Original Articles
Inequality in cancer survival rates among industrial sectors in Japan: an analysis of two large merged datasets
2021 Volume 3 Issue 1 Article ID: 2020-0021-OA
Details
2021 Volume 3 Issue 1 Article ID: 2020-0021-OA
Objectives: Little is known about the specific prognosis of cancer among workers in different industrial sectors. The aim of this study is to demonstrate cancer survival inequality by industry sectors. Methods: Using multicenter inpatient data (1984−2017) and a regional cancer registry in Japan (1995–2018), we merged these two anonymized datasets. Based on standardized national classifications, cases were grouped according to the longest-held employment in primary, secondary, or tertiary industrial sectors. Data regarding smoking, alcohol consumption, and tumor staging at diagnosis were also extracted. We estimated the 5-year survival rates for common cancers using the Kaplan-Meier method to identify inequalities among industrial sectors. Cox proportional hazard model was used to calculate the hazard ratio (HR) of industry sectors. Results: A total of 13,234 cases were merged from two datasets. Among these, 8,794 cases were defined as common cancers (prostate, kidney, bladder, esophagus, stomach, liver, pancreas, colon, breast, and lung). Five-year survival was significantly (p=0.025) shorter for primary industrial sector (43.1%) compared with secondary sector (54.5%) and tertiary sector (56.9%). The adjusted HR for secondary and tertiary sectors versus primary sector was 0.963 (95% confidence interval [CI], 0.649–1.429). Bladder cancer in secondary and tertiary sectors showed a significantly higher survival rate than in the primary sector (p<0.0001), but the HR of secondary and tertiary sectors was 0.049 (95% CI, 0.021–0.153). Conclusions: This study revealed the potential of industrial sector inequalities with regard to the prognosis of cancers in Japan.
One’s occupation in which they are employed is a major social determinant of health1). It is well known that certain occupations are associated with increased disease risk and susceptibility to developing malignant neoplasms2,3,4,5,6). Previous studies have investigated the association between industry and cancer mortality and morbidity7,8,9,10). On the other hand, data on the relationship between industries and cancer survival rate are scarce11,12). Japan is believed to have less inequality in health because of more widespread health coverage of the population13). However, within the Japanese national workers’ compensation scheme, one’s lifetime occupational history influences cancer survival time due to differences in opportunities to diagnose disease and access to medical care10).
The global availability of access to occupational health services and medical information is unequally distributed14). Half of the world’s population is engaged in some form of employment and spends at least one-third of their time in the workplace, and only 15% of laborers have access to basic occupational health services15). It is important that occupational health services provide health education and promote lifetime healthcare15). Pragmatic worksite research and health surveillance into the relationship between occupation and disease is an important step toward developing an approach to reduce disease burden in the workplace16).
Industrial sectors differ with respect to working environment, occupational hazards and exposure, personal health consciousness, and access to medical care. In particular, primary industry workers have a pattern of mortality and disease prevalence distinct from that of other populations17). It is likely that disease characteristics differ among industrial sectors.
The aim of the present study was to evaluate inequalities in cancer survival among different industrial sectors in Japan. Several large governmental statistical databases can be analyzed; however vital information is often missing from such records. We merged two anonymized large datasets, one of which included data on survival time but not occupational information, and the other included occupational information but not data about death. The results provide a significant step forward in elucidating some of the key social factors underlying the relationship between cancer prognosis and industrial risk factors.
We used datasets from the Inpatient Clinico–Occupational Database of the Rosai Hospital Group (ICOD-R) and Kanagawa Cancer Registry (KCR). ICOD-R was produced by the Japan Organization of Occupational Health and Safety (JOHAS), an independent administrative agency in Japan. Rosai Hospitals now comprise 33 hospitals, having grown to cover Hokkaido to Kyushu, in both rural and urban areas, since the establishment of Rosai Hospital Group by the Ministry of Labour of Japan in 1949.
The database contains medical chart information overseen by physicians. The diagnoses were coded according to the International Statistical Classification of Diseases and Related Health Problems, 9th Revision (ICD-9) or 10th Revision (ICD-10). The ICOD-R also included job history (current and past three jobs) as well as smoking and alcohol habits, and history of lifestyle-related diseases. The information is gathered using interviews and a questionnaire completed by patients at the time of admission. Detailed job histories were coded using 3-digit codes according to the standardized national classification of the Japan Standard Occupational Classification and Japan Standard Industrial Classification, which corresponds to the International Standard Industrial Classification and International Standard Occupational Classification. Written informed consent was obtained before patients completed the questionnaires. Trained registrars were in charge of registering the data.
Regarding the KCR, Kanagawa Prefecture is a neighboring prefecture of Tokyo and is the second largest in Japan, with a population of roughly 9 million. Kanagawa Prefecture started its own Regional Cancer Registry from 1970, with the accumulated number of cases being approximately 1,200,000 by December 31, 2018. Details on the cancer registry system in Japan have been reported elsewhere18). Data were collected from neoplasm registration sheets reported by each diagnosing hospital or from clinic and death certificates of residents in Kanagawa Prefecture. The Kanagawa Prefectural Cancer Center collected and consolidated the data into anonymous formats and made these available for academic and administrative purposes.
Accumulated data included the following items: 1) personal identification code, 2) method of registry entry, 3) diagnosing institution, 4) sex, 5) date of birth, 6) date of diagnosis, 7) local government code for the patient’s home address, 8) ICD-10 code for disease name, 9) ICD-O-3 code for pathology, 10) initial or recurrent tumor, 11) therapeutic strategy (very brief), 12) operative procedure (if any), 13) date of death, 14) cause of death, 15) date of last follow-up, and 16) TNM classification and pathological grade according to ICD-O-3 in diagnosed patients. The reporting of TNM classifications became mandatory in 2005. All information was collected by individuals trained in Japan by the Surveillance, Epidemiology, and End Results (SEER) program of the National Cancer Institute in the United States. Information was updated every year from vital statistics and death certificates. The proportion of death-certificate-notification (DCN) cases in the whole database was 8.9% by the end of 201519).
Merge of datasetWe obtained an anonymous dataset from the JOHAS that included hospital admissions between 1984 and 2017 and a dataset from the government of Kanagawa Prefecture that registered subjects between 1998 and 2018. Since all individual data were anonymized by the JOHAS and government of Kanagawa Prefecture, we could not identify specific individuals. From the ICOD-R, we selected the Kanagawa Prefecture area, which comprised registered cases that had been hospitalized at Kanto and Yokohama Rosai Hospital. We conducted a deterministic linkage analysis between the two datasets to extract data on the following: diagnosed or died hospital, sex, month and year of birth, zip-code, and ICD-10. We excluded unmatched samples from the analysis. Finally, we extracted data for those who were diagnosed with cancer between January 1998 and December 2017.
Cancer sites and sectors of industryWith respect to the cancer location, we selected the 10 most common cancers according to national statistics in Japan20,21,22): prostate, breast, kidney, bladder, esophagus, stomach, liver, pancreas, colon, and lung. The prevalence rates of these cancers matched almost identically with Japanese national statistics23).
To categorize each industry and industrial sector, we identified the longest-held job for each patient from their industrial history9,24). Primary industrial sectors included agriculture, forestry, and fisheries; secondary industrial sectors included mining, construction, and manufacturing; and tertiary industrial sectors included all other industries, according to the Japan Standard Occupational Classification system.
Statistical analysisThe 5-year survival rate was estimated using the Kaplan–Meier method for each industrial sector. Cox proportional hazard models were used to calculate the adjusted HR for death of secondary and tertiary sectors versus the primary sector, adjusted for basic characteristics (sex, age, Brinkman Index, and amount of alcohol consumption). The Brinkman Index was calculated as the number of cigarettes smoked per day multiplied by the number of years of cigarette smoking.
All p-values were 2-sided, and p<0.05 was considered statistically significant. All analyses were conducted using STATA/MP15.0 software (Stata Corp LP, College Station, TX, USA).
The study was approved by the ethics committees of The University of Tokyo (No.10891) and Kanto Rosai Hospital (No.2019-26).
The merging of cases is summarized in the flow chart shown in Figure 1. A total of 6,526,387 inpatient cases were registered in the ICOD-R from 1984 to 2017. Among these, complete data, which included birthday, sex, ICD-9 or ICD-10 code, history of smoking, and alcohol consumption, were available for 6,309,852 cases, and of those cases, 298,712 were first-time cancer admissions. Of those, 29,803 cases were registered as having been hospitalized at Kanto and Yokohama Rosai Hospital, which is located in Kanagawa Prefecture. A total of 837,164 cases were registered in the KCR from 1998 to 2018, 27,698 of which were diagnosed or died at Kanto and Yokohama Rosai Hospital. In total, 13,234 cases were merged. Of those cases, 8,794 cases fell within the 10 main cancer categories (i.e., prostate, breast, kidney, bladder, esophagus, stomach, liver, pancreas, colon, and lung), 4,630 cases had available industrial history data, and TNM staging was known for 1,546 cases.
The data extraction flowchart.
Table 1 shows the baseline characteristics for cases with available historical data for industrial sector of employment. The median age at onset for breast cancer was younger than that of the other cancer sites. The Brinkman Index and amount of alcohol consumption were high for esophagus cancer and lung cancer (Brinkman Index, 432 and 550; alcohol consumption, 375 g/day and 150 g/day, respectively). The ages of cancer onset by sectors were: primary, 76 years; secondary, 66 years; and tertiary, 63 years. Tertiary sector workers had the lowest Brinkman Index and amount of alcohol consumption.
| Gender | Age | Brinkman Index† | Alcohol (g/day) | ||
|---|---|---|---|---|---|
| Male (%) | Female (%) | (median:IQR‡) | (median:IQR‡) | (median:IQR‡) | |
| Total | 3,431 (74.1) | 1,199 (25.9) | 65 (56:72) | 100 (0:760) | 0 (0:620) |
| Cancer category | |||||
| Prostate (C619) | 409 (100) | – (–) | 64 (58:71) | 100 (0:680) | 0 (0:620) |
| Kidney (C649) | 207 (83.5) | 41 (16.5) | 61 (50:70) | 120 (0:640) | 0 (0:540) |
| Bladder (C670/679) | 330 (87.5) | 47 (12.5) | 67 (57:75) | 220 (0:780) | 0 (0:660) |
| Esophagus (C150/159) | 166 (92.2) | 14 (7.8) | 66 (59:73) | 432 (0:970) | 375 (0:900) |
| Stomach (C160/169) | 573 (80.8) | 136 (19.2) | 66 (57:73) | 165 (0:760) | 0 (0:660) |
| Liver (C220) | 191 (86.0) | 31 (14.0) | 66 (60:73) | 9 (0:660) | 0 (0:600) |
| Pancreas (C250/259) | 127 (69.8) | 55 (30.2) | 68 (60:74) | 0 (0:840) | 0 (0:620) |
| Colon (C180/189/199/209) | 809 (75.8) | 258 (24.2) | 65 (57:72) | 160 (0:700) | 0 (0:600) |
| Breast (C500/509) | 3 (0.7) | 430 (99.3) | 52 (46:62) | 0 (0:0) | 0 (0:0) |
| Lung (C340/349) | 616 (76.7) | 187 (23.3) | 66 (58:73) | 550 (0:1000) | 150 (0:860) |
| Industry sector | |||||
| Primary | 43 (69.4) | 19 (30.7) | 76 (68:82) | 255 (0:900) | 0 (0:880) |
| Secondary | 1,746 (85.1) | 307 (14.9) | 66 (58:74) | 280 (0:820) | 0 (0:700) |
| Tertiary | 1,642 (65.3) | 873 (34.7) | 63 (54:70) | 0 (0:700) | 0 (0:555) |
Because of rounding, percentages may not total 100.
Table 2 shows TNM staging data according to industrial category for each cancer site. The distributions of tumor staging were similar among industrial sectors for all cancer sites. Among pancreas cancer cases, the percentage of stage 4 cases was high in the secondary and tertiary sector categories (61.3% and 70.3%, respectively). Among esophagus cancer cases, the proportion of advanced tumor stages was higher for secondary (41.2%) versus tertiary sector (21.1%).
| Cancer site (ICD-10) | |||||||
|---|---|---|---|---|---|---|---|
| Prostate (C619) Industry sector | Primary | Secondary | Tertiary | Kidney (C649) Industry sector | Primary | Secondary | Tertiary |
| Total n (%) | 1 (0.5) | 93 (46.5) | 106 (53.0) | Total n (%) | 0 (0) | 42 (47.2) | 47 (52.8) |
| Gender n (%) | Gender n (%) | ||||||
| Male | 1 (0.5) | 93 (46.5) | 106 (53.0) | Male | 0 (0) | 36 (85.7) | 37 (78.7) |
| Female | − | − | − | Female | 0 (0) | 6 (14.3) | 10 (21.3) |
| Age (y) Median (IQR†) | 79 (69:84) | 67 (59:74) | 64 (55:72) | Age (y) Median (IQR†) | − | 67 (59:74) | 64 (55:72) |
| TNM stage n (%) | TNM stage n (%) | ||||||
| Stage 1 | 0 (0) | 48 (51.6) | 42 (39.6) | Stage 1 | 0 (0) | 32 (76.2) | 33 (70.2) |
| Stage 2 | 1 (100) | 29 (31.2) | 33 (31.1) | Stage 2 | 0 (0) | 2 (4.8) | 3 (6.4) |
| Stage 3 | 0 (0) | 4 (4.3) | 7 (6.6) | Stage 3 | 0 (0) | 4 (9.5) | 5 (10.6) |
| Stage 4 | 0 (0) | 12 (12.9) | 24 (22.6) | Stage 4 | 0 (0) | 4 (9.5) | 6 (12.8) |
| Brinkman Index | 490 (−:−) | 230 (0:620) | 180 (0:720) | Brinkman Index | − | 310 (0:800) | 10 (0:500) |
| Alcohol (g/day) | 60 (−:−) | 80 (0:600) | 55 (0:660) | Alcohol (g/day) | − | 240 (0:645) | 0 (0:465) |
| Bladder (C670/679) Industry sector | Primary | Secondary | Tertiary | Esophagus (C150/C159) Industry sector | Primary | Secondary | Tertiary |
| Total n (%) | 1 (1.6) | 32 (52.5) | 28 (45.9) | Total n (%) | 0 (0) | 17 (47.2) | 19 (52.7) |
| Gender n (%) | Gender n (%) | ||||||
| Male | 1 (100) | 29 (90.6) | 25 (85.4) | Male | 0 (0) | 15 (66.2) | 18 (94.7) |
| Female | 0 (0) | 3 (9.4) | 3 (10.7) | Female | 0 (0) | 2 (11.7) | 1 (5.2) |
| Age (y) Median (IQR†) | 79 (69:84) | 67 (59:74) | 64 (55:72) | Age (y) Median (IQR†) | − | 67 (59:74) | 64 (55:72) |
| TNM stage n (%) | TNM stage n (%) | ||||||
| Stage 1 | 0 (0) | 19 (59.4) | 13 (46.4) | Stage 1 | 0 (0) | 5 (29.4) | 10 (52.6) |
| Stage 2 | 0 (0) | 10 (31.3) | 8 (28.6) | Stage 2 | 0 (0) | 2 (11.7) | 0 (0) |
| Stage 3 | 1 (100) | 2 (6.3) | 2 (7.2) | Stage 3 | 0 (0) | 3 (17.6) | 5 (26.3) |
| Stage 4 | 0 (0) | 1 (3.1) | 5 (17.9) | Stage 4 | 0 (0) | 7 (41.1) | 4 (21.1) |
| Brinkman Index | 0 (−:−) | 650 (275:940) | 640 (180:820) | Brinkman Index | − | 900 (400:1050) | 600 (30:800) |
| Alcohol (g/day) | 0 (−:−) | 540 (0:812) | 527 (160:780) | Alcohol (g/day) | − | 800 (412:980) | 400 (0:756) |
| Stomach (C160/169) Industry sector | Primary | Secondary | Tertiary | Liver (C220) Industry sector | Primary | Secondary | Tertiary |
| Total n (%) | 4 (1.7) | 110 (45.5) | 128 (52.8) | Total n (%) | 0 (0) | 30 (42.9) | 40 (57.1) |
| Gender n (%) | Gender n (%) | ||||||
| Male | 4 (100) | 103 (93.6) | 95 (74.2) | Male | 0 (0) | 28 (93.3) | 26 (65) |
| Female | 0 (0) | 7 (6.4) | 33 (25.8) | Female | 0 (0) | 2 (6.7) | 14 (35) |
| Age (y) Median (IQR†) | 79 (69:84) | 67 (59:74) | 64 (55:72) | Age (y) Median (IQR†) | − | 65 (59:71) | 67 (61:75) |
| TNM stage n (%) | TNM stage n (%) | ||||||
| Stage 1 | 1 (25) | 75 (68.2) | 71 (55.5) | Stage 1 | 0 (0) | 11 (36.7) | 15 (37.5) |
| Stage 2 | 2 (50) | 15 (13.6) | 19 (14.8) | Stage 2 | 0 (0) | 9 (30.0) | 7 (17.5) |
| Stage 3 | 0 (0) | 6 (5.5) | 9 (7.0) | Stage 3 | 0 (0) | 5 (16.7) | 10 (25) |
| Stage 4 | 1 (25) | 14 (12.7) | 29 (22.7) | Stage 4 | 0 (0) | 5 (16.7) | 8 (20) |
| Brinkman Index | 670 (355:900) | 620 (315:980) | 198 (0:740) | Brinkman Index | − | 395 (180:820) | 120 (0:830) |
| Alcohol (g/day) | 355 (35:870) | 540 (0:920) | 112 (0:720) | Alcohol (g/day) | − | 370 (0:680) | 0 (0:830) |
| Pancreas (C250/259) Industry sector | Primary | Secondary | Tertiary | Colon (C180/189,C199,C209) Industry sector | Primary | Secondary | Tertiary |
| Total n (%) | 0 (0) | 31 (45.6) | 37 (54.4) | Total n (%) | 4 (1.3) | 121 (40.1) | 177 (58.6) |
| Gender n (%) | Gender n (%) | ||||||
| Male | 0 (0) | 28 (90.3) | 17 (46.0) | Male | 3 (75) | 100 (82.6) | 119 (67.2) |
| Female | 0 (0) | 3 (9.7) | 20 (54.1) | Female | 1 (25) | 21 (17.4) | 58 (32.8) |
| Age (y) Median (IQR†) | − | 67 (49:74) | 64 (55:72) | Age (y) Median (IQR†) | 75 (60:82) | 66 (59:73) | 65 (57:73) |
| TNM stage n (%) | TNM stage n (%) | ||||||
| Stage 1 | 0 (0) | 2 (6.5) | 4 (10.8) | Stage 1 | 0 (0) | 33 (27.3) | 50 (28.3) |
| Stage 2 | 0 (0) | 5 (16.1) | 5 (13.5) | Stage 2 | 2 (50) | 31 (25.6) | 44 (24.9) |
| Stage 3 | 0 (0) | 5 (16.1) | 2 (5.4) | Stage 3 | 2 (50) | 30 (24.8) | 43 (24.3) |
| Stage 4 | 0 (0) | 19 (61.3) | 26 (70.3) | Stage 4 | 0 (0) | 27 (22.3) | 40 (22.6) |
| Brinkman Index | − | 500 (0:1050) | 0 (0:720) | Brinkman Index | 280 (0:720) | 340 (0:760) | 240 (0:697) |
| Alcohol (g/day) | − | 460 (0:900) | 0 (0:320) | Alcohol (g/day) | 0 (0:440) | 300 (0:645) | 100 (0:580) |
| Breast (C500/509) Industry sector | Primary | Secondary | Tertiary | Lung (C340/349) Industry sector | Primary | Secondary | Tertiary |
| Total n (%) | 1 (0.6) | 42 (24.0) | 132 (75.4) | Total n (%) | 2 (0.7) | 119 (39.3) | 182 (60.1) |
| Gender n (%) | Gender n (%) | ||||||
| Male | 0 (0) | 0 (0) | 1 (0.8) | Male | 2 (100) | 107 (89.9) | 129 (70.9) |
| Female | 1 (0.6) | 42 (24.1) | 131 (99.2) | Female | 0 (0) | 12 (10.1) | 53 (29.1) |
| Age (y) Median (IQR†) | 50 (−:−) | 60 (46:72) | 52 (46:63) | Age (y) Median (IQR†) | 80 (73:87) | 69 (62:76) | 66 (59:73) |
| TNM stage n (%) | TNM stage n (%) | ||||||
| Stage 1 | 1 (100) | 14 (33.3) | 63 (47.7) | Stage 1 | 0 (0) | 40 (33.6) | 67 (36.8) |
| Stage 2 | 0 (0) | 19 (45.2) | 54 (40.9) | Stage 2 | 1 (50) | 7 (5.9) | 10 (5.5) |
| Stage 3 | 0 (0) | 5 (11.9) | 11 (8.3) | Stage 3 | 1 (50) | 37 (31.1) | 44 (24.2) |
| Stage 4 | 0 (0) | 4 (9.5) | 4 (3.0) | Stage 4 | 0 (0) | 35 (29.4) | 61 (33.5) |
| Brinkman Index | 0 (−:−) | 0 (0:0) | 0 (0:0) | Brinkman Index | 905 (−:−) | 800 (460:1060) | 705 (200:1000) |
| Alcohol (g/day) | 0 (−:−) | 0 (0:0) | 0 (0:0) | Alcohol (g/day) | 885 (−:−) | 720 (0:980) | 475 (0:920) |
Because of rounding, percentages may not total 100.
Among the cases with available industry history, the 5-year survival rates for each cancer site (Figure 2) were as follows: prostate, 74.6%; kidney, 79.0%; bladder, 75.8%; esophagus, 28.8%; stomach, 58.3%; liver, 21.6%; pancreas, 9.9%; colon, 61.3%; breast, 86.9%; and lung, 32.4%. Figure 3a shows the 5-year survival rates for each industrial sector among all cases. The survival rate was primary sector, 43.1%; secondary sector, 54.5%; and tertiary sector, 56.9% (log-rank test: p=0.025). The HR estimated using Cox regression analysis for secondary and tertiary versus primary sector was 0.963 (95% confidence interval [CI], 0.649–1.429) with adjustment for age, Brinkman Index, and alcohol intake. The adjusted HR of age at diagnosis was 1.028 (95% CI, 1.024–1.033). Figure 3b shows survival estimation limited to case with available TNM staging. The survival rate was primary sector, 23.8%; secondary sector, 50.7%; and tertiary sector, 51.8% (log-rank test: p=0.209).
Kaplan-Meier survival curves for overall survival among 10 most common cancers in merged cases. Survival was estimated using the Kaplan-Meier method in patients with complete information on sex, age, and observation period, and with right censoring at the 5-year mark.
a) Kaplan-Meier survival curves for overall survival among cancer cases with industrial history. The survival curve for primary industry was lower than that of the others (log-rank test: p=0.025) even though the hazard ratio for industry category showed no statistically significant difference. b) Kaplan-Meier survival curves for overall survival in cancer cases with industrial history and TNM staging. The survival curve for primary industry was again lower than that of the others (log-rank test: p=0.209).
Figure 4 shows survival curves for each industrial sector according to cancer site. The 5-year survival rates for bladder cancer were as follows: primary, 0%; secondary, 77.0%; tertiary, 76.1% (log-rank test: p<0.000). The same trend was observed for stomach cancer: primary sector, 40.0%; secondary sector, 60.7%; and tertiary sector, 56.9% (log-rank test: p=0.099). The adjusted HR for secondary and tertiary sectors versus the primary sector was 0.049 (95% CI, 0.021–0.153) and 0.672 (95% CI, 0.355–1.340), respectively. The adjusted HR for age at diagnosis was 1.057 (95% CI, 1.036–1.079) and 1.023 (95% CI, 1.012–1.035). No characteristic findings were found in other cancer types.
Kaplan-Meier survival curves for specific cancer by industry category. Survival rates in primary industry were significantly lower in bladder and gastric cancers compared with other industry categories.
The prognosis for primary industry tended to be short compared with the other sectors. The adjusted HR of primary industry sector was not statistically significant. Notably, the age at diagnosis was higher for primary industry sector versus the other sectors and the HR was significantly related to 5-year survival.
The strongest reason for survival rate inequalities among industrial sectors is most likely the aging population of agricultural workers. Among the Japanese labor force engaged in a primary industry sector, the percentage of workers aged 70 years or older increased from 12.8–19.0% in 2008 to 37.7% in 201825). Agricultural work is more likely to be a life-long occupation than other occupations. Given these growth trends in the elderly population, super-aging of primary industrial sector workers may accelerate further in the future. As a result, the cancer prognosis is likely to worsen compared with other industrial sectors. In our study, age at diagnosis in the primary sector was approximately 10 years older than in the other sectors. This might strongly contribute to a poor prognosis.
Other factors concerning the primary sector must be taken into account with respect to the results of this study. First, the sector must be considered from the point of social structure. Uneven access to medical resources is a major challenge that is encountered worldwide26). Inequality of access to medical information among industries is an inevitable problem that may deeply influence the chances of detecting cancer and its prognosis. One strategy to correct these inequalities is an approach thorough occupational healthcare services. In Japan, companies with 50 or more employees are required by the Occupational Health and Safety Law to employ an occupational health physician, and annual health examinations are required for all employers. This is vital for protecting health in the workplace and promoting general health, well-being, and work ability, as well as preventing illnesses and accidents. The provision of occupational healthcare services in Japan includes all industrial sites, in theory. However, in reality, low access to healthcare services is common within the primary industrial sector, particularly those employed by family-run farms, which typically have only 1 to 3 employees15). The inequality of workplace services in Japan across different occupations and industrial sectors may strongly influence the incidence of different types of cancer and their prognosis. Easy access to medical information enables patients to search for and choose the most appropriate treatment options available26). Family-run farms may be insufficient resources for health activities as compared with large companies in secondary or tertiary sectors. Most are not included in special periodic health examinations, involving health guidance by medical personnel, which is normally provided to enterprise workers. Inequalities in the lack of occupational health interventions such as occupational physician monitoring, medical checkups, and health guidance due to being employed are expected to exist. It is necessary to assess the healthcare status of primary industries in some cases by collating the national cancer registries, including rural areas, with the ICOD-R.
Second, exposures to chemical hazards may influence survival inequality. In our previous study, the risk of most cancers in agriculture workers was low compared with other occupations10). However, the Agricultural Health Study, the largest prospective cohort study of agricultural workers, demonstrated the relationship between individual pesticides and the development of cancer27). Those who are self-employed are not dismissed by an employer for having cancer28), and exposure to hazardous pesticides and other hazardous factors after the diagnosis may also influence the prognosis after cancer detection.
Third, with regard to general risk factors for cancers, smoking and drinking are critically important. Previously, we reported that both the cancer risk and rate of smoking are low in primary industry10). It has been reported that non-smoking policies in the workplace do not differ between primary and secondary industries29), but workers in primary industries have a higher mortality risk for lung, gastric, and colorectal cancers compared with workers in manufacturing9). Studies have examined risks associated with work-related secondhand tobacco smoke exposure30). In particular, the blue-collar sector may be more lax with regard to limiting secondhand tobacco smoke exposure31). In the present study, esophagus and lung cancer were associated with a high amount of smoking and alcohol consumption, especially in the secondary sector. Esophagus cancer in the secondary sector tended to be at the advanced stage at the time of diagnosis. Considering the etiology of esophagus cancer, this result could be related to the amount of smoking or alcohol consumption.
To our knowledge, this is the first study to investigate inequalities among industrial sectors in relation to cancer survival in Japan. The strength of this study is that we were able to combine two completely independent large databases. Two datasets had to be merged to analyze individual industrial employment history and information about length of survival. Every 5 years, in the same year as the national population census, the government collects information on occupation as well as cause of death based on death certificates submitted to local governments32). However, because the occupational information is registered by family members after the patient’s death, it is likely that they reported the last job held just prior to the patient’s death rather than the job they held the longest during their lifetime, possibly resulting in recall bias33). The present merged dataset enabled us to estimate a relationship between a lifework occupation and disease.
The limitations of the study were as follows. First, because the ICOD-R was not a relevant population-based database, the dataset may have had a selection bias; i.e., one-third of the occupational data was unavailable, TNM was registered in only half of the cases. That could have amplified selection bias. This problem arises because it is not mandatory for patients to provide employment data in order to protect their privacy. In addition, we were able to use cases from only two hospitals because the cancer registry data were limited solely to Kanagawa Prefecture. This may have resulted in a low percentage of primary industrial sector cases because Kanto and Yokohama Rosai Hospital is located in a large urban city. Due to missing data, the power of the statistical analysis may have been insufficient for industrial sector categories.
We must also consider the limits of accuracy for the percentage of cases that are linked. There were 2105 more cases with extractable data in the ICOD-R than in the KCR. Considering that the ICOD-R includes cases that were neither diagnosed nor died at the Rosai hospitals but were hospitalized as an intermediate stage, we believe that almost all cases in the KCR should be matched to the ICOD-R. In this study, only about half of the data were matched. We believe this poses a limitation regarding a deterministic linkage34). Follow-up research should involve performing a probabilistic linkage using more conditions.
Second, our study was designed to enable us to assess exposure to hazards related to cancers. Exposure to chemical hazards, overwork, day and night shifts, and socioeconomic status related to occupational have a profound influence on cancer risk12,35). Other relevant factors, such as pathogenic organisms (i.e., Helicobacter pylori in stomach cancer, hepatitis virus and fluke in liver cancer) could not be evaluated due to limited availability of data. In this regard, ICOD-R has the unique feature of being able to collect information back to the hospitalization summary while remaining anonymized36). This task requires an enormous amount of manpower. However, when we are able to conduct a nationwide collation in the future, we will need to go back and supplement the missing data with summaries in order to improve the validity of results.
Last, evaluating industrial risk by using data from the longest-held job may lead to bias. The occupations used in this study were necessarily the occupations in which individuals were mainly engaged throughout their lifetime. An alternate method for assessment of industry risk involves choosing the occupation of the deceased at the time of death9,24). However, this may not always be the most relevant factor for cancer risk. The length of engagement, the occupations held over an entire lifetime, and incubation time to development of cancer must be taken into consideration in future investigations.
We have documented the possibility of inequalities among industrial sectors with respect to the prognosis of cancer in Japan. On the basis of the information gathered, explicit strategies to facilitate access to health resources, such as those in primary industries, small and family-run businesses, and among elderly laborers, may improve the prognosis of cancer as well as the general health of all industrial sectors.
R.K: Conceptualization, formal analysis, writing original draft, review and editing, funding acquisition. Y.S: Supervision. Y.K: Funding acquisition and supervision.
The authors declare that there are no conflicts of interest.
This work was supported by Industrial Disease Clinical Research Grants (No. 170201-01) and Research Funds to Promote Hospital Functions from the Japan Organization of Occupational Health and Safety.
