out routine health check-ups for workers [31, 32]. However, in the case of the semiconductor industry, protecting workers’ health was not given top priority due to the rapidly changing industrial environment and the introduction of new processes. These instances underscore the importance of the precautionary principle in occupational health, emphasizing proactive action to prevent harm rather than reactive measures after damage occurrence.

Our study reported myeloid leukemia, lymphoid leukemia, and non-follicular lymphoma among female operators in the wafer fabrication process who experienced excess mortality from malignant lymphoid, hematopoietic, and related tissues. The female operators died in their 20 s. The age distribution of this group is notable; they were classified as “healthy workers,” and their SMRs for all-cause mortality were significantly lower than the general population in both sexes. Leukemia, especially acute myeloid leukemia, occurs in older adults, with a reported median age ranging from 63 to 71 years [33, 34]. The demographic characteristics of the participants in our study were unusual for individuals with this type of cancer, implying that there may be a cancer cluster in this industry [35]. Although it was difficult to determine whether a definite exposure to carcinogenic chemicals had occurred, recent studies reported that some substances can emit benzene during a chemical reaction, which could lead investigators to infer an association between the semiconductor industry in South Korea and leukemia [7, 8, 36].

The semiconductor industry uses numerous chemicals; however, information on the chemicals that are potentially harmful to workers is limited. In 1981, the Semiconductor Industry Study reported many toxic materials and gases such as arsine, phosphine, and diborane [23]. The semiconductor industry is high-tech, competitive, and fast-growing, so the work condition and environment have changed rapidly. Because of secrecy, rapid change, and short history, workers may not notice the substances they use, and even experts could not fully understand the risks [36]. One of the potential risky chemical exposures is benzene. According to the International Agency for Research on Cancer monograph for chemical agents and the associated occupations, benzene, 1,3-butadiene, formaldehyde, and rubber have been suggested as carcinogens, whose hazardous effects have been sufficiently demonstrated in human studies [37]. The chemicals used in industries, such as TCE, could contain benzene due to the limited purity of the substances [38, 39]. Considering that a low-level benzene exposure of < 10 ppm-year shows a relative risk of 2.2 (95% CI = 1.1–4.2) among benzene-exposed Chinese, the exposure to benzene during a past time could be a possible plausible mechanism for the occurrence of excess mortality due to malignant neoplasms in the lymphoid, hematopoietic, and related tissues. Although the accurate duration of exposure needs to be determined in order to estimate the period of cumulative exposure to benzene [40], a threshold effect between benzene and leukemia implies that high-intensity exposure could be related to a shorter exposure period [41]. Furthermore, as workers in the semiconductor industry could have been exposed to multiple chemical mixtures of more than 500 chemicals, there are limitations to the approach investigating the simple association between exposure to specific risk factors and a specific disease [42]. However, in this study, the insufficient information regarding the level of exposure to potential risk factors and the short exposure window were obstacles to determining and concluding the causal relationship between work environments and the observed excess mortality due to leukemia.

Our study has several strengths. First, information on the changes in the workplace and job classification was obtained from the company. Therefore, the job categories can be classified accurately and completely. Second, as the target population consisted of employees from one semiconductor company, we could minimize the difficulty in interpreting the results compared with the results of employees from different companies.

However, our study has several limitations. First, although we constructed a cohort of 65,782 workers with 878,325 person-years, which is the largest population of employees in the semiconductor industry in a single company, our study has limited statistical power for determining the incidence of rare malignancies such as those occurring in the lymphoid, hematopoietic, and related tissues. Calculating the SMR and SIR is not recommended when only five or fewer cases are observed as the results are considered unreliable [43]. Although we observed ten female operators who died from lymphohematopoietic cancer, special attention should be paid to the interpretation of SMRs for malignant neoplasms in the lymphoid, hematopoietic, and related tissues. Second, the causes of death were possibly misclassified. However, the classification of the causes of death according to the NSO data had an overall accuracy rate of 91.9%, and the causes of death were commonly misclassified as “unusual, unnatural death.” The reliability of the causes of cancer-related deaths was higher than that of other causes [44]. Third, information on the changes in the working environment is lacking. Although the SMRs for malignant neoplasms in the lymphoid, hematopoietic, and related tissues increased, more detailed assessments such as the job exposure matrices could not be carried out in our study. Fourth, we classified workers according to the job they had worked in for most years in cases of shifting job types. Therefore, information on the workers’ jobs during the study period was summarized. Although this could have attenuated our findings, we believe that the direction of the attenuation was null. Fifth, leukemia is a group of different disease subtypes, including acute myeloid leukemia, acute lymphoblastic leukemia, chronic lymphocytic leukemia, and chronic myeloid leukemia [45, 46]. However, the reference mortality by 5-year age and sex group was not provided at the subcategory level; hence, we could not calculate the SMR for myeloid leukemia or lymphoid leukemia separately. Finally, the electronic human resource records from the company only included employee’s information from 1998 and beyond. The company’s semiconductor factory has continuously expanded and been renovated since the establishment of the production line at Giheung, Korea in 1983. Considering that the work environment and protective measures implemented to avoid exposure from hazardous chemicals have continuously improved in the company, more workers died from malignant neoplasms in the lymphoid, hematopoietic, and related tissues. Workers who developed work-related diseases may have left before the year 1998, leading to a potential selection bias in terms of the healthy worker survivor effect. The lack of human resource records for subcontractor employees who are potentially exposed to hazardous work environments is another limitation of our study. Additionally, we hope that social discussions incorporate an intersectional perspective. The semiconductor industry predominantly comprises educated female workers in their twenties. It is possible that susceptibility to occupational diseases may be related to social factors. While this epidemiological study does not focus on this specific aspect, we believe that understanding the social context is crucial for societal improvement and the prevention of occupational diseases.