Radiological Assessments for the Resettlement of Rongelap in the Republic of the Marshall Islands (1994)

The technical activities (environmental sampling and dose measurement) needed to support any decision on resettlement are themselves diverse and scientifically sophisticated. However, the integration and interpretation of the resulting data in a manner that permits reliable dose projections depend on the development of scientifically sound diet models appropriate to the population of concern.

The development of a model that provides useful predictions of dietary intake under a variety of life circumstances is not easy. Nevertheless, accurate estimates of exposure to radioactivity depend on a robust diet model, particularly in the northern Marshall Islands, where a large component of the radiation dose appears to be contributed by cesium-137, which is concentrated in common local foods. A number of groups have studied food habits in the Marshall Islands, including Kotzebue (1821, 1830), Paulding (1831), Finsch (1893), Snefft (1903), Kramer and Nevermann (1938), Lawrence (1943), Murdock (1943), Mason (1947, 1967), Tobin (1952, 1953), and Maramba (1960). However, the data from those sources have little value for use in dose projections because they are old (and, therefore, not in tune with changes in dietary habits) and generally not quantitative. The most recent information on diet that might be relevant to the resettlement of Rongelap comes from the Marshall Islands National Nutrition Survey of 1991, which focused on the growth of children and infant feeding practices (RMI Ministry of Health, 1991); it also does not provide quantitative estimates of food intakes.

Only two surveys have provided data that are of potential value in quantitatively estimating adult dietary intakes and are reasonably appropriate to the proposed resettlement of Rongelap: a 1978 survey of the Ujelang community by Micronesian Legal Services Corporation (Robison et al., 1980, 1992, 1993) and a 1970s survey of the dietary habits of the northern Marshallese by Naidu et al. (1980). A thorough assessment of the strengths and weaknesses of both surveys is needed to interpret any calculations of radiological exposure based on their data; such an assessment is provided below.

The Ujelang diet model is based on a survey of the Ujelang community in 1978 by the Micronesian Legal Services Corporation (MLSC) staff with the aid of a Marshall Islands school teacher on Ujelang (Robison et al., 1992, 1980). Although the survey had weaknesses, it was in many ways superior to national diet surveys in the United States (Human Nutrition Information Service, 1988; Public Health Service, 1979; Science and Education Administration, 1980). In particular, these surveys suffer from poor participation rates, non-optimal statistical selection procedures, lack of timeliness, or inherent limitation of the methods (Guenther and Tippett, 1993). The diet surveys of the Rongelap and related populations provide critical data

for the evaluation of projected doses because of the importance of the ingestion pathway for exposure to radionuclides. In view of the need for precise dose calculations to examine the conditions set forth in the MOU, the requirements for an acceptable diet survey far exceed those for usual surveys.

The positive aspects of the Ujelang diet survey include the following:

• The sampling plan was reasonable. It was based on an enumeration of all members of the population. Subjects were grouped by age and sex, and an attempt was made to interview at least 20% of the subjects in each category.

• The purpose of the study was carefully explained to all subjects. An official (a magistrate) emphasized the importance of giving an accurate recall that neither ignored some foods because they were ''unimportant'' nor overestimated intake of others to please the interviewer.

• Trained interviewers conducted all the interviews, so data collection was standardized and fairly uniform.

• The survey instrument was detailed and so should have helped subjects to remember details about their food intake. An attempt was made to help subjects in estimating serving sizes by making comparisons to a 12-oz. can. Subjects were asked to estimate intakes for days, weeks, and months.

• The survey team identified the situation in which local foods were most apt to be consumed (i.e., famine) and asked detailed questions on both that situation and the typical situation.

The MLSC diet survey in Ujelang also had shortcomings, as presented below, and its deficiencies must be recognized when using the data generated, especially because extrapolations from diet have such a critical effect on estimated doses of ingested radionuclides. Each deficiency must be considered in relation to different contexts of Marshallese life, but the first four presented potentially have the most important impacts from the point of view of data application.

• Although the magistrate emphasized the importance of giving an accurate recall of foods eaten, the climate of the survey was politically charged. Pritchard, who conducted the survey for the MLSC, told the community that "an earlier survey had reported what seemed to be an unrealistically high consumption rate for local foods, particularly coconuts, and if these reports were utilized in assessing the risk, there was almost no likelihood that people would be able to live on Enjebi in the foreseeable future" (Pritchard, 1979:2-3). Because the exiled Enewetok people would, at that point, have done virtually anything to ensure they would be returned to Enewetok Atoll (Carucci, 1992), presenting the request for dietary information in this way almost certainly affected how people responded to the survey, particularly residents

whose lands were mainly in the northern (Enjebi) half of Enewetok Atoll.

• Lack of information on subjects' living situations, especially during famine periods, limits the interpretation of data from this survey. All Ujelang residents over 1 year of age are known to have consumed large quantities of arrowroot, out of necessity, during a famine period in 1978 (the year before the survey). But Ujelang people are extremely mobile. Some of those reporting no arrowroot consumption might not have lived on Ujelang at the time of the famine. The researchers did not keep track of residents' histories, so the causes of discrepancies are not immediately apparent. As noted in the study by Naidu et al. (1980), type A communities (outer islands, where residents depend on imports, not local food) have substantially different patterns of food use from other communities. During the 1978 food shortage, over a dozen Ujelang residents tended tapped coconut trees. General coconut consumption, particularly of coconut liquid and sprouted coconuts, increases when men make copra. During famines, coconut is included in dishes not described in the survey (pikukuk, mokkir, and aojek). That these are Ujelang foods with distinctive names, perhaps not known to the Marshallese translator, contributed to their not being reported in the survey.

• The reported low energy intake (1392 kcal/d) of the Marshallese when eating an "indigenous" diet is surprising and probably unrealistic (Robison et al., 1992). This estimate is less than the average recommended energy intake for 25-to 50-y-old men but not women, 1,800 and 1380 kcal/d, respectively (Food and Nutrition Board, 1989). These recommended intakes constitute a general guideline, but their applicability to Ujelang residents is not known. The food intake of subjects consuming the indigenous diet of 1,392 kcal/d was reported to be only 40% of their typical intake, 3490 kcal/d (Robison et al., 1992). That inverts the probable energy requirements of adults. Island residents would expend more energy in collecting and preparing indigenous foods than in preparing imported, often pre-packaged imported items. A deficit in energy balance of the type indicated by the data would have caused the men to lose about 4 lb of weight per week. There are no reports of such rapid weight loss during the famine period.

• Another major limitation of this survey was the failure to distinguish food events systematically. The researchers did ask about times of famine, as well as everyday consumption when imported foods were available, but they failed to differentiate between food intake at feasts and on weekends, as opposed to weekday consumption. Feasts, particularly first-birthday celebrations, occur frequently, and food intake increases at these times. Food intake also increases substantially during the 4-month period during which local people celebrate Christmas (October-January), Enewetok liberation day, and Easter. As throughout Micronesia and Polynesia, food intake is generally greater on weekends than on weekdays (Pelletier, 1987). Food intake also varies with occupation. The intake of particular foods increase greatly during copra production, and on extended fishing and gathering expeditions. In this survey, subjects might have reported typical intakes on weekdays, weekends, feast days, or hypothetical "average" days. It is likely that Ujelang residents varied in their interpretation of a "typical" day. Subjects might or might not have included in their estimates special labor conditions, weekdays and weekends, feast days, and the lengthy celebration of Christmas.

• The age stratifications by 5-y increments are compromised by the fact that "people on Ujelang [are] unsure of their own and their children's ages and all these figures should probably be read ± 1 or 2 [years]" (Pritchard, 1979a). Parental estimates of "ages" are again inexact and probably should be ± 5 [years]'' (Pritchard, 1979a). Thus, the balance of subjects in each category might be skewed from the 20% representation claimed in the report.

• Interviewers were trained to perform dietary interviews, but they were not experienced nutritionists or dietitians, nor were they familiar with Ujelang foods. Even the local "Marshallese" school teacher was an "in-married" man from Pingelap Atoll who spoke an acceptable variant of the local dialect. In addition, the survey instrument was administered differently over time. Some foods were added as the interviewers became aware of them (Pritchard, 1979a:15); others were measured more accurately as the survey progressed (Pritchard, 1979a:4). During the first 2 d, questions were translated, often with extensive explanation, and the answers were translated back for transcription to the survey form. After the first two days, the Marshallese assistant, who was unfamiliar with the Ujelang dialect of Marshallese, simply stated the question in Marshallese and translated only answers or requests for further explanation to the interviewer (Pritchard 1979a: 2). Thus, many linguistic and nonverbal responses went untranslated after day two.

• The survey instrument was based on a general knowledge of Marshall Islands foods, but it failed to exhibit a specific knowledge of the food items on Ujelang Atoll. Neither the names of the foods nor their modes of preparation were accounted for by the survey team. Researchers "added the term mado to our survey the second day, [but] there may be some underreporting since the word used on Ujelang is Kiraj" (Pritchard, 1979a: 15). At times, the survey team's lack of knowledge of local methods of preparing indigenous foods caused them to overlook numerous variables that directly affect food consumption and composition. For example, pork is most often prepared by boiling, but the interviewers stated that pork was baked or reheated in a frying pan (Pritchard, 1979a:16). Similar problems arose in relation to coconuts, which at that time were considered the most critical local food in relation to radioisotope uptake (Pritchard, 1979b). In spite of a commendable attempt to corroborate the general levels of coconut consumption (Pritchard, 1979b), confusion persists in relation to the stages of coconut, the physical characteristics of coconut at different stages, and how each coconut stage is related to patterns of food preparation and consumption (Pritchard, 1979b:13, 17). Some of the food designations, such as "Marshallese cake," and questions used in the survey were probably confusing to the interviewees. Marshallese cake has multiple forms. Although one type includes substantial quantities of coconut, another is made without shredded coconut. Mokkir, a Ujelang creation, consists of cake broken into bite-sized bits and mixed with warm coconut cream. This common feast food was not mentioned among the list of local consumables. To some extent, the interviewers tried to compensate for these survey limitations by asking people to estimate the amounts of coconut cream added to food in general for a set period (Pritchard, 1979b); people who did not prepare the food would have difficulty in answering this question.

• Although a familiar vessel (a 12-oz cola can) was displayed to help people to estimate

quantities of ingested items, no attempt was made to demonstrate how much food or drink could be contained in it.

• The survey population was small. Unintentional collusion due to the interview situations might have skewed some results. For example, in one situation, both subject A and her daughter, subject B, reported exactly the same intakes of foods (W.L.Robison, personal communication, 1993). Their responses are suspect (although not impossible) because in 1978 subject A was a large, mature woman and subject B was slight. Extraneous values might also have skewed the results of this small survey. For example, subject C, the younger sister of subject B, in 1978 shared the same cookhouse. She estimated that she consumed 180 g of baked bread per day; that is 2 times the next highest woman's estimate in the community and over 8 times the quantity being consumed by subject B. Another example of questionable data concerns subject X, who claimed that she consumed 63 g of arrowroot per day as part of her normal diet, not in famine conditions. That is 3 times the next highest woman's estimate in a field of 38 in which 32 women claimed that they consumed no arrowroot. A translation problem probably caused subject X to make her estimate. Whatever the cause, the estimate skewed the data on arrowroot consumption.

• As is typical of diet surveys, this one made great demands on subjects' memories and abilities to quantify food intakes. It is doubtful that subjects were able to remember all the foods that they consumed or to estimate serving sizes accurately. In recent surveys in the United States and Northern Ireland, average subjects reported energy intakes 18-19% below maintenance intakes (Mertz, 1992; Lichtman et al., 1992; Livingstone et al., 1990). However, the extent of underestimation varies among subjects. Lichtman et al. (1992) reported that a group of obese subjects underreported their energy intake by 47±16%. Robison et al. (1992 and 1980) noted that the estimated food intake was about 20% greater for women than men. Other investigators have noted that more accurate dietary information was obtained from women than men (Rider et al., 1984; Yuhas et al., 1989). That might reflect greater attention to food by women. In the Marshall Islands, food preparation depends on the occasion, but maturing girls and young women do most of the regular cooking. Generally, people remember foods consumed as major components of meals more accurately than foods and beverages consumed as "accessory" foods (Greger and Etnyre, 1978; Mullen et al., 1984). Food events are highly variable in the Marshall Islands, and a great deal is consumed ad hoc as people move around the village accepting invitations to eat at relatives' houses or as they work in the bush. Moreover, it is rude to eat elsewhere and not bring food home to share. Therefore, when interviews were conducted in household settings, the intake of "snack" foods and beverages might well have been underestimated. It is particularly difficult for subjects to report past behaviors accurately (Bradburn et al., 1987; Dwyer et al., 1987); the most recent famine experiences were slightly more than a year before the time of the survey, and the lower food intakes and increased variability in the "famine" statistics might be explained in part by the time lapse.

• There might be at least three other sources of error in extrapolation from the 1978 survey of the Ujelang diet to the 1990s Rongelap diet. First, the diet observations were made

when the Ujelang council members were absent from the island. If their diet differed from the noncouncil population and if the council members should be considered part of the target population, then the sampling was biased. Second, extrapolating from Ujelang to Rongelap assumes similar ecological situations and dietary habits. Third, the use of a 1978 diet survey to estimate food consumption in the 1990s assumes that dietary practices have not changed substantially. The presence and practical consequences of these potential sources of error are unknown.

The Ujelang diet model represents the best data on which to base estimates of a Rongelap diet. Although it has several limitations, it is as good as many major diet studies in the United States (Human Nutrition Information Services, 1988; Public Health Service, 1979; Science and Education Administration, 1980). Even with its drawbacks, it was the logical and appropriate choice for Robison et al. (1992) to use to estimate exposure to isotopes in ingested food and beverages. However, the provisional character of this diet model necessitates a recognition that absolute answers on the average intake and on measures of variability are unlikely. Thus, the committee feels that an additional way to use the Ujelang diet model is in preparation of scenarios that allow for variations in energy intake, food selection (e.g., imports vs. local foods and such activities as coconut-gathering), and effects of potential remediation activities (such as application of KCl fertilizer to soil). Table 3-1 gives several potential scenarios; more could be devised to reflect the concerns and life style of the members of the Rongelap community.

This diet model that has been applied to people of the northern Marshall Islands, was developed by Naidu et al. (1980) of BNL based on intermittent observations by a survey team and qualitative data collected in a formal survey. The survey team lived in local communities for brief periods and reported the food-preparation techniques and food patterns of the islanders. Systematic records of food intake were not made by the survey team. Thus, their report—with those of Murai et al. (1958), Pollock (1970), Carucci (1980), and Maifeld (1982) and the historical accounts mentioned earlier—contribute to a comprehensive set of qualitative descriptions of the dietary habits of the residents of the Marshall Islands. However, the qualitative data collected by the BNL team are not useful for making quantitative estimates of intake of specific nutrients or even foods.

The following points are important for the interpretation and use of the BNL data:

• Like the Ujelang survey data, the BNL study is dated (data were collected before 1980). Food habits have changed dramatically throughout the Pacific as residents have become more integrated into the world economy (Bindon and Baker, 1985; Pelletier, 1987; McGarvey, 1991; Taylor et al., 1991). Local people are probably substantially less dependent on indigenous foods, particularly foods that require extensive preparation, than they were at the time of this survey.

• The BNL study shows no evidence of a sampling plan. The number of subjects interviewed and the demographic situation of each subject were not defined in the major report of data.

  The survey team did not seem to appreciate the principles for estimating average diets. They suggested that it would be impossible for an outside observer to choose a "typical family" to monitor for dietary observation, although the "typical maximum" or "typical minimum" diet would stand out (Naidu et al., 1980). Once a sample population was defined, normative values could be easily derived. Without a demographic profile, it is impossible to define the "meaning'' of the BNL data or to subdivide the data and derive separate profiles with sensitivity to sex, age, residence, day-to-day occupation, and seasonal or weekly variations.

• The survey instrument was poorly designed to elicit quantitative information. For instance, Question 28 asked "How many pumpkins do you cook for your family during a typical year?" and Question 37 asked "How many times do you make a meal of pig during the typical month or year?" (Naidu et al., 1980). Such questions did not elicit quantitative information, e.g., on the sizes of pumpkins or pigs or on the sizes of typical servings for males and females or adults and children. The questions also made unrealistic demands on the subjects' memories; estimating consumption of an item for a typical month or year is very difficult (Dwyer et al., 1987), especially when variations of residence, activity, and season are likely. The interviewers had no systematic mechanism to allow respondents to standardize quantities. Moreover, adults made the estimates of consumption for children—a likely source of reduced variability in sampling.

• The amount of food prepared was measured in the BNL study, not food intake. It is difficult to estimate the percentage of prepared food actually consumed by humans. Refrigeration at the time of the survey was available only in urban areas. The leftovers were probably fed to pigs, chickens, and dogs. No estimates of wastes have been made, so conversion of prepared food to consumed food is very questionable. Cycles of feast and famine have immediate effects on waste and food conservation. The difference in sampling e.g., assessment of prepared food versus consumed food—accounts for a lower estimate of coconut intake and hence of ingested isotopes in the Ujelang diet model than in the BNL diet model. It is also logical that the Ujelang diet data (based on food consumption) correlate more closely with measured isotope body burden than BNL data (based on food preparation with no correction for waste).

• The most useful information from the BNL survey is derived from the discussion of seasonal variability of food intake. Unfortunately, the survey provides no information on differences in food intake based on age, sex, residence history, occupational pursuit, weekday versus weekend, ordinary day versus feast, etc.

• The delineation of diet patterns on the basis of three hypothetical community types provided by the Naidu et al. study is potentially useful, but the basis of the classification and the actual communities that were surveyed were never made apparent (Naidu et al., 1980). The

creation of hypothetical Communities A, B, and C by the authors in an attempt to be objective ultimately compromised the scientific utility of the research, in that the maintenance of community and respondent anonymity means that the data could not be reanalyzed. Although the contrast between Community A and Community C represents the logical extremes in the Marshall Islands, Community B is anomalous. It was not described adequately to make the data useful. Generally, a dense population is subject to heavy fishing pressure and reduced fishing success. Nonetheless, Community B is defined as "overpopulated" with low availability of local food and yet with "excellent fishing"; about 33% of its diet comes from fish. It is extremely unlikely that residents of the ''overpopulated" areas of the Marshall Islands (Uliga, Majuro; and Ebeye, Kwajalein) have access to enough fish to make up 33% of their diet. To make the data usable, terms like "overpopulation" need to be quantified, the communities need to be identified, and for purposes of Rongelap resettlement the living circumstances in the research communities need to be compared with the expected situation on Rongelap Atoll.

• The BNL survey data have other serious anomalies that are unexplainable without further information. For example, a reasonable hypothesis (on the basis of information from other diet surveys) would be that consumption of coconut in general and coconut cream in particular would decline as a community became more urban. Indeed, in Community A (with a low population density), about 10,000 g of grated coconut for cream (milk) per person was produced per person each year; and in Community B (high density but fish plentiful), grated coconut production used for cream dropped to 2,500 g/y—a reasonable expectation. In Community C, however, with high population density and low availability of local produce, grated coconut for cream production was estimated to be 45,000 g/y. This is highly unlikely: coconuts are virtually unavailable in either Ebeye or Uliga, and when available they are expensive. Such anomalies decrease confidence in the BNL diet survey.

The dietary data generated by the BNL diet survey are not quantitative and therefore are of only supplementary value in estimating dietary intake of the Marshallese.

The lack of a comprehensive DOE research plan to relate the BNL and LLNL research endeavors formally has produced mixed results. The Ujelang diet model and the LLNL environmental information can be combined to yield whole-body burdens of cesium-137 that are generally consistent with the BNL measured values. However, some features of the diet model are clearly unrealistic. There are many dietary components in the model, and single confirmation solely by BNL measured values cannot be taken as confirmation that the details of the dietary model are correct. In particular, there is little evidence that the relative importance of some local foods in the diet is being accurately assessed.

Diet information now being obtained might help to reduce the uncertainties, but historically BNL' s reliance on LLNL diet information from the Ujelang diet survey to calculate potential plutonium intake through food has been a risky strategy (Sun et al. 1992b, 1993). Potentials for soil contamination during outdoor food preparation were recognized as early as 1985 (Lessard et al., 1985, as quoted in Sun et al., 1992a), but LLNL measurements were

based on collected foods, rather than prepared foods. Indeed, radioactive assessments of foods prepared under local conditions are lacking. Although improvements in the urine-collection protocol appear to have answered questions about increased plutonium in BNL' s urine-bioassay measurements, absolute assurances about plutonium intake from food (Conard 1992:53) cannot be answered, given our current lack of knowledge about Marshallese food-preparation techniques.

The unanswered dietary questions include these: How accurately does the Ujelang diet survey predict the pattern of food consumption on Rongelap? How have northern Marshall Islands consumption patterns systematically changed since 1978? How have outer-island food-preparation techniques altered the ingestion of cesium and plutonium? Robison et al. (1992) recognize the existence of problems in knowledge of the Ujelang diet. The Committee hopes that a few of these problems will be solved through a more systematic assessment of Marshall Islands diets.

We recommend that a comprehensive research plan be developed to integrate the BNL and LLNL research endeavors. Historically, the dietary and dosimetric data have not reached their optimal use, because well-defined research objectives and program review have been lacking. A comprehensive research plan could have brought the individual laboratory efforts together toward a common goal with greater effect than that achieved separately. The laboratories should make ample use of experts in nutrition, diet, dosimetry and in the cultural characteristics of the local people in developing their comprehensive research plan.

The following suggestions might help in the interpretation of the total data set but will not alter the Committee's basic conclusions.

• Individual dietary differences are not well described in the current diet models. Individuals differ in both total amount of food consumed and choice of items in the diet. The cesium intake of each person in the Ujelang survey should be calculated individually from food intake and reported in terms of grams per day and picocuries per day. This is under way at LLNL. Separate values (means plus or minus standard derivation) for food and isotope intakes for each age and sex groups should also be calculated.

• Measurement of food prepared and food left at the end of meals before animals are fed would yield information on waste. If such measurements were made in 10 households under a variety of circumstances—e.g., feast versus everyday and weekend versus weekday—it would be possible to estimate waste and apply this correction factor to the BNL diet model. These estimates could not be statistically validated, but they would provide a way to establish closer correlations between the BNL diet model and the Ujelang diet model. Considering the other limitations of the BNL diet model, however, such corrections may not be considered worth the effort.

• It is not realistic to expect the Marshallese to return to a totally indigenous diet. Active adults need more than 1,392 kcal/d (1,541 g/d)—the estimated average energy intake of adults subsisting on a totally indigenous diet on Ujelang—but are unlikely to consume 6,000 g/d, the approximate amount consumed by people with the highest consumption in the Ujelang survey (Robison et al., 1993:37). Therefore, estimating dietary intakes for a 100% indigenous food diet is probably not useful. The best way to estimate maximum intake of indigenous food is debatable, but it might be possible to survey food intake on Ujae or Lae Atoll. Residents of these outer atolls in the central western Marshall Islands use more indigenous foods than those of Rongelap Atoll because their financial resources are smaller.

• The study by Calf (J. Calf, personal communication, 1993) will update information on dietary intake of the Marshallese from before 1980 to 1993. It should provide important information on infant feeding practices and might help to identify food intakes of children, young single males and females, nursing women, adult males, nonnursing females, and the elderly. Such data will yield estimates of variance in food intake for each group. More important, this survey was designed to provide realistic estimates of energy intakes for subjects of different ages and should accomplish that goal. However, the survey conducted by Calf consisting of 1-d diet recalls of all residents on Mejatto during May 1993 will not provide data on seasonal variation. A study being conducted by Burton and Nero of the University of California, Irvine (Burton, personal communication, 1993), although not quantitative, will be a useful source of supplementary data on shifts in food intake due to seasonal variations, feasts, and special activities, such as intensive food-gathering or copra production. This information will be helpful in formulating potential diet scenarios involving indigenous foods and people with varied activities.

• An accounting of all shipping records for imported foods arriving in Mejatto would provide a cross-check of data collected in the Ujelang and Calf diet surveys and provide another means of assessing the relative uses of imported foods. This accounting should include all government subsidies (U.S. Department of Agriculture: school lunch and supplemental food supplies that are part of nuclear-related compensations) and all purchased foods. Some sort of food-supplement program or agricultural program will be essential if energy intakes are to be maintained after resettlement of Rongelap.

• Data on the radiation content of selected prepared indigenous foods would aid in the interpretation of current data on the radiation content of raw food products. The result would reassure northern Marshall Islands residents about one of the possible ingestion pathways of plutonium and strengthen the case for multiple uses of the Ujelang diet model.

If residents of Rongelap agree on a reasonable goal for indigenous-food intake, the Ujelang diet model will be sufficient for scenario-based calculations of dose.