8
Exposure Definition, Measurement, and Verification
Introduction
Chapter 6 describes at length how the participant cohort was defined and assembled for this study. Status as an Operation CROSSROADS participant is the most reliable (though broad) measure of exposure and, therefore, the core variable on which to compare groups in this study. The participant vs. nonparticipant (comparison group) dichotomy provides the largest group of people to study, size being important when considering rare outcomes such as leukemia. Using participant status to represent exposure, in this case exposure to an atmospheric nuclear test (including ionizing radiation and possibly other environmental factors) has many limitations for use in an epidemiologic study. The fact of participant status does not provide individual-level information regarding:
- differences in potential radiation exposure among participants (see note 1 for the Federal Register's definition of participation);
- potential for internal as well as external radiation exposure;
- extent, if any, of participation (and, therefore, potential exposure) in test series other than CROSSROADS;
- exposure to other ill-defined environmental or occupational factors, either related to or independent of nuclear test participation (for example, post-
- CROSSROADS employment as a radiological technician or a radiation worker in the nuclear power industry).
Therefore, while retaining as basic the participant vs. nonparticipant comparison, we attempted to develop more refined measures of radiation exposure. After providing a historical backdrop to the dosimetry issues, this chapter describes the approaches we considered, details about the validity of those options, and the decisions upon which we base our analyses.
Dosimetry Background
The National Research Council's 1985 study (Robinette et al. 1985) used dose data provided by the Defense Nuclear Agency's (DNA) Nuclear Test Personnel Review program, which attempted to assign to each individual participant a valid estimate of the radiation dose he had received (DNA 1984). Therefore, the initial plans for the Medical Follow-up Agency (MFUA) study of CROSSROADS participants reported here included the use of these individual dose assignments.
Later, a committee of another MFUA nuclear test exposure project17 reviewed the DNA dosimetry estimating procedures and results and issued a letter report (IOM 1995), stating that the dosimetry estimates were not appropriate for dose response analyses used in epidemiologic studies. The CROSSROADS committee and staff, based on that letter report and their own judgment, decided not to use the individual dose data in this study. We set out to explore other exposure proxies, determining how they correlated with each other and with the DNA-assigned individual dose estimates.
Definition of Potential Surrogate Dose Groups
We consider four broad ways in which to categorize exposure:
- DNA-assigned doses,
- participant status,
- target ship boarding status, and
- Engineering & Hull specialty status
Each of these adds intuitive interpretive possibilities to the analysis, while at the same time involving a wide range of validity and precision. We discuss each in turn, offering definitions, advantages, and both practical and theoretical drawbacks to use as a dose surrogate.
DNA-Provided Dose Estimates
Overall
The Nuclear Test Personnel Review (NTPR) database contains a dose assignment for each participant derived, in most cases, by reconstruction based on duty assignments. In a small percentage of cases, the assigned dose is based on one or more film badges worn by the participant or on a film badge worn by another participant in the same unit at CROSSROADS (cohort badging).
Ideally, exposure measurements would be (a) individual-specific; (b) recorded by time, duration, and dose; (c) sensitive to different components of exposure (e.g., alpha, beta, or gamma rays); (d) validated in similar conditions prior to their use, (e) quantitative and at least theoretically reproducible (f) complete, in that they covered all exposures for all involved people; and (g) accepted by all interested parties. As stated above, based on our own examination of the DNA dosimetry data, their consideration by others, we do not believe the data are appropriate for the individual-specific assignments necessary for the type of epidemiologic comparisons on which this report is based. We offer our justification of this decision in this section, along with a discussion of our consideration of alternative measures of dose, using subsets of the dosimetry data.
A Working Group of the Five Series Study Committee assessed the basis and quality of the data upon which dose assignments were made and concluded that they were not suitable for dose-response analyses in epidemiology (IOM 1995).
The Working Group concluded that there has been a lack of consistency over time in NTPR dose estimation methods and, in particular, in the methods of assigning ''high-sided" doses, that is, doses in which uncertainties are resolved in favor of assigning higher doses rather than lower doses. In some cases, because of the existing compensation program, procedures for assigning doses have been different for those who did and did not file a claim for a radiogenic cancer. Neither the dose assignment methods nor the database itself are thoroughly documented. In addition, uncertainties have not been estimated in a consistent manner and do not incorporate all potential sources of variability inherent in the dosimetry. (p. 2)
The conclusions also state, "Although there is anecdotal evidence that individual doses may have been greatly underestimated in individual cases, the overall tendency may have been to overestimate both external and internal doses."
The concurrence of both veterans with cancer having higher likelihood of individually reconstructed doses and reconstructed doses being more likely to be overestimated than others could introduce serious bias into the epidemiologic
analysis of these data. However, the data may still be a useful indicator on a group basis in contrasting exposure surrogates in CROSSROADS. For their part, veterans have expressed concerns that the assigned doses are significantly lower than justified based upon their first-hand experiences at the test site.
Alternatives
Because the dose assignments are the subject of such hot debate, we looked for other indirect quantitative dose measures obtainable from the NTPR database. We hypothesized that either number of badges issued to an individual or total dose derived from badge data might be more reliable measures of individual exposure than the reconstructed total dose discussed above.
Based on DNA background material (DNA 1984), we hypothesized that participants most likely to be exposed to ionizing radiation would have been issued more badges than those believed less likely to be exposed. We looked, therefore, for relationships between the number of badges issued to an individual and both the total dose assigned to that individual and the dose assigned to that individual using badge data alone. Finding none, we rejected using number of badges as an exposure surrogate.
The badging in CROSSROADS was sparse—few people were badged and those who were did not wear their badges continuously during their exposure, according to participant and DNA accounts. Thus, an individual's cumulative dose from film badges may well give an incomplete picture of total dose in CROSSROADS. We examined the dosimetry data to determine if an individual's assigned dose was proportional to his badged dose, which would allow an assumption that badges were indicative of the total dose accrued by the individual. We found that individuals with very similar badged totals had widely disparate assigned doses due to differing dose reconstructions.
In summary, seeing no evidence that the film badge data provided a useful exposure surrogate, we chose not to use them.
Participant Versus Nonparticipant Cohorts as an Exposure Surrogate
Chapter 6 contains our discussion of the use of participation status as a proxy for exposure.
Boarders Versus Nonboarding Participants Versus Nonparticipants as Exposure Surrogates
One would expect individuals who boarded target ships after shot Able or Baker (or both) to be more highly exposed to radiation than others, since the ships were radiologically contaminated, in some cases very heavily. The boarders group does not correspond to a particular occupation; it consists of a variety of different enlisted ratings and officer ranks.
We define a boarder as someone who has a record of being on a target ship after one (or both) of the detonations or who has a unit designation of "RADSAFE" or "BOARDING TEAM." Ships assigned to CROSSROADS were categorized in DNA historical records as "target" or ''support" ships. The former were literally the targets of the nuclear explosions. While all personnel were removed from target ships before the detonations, many reboarded those ships to monitor experiments, retrieve instruments, or decontaminate the vessel. Radiation safety officers were mostly medical and scientific personnel who carried radiation detection devices (e.g., Geiger counters) and were responsible for providing clearance or evacuation advice regarding other boarding party members. Boarding team personnel were so designated in their military assignment records.
This boarder variable has high face validity and makes what may be the best use of available data. The data, however, have not been validated regarding either accuracy of actual assignments or actual associations with ionizing radiation exposure. Furthermore, not all potential high-exposure groups are captured by the boarder variable. DNA narratives mention diving teams, for example, that may have been exposed but not recorded as such in assignment records (DNA 1984).
While the boarders represent a more exposed group of CROSSROADS participants, on balance the remaining participants constitute a less exposed group and the nonparticipants are an unexposed group. The distribution of individuals in occupation and grade categories and their status as boarders or nonboarders is reasonably balanced.
Engineering & hull Versus Other Specialty Status as an Exposure Surrogate
An individual's military classification—by rank, rate, paygrade, or occupation—provides clues to his age, education, physical exposures, military salary, and postmilitary occupation (and concomitant exposures). While these classifications are relevant to this study as proxies for the confounding effects of
socioeconomic characteristics (which we discussed in Chapter 3), in this section we consider their usefulness in developing proxies for an ionizing radiation exposure measure.
Military paygrades fell into three categories: enlisted (El to E718), warrant officer (W1 to W4), and commissioned officer (O1 to O10). By the alphabetic letter we categorize—however grossly—something about the individual's position in the military hierarchy, while the number reflects, for one thing, time in service. Ranks provide service-specific nominal titles to the military-wide paygrades and provide descriptive information. Army examples of ranks are First Lieutenant (1LT, O2), Chief Warrant Officer (W2), and Staff Sergeant (E6). The Navy assigns ranks (for example, Captain, O6) to all officers, while a rating is assigned to all enlisted individuals. The rating combines information about a sailor's rank and occupational specialty (for example, Chief Electrician's Mate, E7, or Seaman Apprentice, E2). Navy enlisted ratings are unique in that they provide valuable information about typical duties. Similar information was unavailable to us for enlisted persons of other services or for officers of any service. We used the "Navy Career Path" (NAVPERS 1949) to reduce over a thousand variants of Navy ratings to 14 broad occupational specialty groups (Table 8-1).
TABLE 8-1. Broad Occupational Specialty Groupings of Navy Enlisted Ratings Developed from "Navy Career Paths" (1949)
Occupational Group |
Major Job Fields (examples) |
Administrative and Clerical |
Communications Technician, Disbursing Clerk |
Aviation |
Aviation Machinist's Mate, Aviation Photographer's Mate |
Construction |
Surveyor, Steelworker |
Deck |
Boatswain's Mate, Radarman |
Dental |
Dental Technician |
Electronics |
Electronics Technician |
Engineering & Hull |
Boilerman, Machinist's Mate |
Medical |
Hospital Corpsman |
Miscellaneous |
Printer, Musician |
Ordnance |
Gunner's Mate, Mineman |
Seaman |
Seaman |
Precision Equipment |
Instrumentman, Opticalman |
Steward |
Steward, Cook |
Unknown |
No occupational information available |
Each classification by itself has limitations for epidemiologic purposes. We therefore used the following criteria in our attempts to create a classification variable for our analyses:
- balanced cell sizes;
- provision of some information about potential exposure to ionizing radiation;
- provision of some control of unmeasured characteristics of individuals that could be related both to past, current, and future exposures and to health outcomes;
- efficient use of available information; avoidance of large unknown categories;
- a categorization that is both understandable to the reader and meaningful within a mortality analysis.
Prior nonmilitary research has shown that age, socioeconomic status, and occupation are all related to health outcomes. Naval Health Research Center reports associate specific occupations and grades with, for example, respiratory disease, mental disorders, accidents, and hospitalization rates (Gunderson 1976; Helmkamp and Colcord 1984; Helmkamp and Bone 1985a, 1985b). Available data do not include occupation or task-related information about officers. DNA materials (Appendix B) suggest that some job classifications may be associated with higher than normal radiation exposure potential at CROSSROADS. Because a major source of radiation at the CROSSROADS test site was contaminated seawater and because this seawater flowed through pipes in the Engineering & Hull spaces of the ships, one might expect individuals working in that environment to have been more highly exposed to radiation than other personnel. For that reason, DNA assigned higher doses to these individuals, based upon their Engineering & Hull rating. We therefore developed a seven-level variable (Table 8-2) that used information from occupational group and paygrade. We use this grouping in calculating standardized mortality ratios for Navy all-cause and all-malignancy mortality (Appendix C).
TABLE 8-2. General Rank/Rating and Occupational Specialty Categories for Navy Personnel
Engineering & Hull specialties, junior enlisted paygrades E 1–E3 |
Engineering & Hull specialties, midlevel enlisted paygrades E4–E5 |
Engineering & Hull specialties, senior enlisted paygrades E6–E7 |
All other enlisted specialties, E1–E3 |
All other enlisted specialties, E4–E5 |
All other enlisted specialties, E6–E7 |
All officers, W1–W4 and O1–O10 |
There is great variability in occupation and activities within each of the above groups. Engineering & Hull occupations in the Navy are held by those individuals who were most likely to be found working below decks (Appendix
B). Included in this group are Boiler Tenders, Pipe Fitters, Machinists Mates, and similar enlisted ratings.
Other enlisted occupations range across Fire Control Technician, Steward, Journalist, Aviation Ordnanceman, and Accountant. Officers' responsibilities can include physical exposures in whatever area they command. Furthermore, paygrade provides a very imprecise and ''noisy" measure of age and socioeconomic characteristics. For the enlisted men, we have incorporated three grade levels, junior (E1–E3), midlevel (E4–E5), and senior (E6–E7). There were insufficient numbers of officers to break them into grade subgroups while retaining statistical power. The combination of occupation and paygrade may also control for exposure potential determined in part by whether particularly "dirty" or ''interesting" assignments were directed to junior or senior personnel.
Relationships of Surrogates with DNA Dose Data
Validating our surrogate measures is impossible without an accepted comparison standard. While holding to our critique of both measured (badged) and reconstructed doses to individuals, we recognize that those data may be useful for other than dose-response epidemiologic analyses with the individual as the unit of analysis.
We tabulated the surrogate measures of exposure described in this chapter—occupation, paygrade, boarder status, participant status—with the DNA dose estimates assigned to the personnel in those categories, thinking that general concordance would add support to our use of those surrogates. Note that this comparison was done in retrospect. Dose data were neither used nor considered in the conceptualization or development of the exposure surrogate variables. The validating comparisons were made after the decision was made to use the surrogates.
In retrospect, the DNA-assigned doses do confirm our exposure expectations in the surrogate groups we selected. Both the Engineering & Hull and the boarders19 have higher assigned doses than those who are not in these categories. For the Engineering & Hull group, however, this is in some respects a self-fulfilling prophecy, since DNA increased their assigned dose estimates because they thought them to be at higher risk of exposure.
Our expectation of relatively high exposure in the RADSAFE and Able-only groups is reflected in both the badge and total assigned dose data. The total
assigned dose data also validated our assumption of higher exposure in the Engineering & Hull, Baker-only, and both-shot groups. The badge data do not validate the last three groups, but their appropriateness as a validation tool is questionable because of the limitations of film badge dosimetry discussed earlier. Overall, we believe our selections of dose surrogate groups are consonant with the assigned total dose data and are not refuted by the badge data.
Decisions for the Analyses in this Report
Based on our considerations as described in the preceding sections, the committee and staff decided not to use dosimetry data in the analysis. This decision was not taken lightly. The amount of painstaking sifting through military records by DNA to develop the dose data was immense. So, too, is the information gained about a physical exposure that no one, until then, had been required to measure. The dose data, however, as previously described, do contain biases that could affect the study's results in ways that are not well defined. To prevent that, we made the decision not to use the individual-specific reconstructed or badged doses before looking at exposure-outcome correlations.
Our decisions on the establishment of dose surrogates were:
- Use participant vs. nonparticipant comparison.
- Use a consolidated boarder variable as one exposure surrogate.
- Use dosimetry data to provide an overall perspective regarding the amount of ionizing radiation exposure involved (see Appendix D).