Wojcik, Nancy C. MS; Schnatter, A. Robert PhD; Huebner, Wendy W. PhD
Mesothelioma is a cancer of the mesothelial cells, which form thin membranes that line the cavities of the chest (pleural), abdomen (peritoneal), heart (pericardium), and the surface of most organs.1 The most common anatomical site for malignant mesothelioma occurrence is the pleura, followed by the peritoneum.1 Malignant mesothelioma is primarily associated with occupational exposure to airborne asbestos1,2 and is known for its long latency of 20 to 40 years between exposure and appearance of disease.2–4
In the United States, occupational exposure to asbestos increased during World War II, most notably in shipbuilding, asbestos product manufacturing such as insulation, and construction operations.5,6 After peak use in the 1950s5,7 through the early 1970s, asbestos consumption declined over the next several decades2,4,6 as awareness of the hazard increased.
In many petroleum refineries, asbestos was used for thermal insulation, as a gasket material, and as protective screening during welding operations. These uses changed in the United States in the mid- to late-1960s when refineries began to replace asbestos with other materials.
Recognizing the potential hazard from asbestos exposure in earlier eras and the long latency of mesothelioma, we and others have conducted several cohort mortality studies to monitor mesothelioma risk in petroleum industry workers. Studies of early era petroleum industry workers in the United States and elsewhere8–18 report a 2- to 5-fold elevation in mesothelioma mortality with higher risks among subgroups of maintenance workers.9,10,13,15 Most decedents were hired in the 1950s or earlier, worked in refineries as insulators or pipe-fitters, and had long job tenure and latency (some 40 to 50+ years).
Given mesothelioma's long latency, continued surveillance among petroleum industry workers is warranted. Additional information can contribute to understanding the potential risk among workers hired after 1950 as they accumulate more follow-up time. Surveillance can also provide signals related to the effectiveness of current asbestos controls and signals related to potential risks from deterioration of contained asbestos and exposures to nonasbestos sources such as nanoparticles, which have not been well-studied.2,19
Accurate surveillance relies on the completeness of mesothelioma case capture. This is inherently difficult when studying mesothelioma mortality. First, mesothelioma lacked a specific cause-of-death code until the Tenth Revision of the International Classification of Diseases (ICD). In earlier ICD revisions, mesothelioma cases could be coded under other anatomical cancer sites. For example, in the ICD-9, mesothelioma cases could be coded as 158.9 for peritoneal cancer, 162.9 for lung cancer, 163.9 for pleural cancer, and 199.1 for cancer of unspecified or other sites. Nevertheless, these codes are not specific to mesothelioma. Second, because of the lack of specific coding, there are no national death rates for mesothelioma before 1999, making it difficult to assess patterns and trends of mesothelioma mortality among the general population, which frequently serve as the comparison population for worker cohorts.
These challenges have been acknowledged by others.6,20–25 Several studies have investigated the accuracy between histologically confirmed mesothelioma diagnoses and death certificate cause-of-death data to assess how well mesothelioma is captured using ICD codes.21,22,25,26 It is clear from these studies that relying on the ICD codes for malignant neoplasms of both the pleura and peritoneum fails to capture all malignant mesotheliomas.21,22,25,26
In the 1970s, investigators began applying modern life table techniques to study mortality in petroleum workers.27,28 Software used at the time29,30 did not tabulate expected numbers for mesothelioma, cancer of the pleura, and cancer of the peritoneum, because these outcomes were not included in the programs' default mortality rates. Thus, in early studies, mesothelioma deaths were interspersed among results for several default causes such as cancer of the lung and bronchus, other respiratory cancer, other digestive system cancer, benign neoplasms, and cancer of unspecified site (ICD-8 code 199).
The first study to examine mesothelioma and document an elevation (2.4-fold) in petroleum workers was reported in 1986, although ascertainment procedures were not detailed in this report.8 The first study to clearly document mesothelioma ascertainment procedures in a petroleum worker cohort9 was published shortly later, but case ascertainment procedures are not always specified in subsequent research.10–16
In this article, we describe our methods for more complete case capture and present results regarding the distribution of mesothelioma deaths by ICD code from two large cohort studies of men in North American operations. To test the accuracy of our ascertainment method, we compared New Jersey death certificate data from our company's health status registry with histologically confirmed cases from the New Jersey State Cancer Registry (NJSCR). The methods described ensure complete capture of mesothelioma diagnoses for comparison to mesothelioma incidence rates.
Mesothelioma Death Ascertainment
For mesothelioma death ascertainment, we conduct a manual review of all death certificates for any mention of mesothelioma. In this company, the practice of manually reviewing each death certificate for mention of mesothelioma was implemented in the late 1980s.
A nosologist, trained by the National Center for Health Statistics, developed a comprehensive list of ICD-7, ICD-8, and ICD-9 codes that could possibly contain a diagnosis of mesothelioma. Table 1 lists the ICD codes that we recognize as possible mesothelioma codes from the ICD-7, ICD-8, and ICD-7, as well as the recent ICD-10, that established specific, all inclusive codes for mesothelioma. The nosologist ascertains mesothelioma occurrence by manual review of the underlying and contributory causes on death certificates for any mention of mesothelioma. The nosologist then applies ICD rules using the revision in effect at the time of death to determine underlying cause-of-death codes (Table 1).
For all death records identified as mesothelioma, a mesothelioma flag is assigned in the company's health status registry system—a computerized system containing demographic, job, and mortality information for former and current US-based employees.31
Test for Agreement
To test for accuracy of our ascertainment method, we compared the New Jersey death certificate data from our health status registry with histologically confirmed mesothelioma cases from the NJSCR.
For the test cohort, we selected all male decedents in the health status registry with a New Jersey death certificate and year of death between 1979 and 2000. The death certificate and demographic data were obtained through company benefits and human resources. The decedents include members from previous company cohort studies, as well as more recent employees from all US business segments of the company. The period of time reflects the first year of available state/national cancer information (1979) to the last year of complete company death ascertainment (2000) in our latest published cohort study.32 Cause-of-death information, including underlying and secondary causes of death, was reviewed to distinguish mesothelioma deaths from all other causes of death.
Linkage was performed by NJSCR staff using AUTOMATCH probabilistic linkage software (MatchWare Technologies, Inc., Kennebunk, ME). Decedents were matched to NJSCR data, using social security number, name, sex, date of birth, and race (Fig. 1).
Company and NJSCR personnel manually reviewed matching results, using cancer site, cancer subgroup, histology codes, pathological findings, and dates of diagnosis, for case verification. Work location information within 5 years of death was included in the study database to help verify one type of nonmatch, that is, a person who died in New Jersey during the study period but who may have lived elsewhere when diagnosed with mesothelioma and later had an unspecified cause of death code on the New Jersey death certificate.
On completion of matching, name, social security number, day of birth, and day of diagnosis fields were deleted. Study identity numbers for the analytic file were randomly assigned to each study member. The protocol was approved by the University of Medicine and Dentistry of New Jersey institutional review board.
Test for Agreement
More than 330,000 US employee records in the health status registry were searched to construct the company test cohort. The test cohort consisted of 3834 male decedents with New Jersey death certificates from 1979 through 2000. Of these, 28 had been flagged as mesothelioma.
For the 28 company-identified mesothelioma diagnoses, 57% of the deaths occurred before 1990 (Table 2). Most decedents (75%) were older than 64 years, and 60% died within 12 months of diagnosis (Table 2).
Test cohort diagnoses were matched to 917,190 cancer cases diagnosed between 1979 and 2000 by the NJSCR. The results showed that 24 of the 28 company-identified mesothelioma deaths (cohort) were histologically confirmed as mesothelioma cancer cases by the NJSCR (Table 3).
Four cohort mesothelioma death certificate diagnoses did not match the NJSCR. Among these four, one decedent died during the first year of the NJSCR registration, so no earlier NJSCR records exist to help examine the accuracy of the death certificate diagnosis. Two decedents had records in the NJSCR shortly before their deaths but did not include mesothelioma cancer diagnosis. The fourth decedent did not have any NJSCR records and died well into the NJSCR catchment period. The absence of any NJSCR records along with the death certificate diagnosis of 199.1 (site unspecified) makes it difficult to confirm the accuracy of the lack of mesothelioma diagnosis on this death certificate. Overall, given the rapid progression of this disease, it is possible that a finite number of cases may not have been accurately diagnosed before death and would not have a registered mesothelioma diagnosis. Confidentiality procedures prevented further investigation of these four nonmatches.
Four other cohort death certificates did not contain a mesothelioma diagnosis but did have microscopic confirmation as malignant mesothelioma by the NJSCR (Table 3). It is reasonable to have some finite number of such nonmatches, given that not all histologically confirmed mesothelioma cases will be fatal and cases may die from underlying causes that are not related to mesothelioma. Confidentiality procedures prevented further analysis of these four nonmatches.
Altogether, the agreement between the company mesothelioma death ascertainment method and the NJSCR ascertainment is reasonably good, with a κ coefficient of 0.86 (95% confidence interval, 0.76 to 0.95). This exercise does not allow specification of either database as a criterion standard, or the precise degree of underascertainment present in the company mortality data for comparison with baseline mesothelioma incidence rates.
Mesothelioma Death Ascertainment by ICD Code
For the 28 company-identified mesothelioma deaths, most were coded to ICD 199.1 (other, unspecified site) and ICD 162.9 (bronchus and lung, unspecified) (Table 4). The same trend was observed among the 24 test cohort deaths that matched the NJSCR. Histologically, most of the matched cases were classified as pleura by the NJSCR. Because of the small sample size, the distribution of ICD codes and topography cannot be described in detail.
Test Cohort Characteristics
Albeit the small sample size, the characteristics of the 24 matched decedents mirror those observed for malignant mesothelioma mortality and incidence in the current US population. In the US population, for deaths from 1999 through 2005, malignant mesothelioma is most commonly reported among white men aged 75 years or older.2 Similar trends are reported for mesothelioma incidence rates in the United States from 2003 through 2008.33 Survival times for mesothelioma cases are reported to average between 4 and 18 months.1
Mesothelioma Death Ascertainment by ICD Code
Mesothelioma diagnoses are difficult to identify because of the lack of specific ICD coding before the Tenth Revision. Our test with the NJSCR demonstrates that mesothelioma diagnoses are frequently found within various ICD-9 codes and particularly in ICD-9 code 199.1, site unspecified (Table 4).
This is also confirmed by data from our latest North American company cohort studies,17,32 which may include some of the test cohort decedents. The cohort studies captured 88 mesothelioma deaths in our North American cohorts. The distribution of ICD codes is presented in Table 5. Among the 77 deaths coded to ICD-9, 60% are coded to 199.1, in which anatomical site is not specified (data not shown). This attests to the importance of manual review of all 199.1-coded death certificates to check for possible mesothelioma deaths. Most other ICD-9 codes were evenly distributed between unspecified pleura and unspecified bronchus and lung. Similarly, in another company study of two older refinery/chemical plant cohorts (with some overlap with the US cohort study included earlier), 12 of the 18 mesothelioma deaths (67%) from 1970 to 1997 are coded to 199.1.18 The predominance of mesothelioma diagnoses found within 199.1 is seen among earlier petroleum industry studies8,9,11 and within more current US mortality statistics, which report that nearly 75% of mesothelioma deaths are classified as unspecified anatomical site.2
Hence site-specific ICD-9 codes of 163 (pleura) and 158 (peritoneum) capture only a fraction of all mesothelioma deaths in the United States and Canada. In our test with the NJSCR, we found only 13% of all mesothelioma deaths coded to ICD 163 and 4% coded to 158. Similarly, in our latest North American cohorts, we found only 22% of all mesothelioma deaths coded to ICD 163 and none coded to ICD 158 (17 of a total 77 mesothelioma deaths coded to ICD-9). More recent US mortality statistics for malignant mesothelioma also report a lower proportion of deaths classified as pleura (less than 9%) and peritoneum (less than 4%).2 As such, limiting ascertainment to pleura and peritoneum codes, which has been a past practice in some studies, underestimates mesothelioma deaths. Although manually reviewing death certificates to ensure complete capture of mesothelioma diagnoses is reportedly done in some studies,9,11–13,15,23,32 it is unknown whether this practice has been adopted as a universal standard in all epidemiological studies. If manual review of all death certificates is not feasible, researchers can restrict their death certificate review to those coded to the possible mesothelioma codes listed in Table 1. For example, National Death Index Plus users can request death certificates for the National Death Index records containing these codes.
Topography findings from our test cohort suggest that when anatomical site is known, a higher proportion of mesothelioma cases are coded as pleural. This is consistent with recent malignant mesothelioma incidence trends, where pleura classifications made up 85% of all diagnoses during 2003 and 2008.33
Test for Agreement
Our test found reasonably good agreement between company mesothelioma ascertainment and that of the NJSCR (κ coefficient 0.86). This is consistent with a test done in Louisiana by another petroleum company, which found very good agreement between mortality and incidence counts for all cancers.34
It is reasonable to expect a finite number of nonmatches between the death certificate diagnoses and the diagnoses contained within the NJSCR. Nonmatches may have occurred for a few reasons: (1) misdiagnosis on the death certificate record that was used to identify mesothelioma, (2) miscoding of the NJSCR record, (3) death from competing causes after diagnosis of a mesothelioma, or (4) missing records from either source. We were unable to identify which of these reasons applied to each nonmatch.
The findings of this test are strengthened by having access to complete death record information for the 24 mesothelioma diagnoses in our sample. In addition, there was full cooperation between a well-regarded cancer registry and a company with an established, historical and current health status registry. We note that company and NJSCR personnel had no disagreement regarding what records matched or did not match. Merging of these two data sets has allowed us to test case capture as well as the comparability of our mesothelioma death counts with baseline incidence rates for comparison.
The main limitation of this test was our inability to completely research nonmatching cases. This was partly because of constraints in what data were available and/or included in the matching data set, and partly because of our mutual decision to remove all identifiers immediately after merging so that there could be no follow-back for either party. The investigation was also limited by the relatively shorter periods of time of cancer registration compared with longer periods of mortality ascertainment. Thus, we were unable to confirm or refute earlier mesothelioma death diagnoses.
Other Considerations—Comparison Rates
After identifying all possible mesothelioma cases, comparisons are made to baseline rates. For the earlier ICD revisions, researchers in the United States, as well as Australia,35 Canada,36 France,26 and Scotland,25 have used cancer incidence data obtained from national cancer registries to estimate the baseline mesothelioma mortality rate. In the United States, we and others8,11,14,15,21 have used invasive mesothelioma incidence rates from the National Cancer Institute's Surveillance, Epidemiology, and End Results 37 Program as a baseline to calculate standardized mortality ratios.
Because mesothelioma is not always fatal, baseline incidence rates will overestimate the expected number of deaths (reduce the standardized mortality ratio). Nevertheless, incidence rates are recommended5,15,24,26,35 because mesothelioma is typically fatal within 1 to 2 years of diagnosis.1 Thus, overascertainment bias is expected to be small, in the order of 5% to 10%, on the basis of 5-year survival rates.1 To more closely approximate mesothelioma incidence, we include underlying and secondary (contributory) causes of death in the observed number.
Mesothelioma incidence continues to be of interest, particularly in comparing risks in more recent periods of time, as asbestos controls became more widespread. Cohort updates and new studies of mesothelioma mortality risk in US industry studies should set goals of accurate and complete ascertainment and comparison to an appropriate control population.
To this end, and within the parameters of available information on death certificates, we recommend the following uniform strategy for North American industry studies:
1. To ascertain cases as fully as possible, manually review the wording on all death certificates for the relevant anatomical sites (for all ICD revisions, including the Tenth Revision). If this is impractical, review all death certificates under the codes listed in Table 1 as an efficient alternative.
2. Use incidence rates based on “invasive mesothelioma” from the Surveillance, Epidemiology, and End Results Program (United States) or comparable national cancer registries for comparisons to baseline (observed/expected).
3. Calculate risks on the basis of complete ascertainment as possible and compare to national incidence rates such as Canadian or the Surveillance, Epidemiology, and End Results Program.
Special thanks to Eugene Sales for his nosology expertise, Pamela Agovino and Thuy Lam for their assistance with the New Jersey State Cancer Registry data, and the dedicated professionals of the National Death Index and the Surveillance, Epidemiology, and End Results Program.
2. Centers for Disease Control and Prevention. Malignant mesothelioma mortality—United States, 1999–2005. MMWR Morb Mortal Wkly Rep. 2009;58:393–396.
3. Siemiatycki J, Bofetta P. Invited commentary: is it possible to investigate the quantitative relation between asbestos and mesothelioma in a community-based study? Am J Epidemiol. 1998;148:143–147.
4. Price B, Ware A. Mesothelioma trends in the United States: an update based on surveillance, epidemiology, and end results program data for 1973 through 2003. Am J Epidemiol. 2004;159:107–112.
5. Price B. Analysis of current trends in United States mesothelioma incidence. Am J Epidemiol. 1997;145:211–218.
6. Pinheiro GA, Antao VCS, Bang KM, Attfield MD. Malignant mesothelioma surveillance: a comparison of ICD 10 mortality data with SEER incidence data in nine areas of the United States. Int J Occup Environ Health. 2004;10:251–255.
7. Nishikawa K, Takahashi K, Karjalainen A, et al. Recent mortality from pleural mesothelioma, historical patterns of asbestos use, and adoption of bans: a global assessment. Environ Health Perspect. 2008;116:1675–1680.
8. Kaplan SD. Update of a mortality study of workers in petroleum refineries. J Occup Med. 1986;28:514–516.
9. Schnatter AR, Theriault G, Katz AM, Thompson FS, Donaleski D, Murray N. A retrospective mortality study within operating segments of a petroleum company. Am J Ind Med. 1992;22:209–229.
10. Gennaro V, Ceppi M, Boffetta P, Fontana V, Perotta A. Pleural mesothelioma and asbestos among Italian oil refinery workers. Scand J Work Environ Health. 1994;20:213–215.
11. Honda Y, Delzell E, Cole P. An updated study of morality among workers at a petroleum manufacturing plant. J Occup Environ Med. 1995;37:194–200.
12. Tsai SP, Gilstrap EL, Cowles SR, Snyder PJ, Ross CE. Long-term follow-up mortality study of petroleum refinery and chemical plant employees. Am J Ind Med. 1996;29:75–87.
13. Tsai SP, Waddell LC, Gilstrap EL, Ransdell J, Ross C. Mortality among maintenance employees potentially exposed to asbestos in a refinery and petrochemical plant. Am J Ind Med. 1996;29:89–98.
14. Tsai SP, Wendt JK, Cardarelli KM, Fraser AE. A mortality and morbidity study of refinery and petrochemical employees in Louisiana. Occup Environ Med. 2003;60:627–633.
15. Divine BJ, Hartman CM, Wendt JK. Update of the Texaco mortality study 1947-93: Part II. Analysis of specific causes of death for white men employed in refining, research, and petrochemicals. Occup Environ Med. 1999;56:174–180.
16. Gamble JF, Lewis RJ, Jorgensen G. Mortality among three refinery/petrochemical plant cohorts. II. retirees. J Occup Environ Med. 2000;42:730–736.
17. Lewis RJ, Gamble JF, Jorgensen G. Mortality among three refinery/petrochemical plant cohorts. I. 1970–1982 active/terminated workers. J Occup Environ Med. 2000;42:721–729.
18. Huebner WW, Wojcik NC, Rosamilia K, Jorgensen G, Milano CA. Mortality updates (1970–1997) of two refinery/petrochemical plant cohorts at Baton Rouge, Louisiana, and Baytown, Texas. J Occup Environ Med. 2004;46:1229–1245.
19. Lange JH. Re: “Mesothelioma trends in the United States: an update based on surveillance, epidemiology, and end results program data for 1973 through 2003” [letter]. Am J Epidemiol. 2004;160:823.
20. Spirtas R, Beebe GW, Connelly RR, et al. Recent trends in mesothelioma incidence in the United States. Am J Ind Med. 1986;9:397–407.
21. Connelly RR, Spirtas R, Myers MH, Percy CL, Fraumeni JF. Demographic patterns for mesothelioma in the United States. J Natl Cancer Inst. 1987;78:1053–1060.
22. Davis LK, Martin TR, Klingler B. Use of death certificates for mesothelioma surveillance. Public Health Rep. 1992;107:481–483.
23. Satin KP, Wong O, Yuan LA, et al. A 50-year mortality follow-up of a large cohort of oil refinery workers in Texas. J Occup Environ Med. 1996;38:492–506.
24. Steenland K, Burnett C, Lalich N, Ward E, Hurrell J. Dying for work: the magnitude of US mortality from selected causes of death associated with occupation. Am J Ind Med. 2003;43:461–482.
25. Camidge DR, Stockton DL, Bain M. Factors affecting the mesothelioma detection rate within national and international epidemiological studies: insights from Scottish linked cancer registry-mortality data. Br J Cancer. 2006;95:649–652.
26. Iwatsubo Y, Matrat M, Michel E, et al. Estimation of the incidence of pleural mesothelioma according to death certificates in France. Am J Ind Med. 2002;42:188–199.
27. Hanis NM, Stavraky KM, Fowler JL. Cancer mortality in oil refinery workers. J Occup Med. 1979;21:167–174.
28. Theriault G, Goulet L. A mortality study of oil refinery workers. J Occup Med. 1979;21:367–370.
29. Monson RR. Analysis of relative survival and proportional mortality. Comput Biomed Res. 1974;7:325–332.
30. Marsh GM, Preininger M. OCMAP: a user-oriented occupational cohort mortality analysis program. Am Stat. 1980;34:245.
31. Wojcik NC, Huebner WW, Jorgensen G. Strategies from using the National Death Index and the Social Security Administration for death ascertainment in large occupational cohort mortality studies. Am J Epidemiol. 2010;172:469–477.
32. Huebner WW, Wojcik NC, Jorgensen G, Marcella SP, Nicolich MJ. Mortality patterns and trends among 127,266 U.S.-based men in a petroleum company: update 1979–2000. J Occup Environ Med. 2009;51:1333–1348.
33. Henley SJ, Larson TC, Wu M, et al. Mesothelioma incidence in 50 states and the District of Columbia, United States, 2003–2008. Int J Occup Environ Health. 2013;19:1–10.
34. Tsai SP, Chen VW, Fox EE, et al. Cancer incidence among refinery and petrochemical employees in Louisiana, 1983–1999. Ann Epidemiol. 2004;14:722–730.
35. Threlfall TJ, Thompson JR, Olsen N. Cancer in Western Australia: Incidence and Mortality 2003 and Mesothelioma 1960–2003. Perth, Australia: Department of Health; 2005. Statistical series number 74.
36. Lewis RJ, Schnatter AR, Katz AM, et al. Updated mortality among diverse operating segments of a petroleum company. Occup Environ Med. 2000;57:595–604.
37. National Cancer Institute. Surveillance, Epidemiology and End Results (SEER) Program. Cancer statistics, cancer query systems, cancer incidence statistics. Available at http://seer.cancer.gov/canques/incidence.html
. Accessed November 16, 2012.
Copyright © 2014 by the American College of Occupational and Environmental Medicine