Recall of residential history and dietary habits during pregnancy and lactation in the distant past: reliability of questionnaire-based radiation doses for persons exposed in utero and early life

Abstract

This study evaluates the reliability of information obtained by standardized questionnaires used in by personal interviews for estimation of radiation thyroid doses of 1065 individuals in the Belarusian cohort of individuals who were exposed in utero and early life following the Chernobyl accident in April 1986. Data from two interviews conducted in 2012–2017 and in 2018–2022 with mothers, who were pregnant or gave birth shortly after the Chernobyl accident, were analysed. The most reliable answers dealt with various attributes related to residential history. In contrast, the reliability of answers regarding consumption rates of milk from privately owned cows or trade network was moderate, while the agreement in responses for consumption of milk products and leafy vegetables was fair. Information from the two interviews was used to calculate thyroid doses received by the cohort members. Specifically, ‘model-based’ thyroid doses due to ¹³¹I were estimated using input data on individual residential history and food consumption reported during the personal interviews and ecological data (¹³¹I ground deposition in the corresponding settlements). In addition, for a subset of cohort subjects (n = 205) whose mothers were measured for ¹³¹I thyroid activity, ‘measurement-based’ thyroid doses were calculated by adjusting the model-based dose using a scaling factor that is defined as the ratio of measured ¹³¹I thyroid activity to model-based ¹³¹I thyroid activity calculated for the date of measurement. A moderate agreement was observed for total (prenatal and postnatal) model-based thyroid doses due to ¹³¹I intake, the arithmetic mean ± standard deviation for the Jaccard similarity coefficient (J_sim) was 0.45 ± 0.34 (median = 0.39), while measurement-based doses showed a much better agreement with a J_sim of 0.78 ± 0.29 (median = 0.93). For model-based thyroid doses from external irradiation and from ingestion of ¹³⁴Cs and ¹³⁷Cs, J_sim was 0.82 ± 0.23 (median = 0.90) and 0.84 ± 0.24 (median = 0.96), respectively. Measurement-based doses due to ingestion of radiocaesium isotopes resulted in an almost perfect agreement, J_sim was 0.91 ± 0.19 (median = 1.0). The present findings suggest that long-term memory recall can be reliable, if a person is asked about unique or important life events, such as pregnancy and childbirth occurring around the time of a nuclear reactor accident. However, the substantial difference (more than 10 times) observed for model-bases doses calculated using the two questionnaires represents an important source of human factor uncertainties that needs to be considered in any dose response analyses. Other lessons learned from this study are that (i) individual measurements of radionuclides in the human body are the most valuable source of information for estimating radiation doses, and (ii) whenever a radiation accident occurs, a sample of affected people should be asked to keep a diary, if at all possible.

Introduction

In many epidemiological studies of environmental exposure to ionising radiation personal interviews are used to collect retrospective information needed to assess the nature and magnitude of exposure. Information collected should accurately reflect the events that occurred at and around the time of exposure. However, the reliability of such information, based on memory recall, is often unknown or at least difficult to assess when interviews are done years and decades after the time of possible exposure. Consequently, memory recall must be considered as an important source of uncertainty, that can affect the reliability of dose estimates (Drozdovitch et al. 2022a).

In 2017 the U.S. National Cancer Institute established—in collaboration with the Republican Research Center for Radiation Medicine and Human Ecology (RRCRM&HE) in Gomel, Belarus—a cohort of 2965 Belarusian individuals exposed in utero and during early years of life to radioactive fallout from the Chernobyl accident in April 1986 (Yauseyenka et al. 2020). Individual thyroid radiation doses were estimated for pre- and postnatal exposure using residential history and dietary consumption data obtained by personal interviews with mothers of cohort subjects and, when available, supplemented by information on measurements of ¹³¹I thyroid activity in mothers (Drozdovitch et al. 2020). It was previously indicated that if dose-related measurements are not available and only modeling is used for dose estimation, high-quality data on individual behavior and diet are essential in attaining realistic and reliable dose estimates (Drozdovitch et al. 2016; Kukhta et al. 2021). For the Belarusian cohort of the present study, dose-related measurements (measured ¹³¹I thyroid activity) were available only for a small fraction (about 10%) of mothers of the study subjects. Therefore, it was considered important to assess the reliability of thyroid doses based on information obtained by interviews of the mothers.

Consequently, a study was conducted during 2018–2022 to re-interview a sample (36%) of mothers of the Belarusian cohort of persons exposed in utero and early life using the same study questionnaire that was used for the interviews conducted at the time of cohort construction during 2012–2017. Here, results on the reproducibility of interview data on residential history, dietary habits, and other behavior factors and of radiation doses obtained are reported.

Materials and methods

Study population

The studied Belarusian cohort of individuals exposed in utero consists of 2965 child–mother pairs, including 26 pairs of twins born to 2939 mothers. The children (referred here as ‘subjects’) were born between 26 April 1986 and 31 March 1987, so their mothers were pregnant at some point between 26 April 1986 and 30 June 1986, a 2-month period of increased exposure to ¹³¹I from Chernobyl fallout. Most critical for individual thyroid dose estimates was personal information on residential history and food consumption patterns after the accident. Such information was collected for all study subjects during the first personal interview with their mothers between 19 December 2012 and 22 July 2017.

Several sub-groups of mothers were selected for the second interview conducted for the present study, including (i) mothers with data from measurements of ¹³¹I thyroid activity; (ii) mothers who breastfed their children during the period of exposure to ¹³¹I; (iii) mothers whose children received highest prenatal thyroid doses (500 + mGy) from ¹³¹I calculated using data from the first interview; (iv) mothers who received highest thyroid doses (500 + mGy) from ¹³¹I calculated using data from the first interview; and (v) mothers whose children received prenatal thyroid doses from ¹³¹I between 100 and 500 mGy (randomly selected every fourth mother) and between 20 and 100 mGy (randomly selected every fifth mother). A total of 1,200 mothers were invited for the second interview. Among them, 64 women could not be included in the study, because they had moved out of the study area, had serious physical disability, deceased (n = 49, 4.1%) or refused to participate in the study (n = 15, 1.2%). Some women (n = 77, 6.4%) were not interviewed because of logistical challenges associated with traveling to remote areas of their residence and pandemic restrictions due to the COVID-19 outbreak. Finally, 1059 of 1200 (88.3%) mothers, were interviewed (six with twins) between 14 January 2018 and 24 February 2022.

At the time of the accident (ATA), most of the mothers (n = 1022, 96.5%) lived in Gomel Oblast, the most contaminated area, while 27 mothers (2.5%) lived in Mogilev Oblast; 4 (0.4%) in Minsk Oblast or City, and 6 (0.6%) in the Russian Federation or Ukraine (Table 1). Among them, 635 (60.0%) resided in rural settlements and 205 (19.4%) had their ¹³¹I thyroid activity measured between 26 April and 7 June 1986. Among the 1065 children (study subjects), 428 (40.0%) were breastfed during the period of exposure to ¹³¹I, and 166 (15.6%) had whole-body counter measurements (WBC) of radiocaesium body burden in 1987–1991.

Table 1.

Distribution of mothers or study subjects in the Belarusian in utero cohort according to different subgroups

Subgroup	Characteristics	Na	%
Oblast of residence ATAb	Gomel	1022	96.5
	Mogilev	27	2.5
	Minsk	4	0.4
	Other	6	0.6
Type of settlement of residenceb,c	Rural	635	60.0
	Urban	424	40.0
Direct thyroid measurement of mother	Yes	205	19.4
	No	854	80.6
Total mothers		1059	100.0
Child was breastfed on 26 April–30 June 1986b,d	Yes	428	40.0
	No	637	60.0
Child was measured by whole-body counter in 1987–1991d	Yes	166	15.6
	No	899	84.4
Total children (subjects)		1065	100.0

ATA at the time of the accident

^aNumber of mother unless otherwise indicated

^bAccording to information provided during the first interview

^cType of place of residence at the time of the accident

^dNumber of children

Figure 1 shows the study timeline. The time span between the recalled period of interest, i.e., the period of exposure from 26 April to 30 June 1986, and the time of recollection was 28.7 ± 1.1 years for the first interview and 33.7 ± 1.4 years for the second interview. The median and inter-quartile time spans between interviews were 4.9 years and 4.2‒6.0 years, respectively. 228 of the 1065 questionnaire pairs (21.4%) were performed within 4 years, and 259 (24.3%) interviews were performed within a time interval of more than 6.0 years. The mean age of mothers ATA was 25.2 ± 4.6 years, while their ages were 53.7 ± 4.7 years and 58.7 ± 4.8 years at the time of the first and second interviews, respectively.

Fig. 1.

Study timeline. Given uncertainties represent standard deviations

The reproducibility of data was evaluated from the repeat interviews for a number of specific questions related to residential history and cow’s milk consumption; such questions were considered most important in dose assessment. Specifically, measures of reliability were compared according to whether mothers (i) were measured for ¹³¹I thyroid activity, (ii) breastfed, (iii) were pregnant ATA, (iv) resided in urban or rural areas, and (v) whether time span between interviews was < 4.0 years or 6.0 + years (i.e., lowest and highest quartiles).

Study questionnaire and personal interview

As with the first interviews, the second interviews were conducted in person at the Republican Research Center for Radiation Medicine and Human Ecology (RRCRM&HE, Gomel, Belarus) or by mobile teams of the RRCRM&HE at the places of residence of the mothers. The same questionnaire as that in the first interviews was used. It included questions that dealt with the following aspects, all necessary for estimating prenatal and postnatal thyroid doses among persons exposed in utero and early life:

• Mother’s and subject’s residential history, i.e., place of residence and construction material of house ATA, and, if relocated, places and dates of residence between 27 April and 30 June 1986 (period of exposure to ¹³¹I), and those until the child reached the age of 5 years, i.e., between 1 July 1986 and 31 March 1992.

• Mother’s consumption rates and dates of consumption of milk from privately owned cows, of milk from a commercial trade network, of milk products (milk soup or porridge, cottage cheese, sour milk, kefir, cream, sour cream) made from the milk from privately owned cows or purchased in a commercial trade network, and of leafy vegetables, all for the period between 26 April and 30 June 1986.

• Dates and duration of stable iodine administration by the mother between 26 April and 31 May 1986.

• Date of beginning and duration of breastfeeding.

• Mother’s consumption rates and dates of consumption of milk from privately owned cows, of cow’s milk from a commercial trade network, and of milk products after 30 June 1986 (during pregnancy and/or during breastfeeding).

• Consumption rates of child of privately owned cow’s milk, of cow’s milk from a commercial trade network, and of milk products at age 0 – 1, 1 – 2, and 2 – 5 years.

The questionnaires included “basic” and “follow-up” questions. A positive response to a basic question, such as “Did you drink milk?” triggered follow-up questions, e.g., “When did you drink milk?”, “What kind of milk?”, “How much?”, “How often?” etc., whereas follow-up questions were not needed when a negative answer such as “No” or “I do not remember” was given. If the mother did not recall the exact date of, for example, relocation from one settlement to another, she was prompted to estimate the general period during which the event occurred, such as “end of April [1986]”, “beginning of May”, “middle of May”, “end of May” or “June”.

The reliability and validity of information collected during the interview is likely to be influenced by the training and experience of the interviewer (Lee-Han et al. 1989). The interviewers who participated in this study had an M.S. (Master of Science), B.S. (Bachelor of Science), or A.B.S. (Associate Bachelor of Science) degree and received 24 h of training. Specifically, they were trained to behave during the personal interview in a neutral manner and not to show any type of emotion either when asking the questions or when listening to the answers. The interviewers were also trained to use probing techniques, to stimulate the memory recall of the respondent using special probing cards, such as a calendar for April–May 1986 with indication of major holidays and pictures of cups, glasses, and bottles to recall the consumption pattern of cow’s milk. The interviewer consistently read questions from a paper questionnaire and recorded the responses.

The database of information collected during the personal interview was designed for storage, further analysis, and used to calculate radiation doses. Each response was entered independently by two operators. Database verification included comparison of the responses entered by the operators in two databases. When discrepancies occurred, the correct answer was checked in the paper questionnaire and the error was corrected in the database.

Calculation of thyroid doses

As before (Drozdovitch et al. 2020), individual thyroid doses were re-calculated using the same methods for the following exposure pathways:

• Prenatally due to ¹³¹I intake by mothers from inhaled air and/or consumption of foodstuffs, such as cow’s and/or goat milk, milk products and leafy vegetables.

• Postnatally from ¹³¹I intake of infants from breast milk using a transfer coefficient of 0.37 d L⁻¹ from mother’s intake to breast milk (Simon et al. 2002) or/and other foodstuffs if individual was born between 26 April and 30 June 1986.

• Prenatally and postnatally (until age of 5 years) due to external irradiation from gamma-emitting radionuclides deposited on the ground.

• Prenatally from mother’s consumption of foodstuffs contaminated with ¹³⁴Cs and ¹³⁷Cs.

• Postnatally (until age of 5 years) from infant’s consumption of foodstuffs contaminated with ¹³⁴Cs and ¹³⁷Cs.

The thyroid doses due to ¹³¹I were estimated using input data specific to the mother of an infant (individual residential history, consumption data and stable iodine administration reported during the personal interview, and direct thyroid measurement of ¹³¹I activity, when available) and ecological data (¹³¹I ground deposition in the settlements). Ecological and biokinetic models were used to reconstruct the transport of ¹³¹I from the ground to the mother’s thyroid via radionuclides in air and foodstuffs (Drozdovitch et al. 2020). Estimates of ‘model-based’ thyroid dose for the mother served as input to estimate model-based dose to the fetus’ thyroid gland using the model from Publication 88 of the International Commission on Radiological Protection (ICRP) (ICRP 2001). For the subjects included in the present study whose mothers were measured for ¹³¹I thyroid activity, the ‘measurement-based’ thyroid dose was calculated by adjusting the model-based dose using the so-called scaling factor that is the ratio of the measured ¹³¹I thyroid activity to the model-based ¹³¹I thyroid activity calculated for the date of measurement.

The approach used to estimate doses from external irradiation was based on the integration of the time-dependent dose rate in air per unit deposition of the radionuclide mix at the place(s) of residence, considering the shielding properties of the residential environment and the time spent outdoors (Drozdovitch et al. 2022b; Minenko et al. 2006). To estimate doses to the fetus, this model was adapted considering the data provided in ICPR Publication 116 (ICRP 2010).

Exposure to ingested radiocaesium isotopes (¹³⁴Cs and ¹³⁷Cs) was assessed using a semi-empirical approach based on the relation between environmental contamination (¹³⁷Cs deposition density and aggregated ¹³⁷Cs soil to cow’s milk transfer coefficient) and dose due to radiocaesium ingestion derived from WBC measurements of radiocaesium body burden (Drozdovitch et al. 2022b; Minenko et al. 2006). For subjects with WBC measurements (n = 166, 15.6% of the total), the measurement-based doses from radiocaesium ingestion were calculated by adjusting the model-based doses using the result of WBC measurements. Measurement-based doses were calculated only for postnatal exposure as study subjects, not their mothers, were measured by WBCs in 1987–1991. Results of WBC measurements of mothers during pregnancy between 26 April 1986 and 31 March 1987 were not available, and therefore, prenatal measurement-based doses from radiocaesium ingestion could not be calculated.

For each subject, two sets of model-based and two sets of measurement-based thyroid doses (when measurements of ¹³¹I thyroid activity or WBC radiocaesium body-burden were available) were calculated for exposure due to ¹³¹I intake and due to radiocaesium ingestion using personal information obtained during the first and the second interviews. The same results of measurements and parameter values for the dosimetry models were used for each study subject to calculate the corresponding thyroid doses; the only values that were different in the calculation of these sets of doses were those related to individual relocation history and dietary data that were reported during the two interviews. A detailed description of the dosimetry models to calculate model-based and measurement-based pre- and postnatal thyroid doses is provided in (Drozdovitch et al. 2020).

Statistical analysis

Here, kappa statistics was used as the main measure of agreement for qualitative (categorical) variables. Kappa statistics κ < 0 indicates no agreement, while a 0–0.20 range for κ corresponds to slight, 0.21–0.40 to fair, 0.41–0.60 to moderate, 0.61–0.80 to substantial, and 0.81–1.0 to almost perfect agreement (Landis and Koch 1977). In addition, the nonparametric Spearman’s rank-correlation coefficient, r_s, was used to measure the degree of agreement of the responses for whole numbers, i.e., number of settlements of residence and gestational age at birth. Values of |r_s| between 0 and 0.3 indicate a weak correlation, between and 0.7—a moderate, and between 0.7 and 1.0 a strong correlation (Ratner 2009). For all responses, the percentage of agreement was also used as a measure for the degree of agreement.

For thyroid doses, the present study used the Jaccard index, also known as the Jaccard similarity coefficient (Jaccard 1912), J_sim. Same answers regarding date, place of residence, consumption rates of foodstuffs, breastfeeding and administration of stable iodine that were provided during the two interviews resulted in the same dose, referred to in this study as intersection of doses, D₁ ∩ D₂, calculated using information from the first and second interviews, D₁ and D₂, respectively. In contrast, different answers that were provided during the two interviews resulted in a difference between D₁ and D₂, which was numerically calculated as D₁ ∪ D₂ – D₁ ∩ D₂. Values of J_sim range from 0 to 1.0, where 0 means complete discrepancy and 1.0 means complete agreement. J_sim was calculated as [Eq. (1)]

$$J_{sim}=\frac{D_1\cap D_2}{D_1\cup D_2-D_1\cap D_2}.$$ (1)

The differences in responses about consumption rates from two interviews and two sets of thyroid doses were evaluated using the Wilcoxon test, because values were not normally distributed; the p_w value represents the significance level of whether data sets differ.

Results

Questionnaire data

Table 2 presents data describing the extent of agreement, in percent agreement, κ and r_s, between responses given by 1059 mothers of 1065 cohort members during the two interviews (Table 2).

Table 2.

Agreement in responses between two interviews

Characteristics	N	Agreed (%)	κ (rs)a	Degree of agreement
Name of settlement ATA	1059	92	0.921	Almost perfect
Mothers with direct thyroid measurements	205	95	0.950	Almost perfect
Mothers who breastfed child	422	92	0.916	Almost perfect
Mothers at first trimester ATA	157	93	0.929	Almost perfect
Mothers at third trimester ATA	470	92	0.916	Almost perfect
Ruralb	611	92	0.918	Almost perfect
Urbanb	400	93	0.920	Almost perfect
Time between interviews < 4.0 years	228	92	0.914	Almost perfect
Time between interviews ≥ 6.0 years	259	97	0.964	Almost perfect
Number of settlements of residence during 26 April–30 June 1986	1059	75	0.77	Strong
Mothers with direct thyroid measurements	205	76	0.83	Strong
Mothers who breastfed child	422	77	0.80	Strong
Mothers at first trimester ATA	157	74	0.80	Strong
Mothers at third trimester ATA	470	77	0.79	Strong
Rural	611	80	0.82	Strong
Urban	400	75	0.74	Strong
Time between interviews < 4.0 years	228	72	0.73	Strong
Time between interviews ≥ 6.0 years	259	82	0.84	Strong
Date of first relocationc	580	70	0.692	Substantial
Mothers with direct thyroid measurements	135	69	0.633	Substantial
Mothers who breastfed child	311	75	0.744	Substantial
Mothers at first trimester ATA	66	62	0.714	Substantial
Mothers at third trimester ATA	337	75	0.732	Substantial
Rural	377	76	0.731	Substantial
Urban	173	65	0.646	Substantial
Time between interviews < 4.0 years	115	61	0.596	Moderate
Time between interviews ≥ 6.0 years	177	86	0.826	Almost perfect
Name of settlement of first relocation	568	80	0.792	Substantial
Mothers with direct thyroid measurements	131	80	0.796	Substantial
Mothers who breastfed child	314	81	0.803	Almost perfect
Mothers at first trimester ATA	65	74	0.731	Substantial
Mothers at third trimester ATA	335	81	0.803	Almost perfect
Rural	359	84	0.833	Almost perfect
Urban	170	81	0.803	Almost perfect
Time between interviews < 4.0 years	112	73	0.720	Substantial
Time between interviews ≥ 6.0 years	173	90	0.893	Almost perfect
Residential history for evacuees
Name of settlement of residence ATA	53	96	0.961	Almost perfect
Number of settlements of residence during 26 April–30 June 1986	53	57	0.77	Strong
Date of evacuationc	50	68	0.570	Moderate
Name of first settlement of residence after evacuation	50	68	0.715	Substantial
Source of milk ATA	1044	64	0.455	Moderate
Mothers with direct thyroid measurements	204	68	0.530	Moderate
Mothers who breastfed child	422	67	0.517	Moderate
Mothers at first trimester ATA	157	55	0.364	Fair
Mothers at third trimester ATA	470	66	0.512	Moderate
Rural	604	66	0.388	Fair
Urban	393	63	0.451	Moderate
Time between interviews < 4.0 years	226	54	0.295	Fair
Time between interviews ≥ 6.0 years	255	76	0.651	Substantial
Consumption rate ATA of privately owned cow milk, V26April	380	71	0.498	Moderate
Mothers with direct thyroid measurements	70	75	0.586	Moderate
Mothers who breastfed child	146	76	0.556	Moderate
Mothers at first trimester ATA	44	66	0.499	Moderate
Mothers at third trimester ATA	163	75	0.551	Moderate
Rural	301	71	0.499	Moderate
Urban	68	68	0.492	Moderate
Time between interviews < 4.0 years	85	62	0.315	Fair
Time between interviews ≥ 6.0 years	99	80	0.661	Substantial
Consumption rate ATA of milk from a trade network, V26April	197	72	0.409	Moderate
Mothers with direct thyroid measurements	42	86	0.734	Substantial
Mothers who breastfed child	88	70	0.483	Moderate
Mothers at first trimester ATA	34	71	0.444	Moderate
Mothers at third trimester ATA	95	67	0.436	Moderate
Rural	53	75	0.592	Moderate
Urban	142	70	0.424	Moderate
Time between interviews < 4.0 years	30	60	0.322	Fair
Time between interviews ≥ 6.0 years	62	79	0.612	Substantial
Consumption rate of privately owned cow milk averaged over 26 April–10 May 1986,Vaver	536	63	0.401	Moderate
Mothers with direct thyroid measurements	97	60	0.389	Fair
Mothers who breastfed child	200	68	0.450	Moderate
Mothers at first trimester ATA	68	65	0.499	Moderate
Mothers at third trimester ATA	219	66	0.445	Moderate
Rural	401	65	0.439	Moderate
Urban	114	56	0.318	Fair
Time between interviews < 4.0 years	125	59	0.346	Fair
Time between interviews ≥ 6.0 years	129	68	0.497	Moderate
Consumption rate of milk from a trade network averaged over 26 April–10 May 1986,Vaver	329	67	0.364	Fair
Mothers with direct thyroid measurements	67	75	0.492	Moderate
Mothers who breastfed child	142	75	0.452	Moderate
Mothers at first trimester ATA	57	67	0.315	Fair
Mothers at third trimester ATA	151	74	0.443	Moderate
Rural	116	68	0.359	Fair
Urban	202	65	0.360	Fair
Time between interviews < 4.0 years	59	66	0.224	Fair
Time between interviews ≥ 6.0 years	96	68	0.399	Fair
Milk products consumption during 26 April–30 June 1986
Yes/No	1003	97	0.421	Moderate
Consumption rate of milk products	947	61	0.367	Fair
Leafy vegetables consumption during 26 April–30 June 1986
Yes/No	1035	83	0.501	Moderate
Date of beginning of consumption	708	38	0.188	Slight
Date of ending of consumption	15	60	0.444	Moderate
Consumption rate	708	46	0.250	Fair
Stable iodine administration during 26 April–31 May 1986
Yes/No	1015	90	0.622	Substantial
Date of beginningc	88	57	0.060	Slight
Duration	89	46	0.250	Fair
Gestational age at birth	1053	99	0.99	Strong
Breastfeeding during 26 April–31 May 1986
Yes/No	450	98	0.753	Substantial
Duration	421	82	0.732	Substantial
Breastfeeding during 26 April 1986–31 March 1987
Yes/No	1065	97	0.793	Substantial
Duration	972	83	0.753	Substantial
Residential history after 30 June 1986
Name of settlement of residence on 1 July 1986	1065	87	0.366	Fair
Number of settlements of residence after 30 June 1986	1065	67	0.66	Moderate
Name of settlement of residence on 31 March 1992d	1065	91	0.901	Almost perfect
Consumption rates by child until age 1 year
Consumption rate of privately owned cow milk	306	58	0.336	Fair
Consumption rate of milk from a commercial trade network	230	55	0.300	Fair
Consumption rate of milk products	790	48	0.164	Slight
Consumption rates by child at age 1–2 years
Consumption rate of privately owned cow milk	405	56	0.316	Fair
Consumption rate of milk from a commercial trade network	481	59	0.304	Fair
Consumption rate of milk products	888	48	0.213	Fair
Consumption rates by child at age 2–5 years
Consumption rate of privately owned cow milk	450	55	0.352	Fair
Consumption rate of milk from a commercial trade network	644	51	0.277	Fair
Consumption rate of milk products	893	50	0.195	Slight

ATA at the time of the accident

^aKappa statistic, κ (shown in plain numbers), and Spearman’s rank-correlation coefficient, r_s (shown in italic numbers), provide measure of agreement

^bType of place of residence at the time of the accident

^cExact date, date ± 1 day, or period

^dEnd of period for dose reconstruction

Residential history between 26 April and 30 June 1986

Reliable information on residential history is crucial for thyroid dose calculation. The deposition of ¹³¹I on the ground surface in the settlement of residence is the starting point of the ecological model that describes the processes of ¹³¹I transfer through the environment into the thyroid gland.

Name of the settlement ATA

Overall, there was an outstanding agreement in naming the settlement ATA as responses provided in the two interviews agreed in 977 mothers (92% of all mothers), with κ = 0.921, which was regarded as “almost perfect”. The high level of agreement varied little across the various subsets; the highest agreement was observed when the time elapsed between interviews was 6.0 years and more (97%, κ = 0.964), while the lowest was observed when the time elapsed between interviews was less than 4.0 years (92%, κ = 0.914).

Number of settlements of residence from 26 April to 30 June 1986

Levels of reliability in interview responses regrading number of settlements during this 2-month period were generally lower than those for naming of the settlements ATA but were considered mostly substantial. The same number of settlements was reported by 797 respondents (75% agreed, Spearman’s rank-correlation coefficient r_s=0.77, p<0.001) (Table 2). Levels of agreements tended to be higher when the time elapsed between interviews was 6.0 years and more (82% agreed, r_s=0.84, p<0.001), for mothers lived in rural area (80% agreed, r_s=0.82, p<0.002), and when mothers had direct thyroid measurements (76%, r_s=0.83 p<0.005). It is of note that 422 (40%) and 439 (41%) mothers, at the first and second interviews, respectively, reported that they did not move their place of permanent residence between 26 April through 30 June 1986 (not shown in Table 2).

Date of first relocation

Among mothers who changed their residence after the accident, the most important contribution to the thyroid dose occurred, while they were in the settlement ATA and accumulated further until first relocation. There was a substantial agreement in reporting on date of first relocation: with an overall 70% agreement and κ = 0.692 (Table 2). At the first interview 137 (13%) of mothers reported having relocated within 5 days of the accident (between 26 and 30 April 1986); 110 (10%) mothers reported the same at the second interview. It should be noted that most of the ¹³¹I intake with cow’s milk took place before 10 May 1986. At the first interview 236 (22%) of mothers reported that they had first relocated between 1 and 10 May 1986, while 237 (22%) mothers did so at the second interview. There was an almost perfect consistency in the answers (86%, κ = 0.826) when the time elapsed between interviews was more than 6.0 years, while there was only a moderate agreement when less than 4 years elapsed (61%, κ = 0.596).

Name of settlement of first relocation

A substantial agreement was found in naming the settlement to which mothers were first relocated; the same name of the settlement was reported at the first and second interviews by 80% of women (κ = 0.792) who reported their first relocation between 26 April and 30 June 1986. Therefore, the reproducibility of name of settlement of first relocation was somewhat better than reporting date of relocation but tended to be lower when mothers were in the first trimester ATA and the time elapsed between interviews was less than 4 years (Table 2).

Residential history for evacuees from the 30-km zone

A separate analysis of residential history was conducted for the residents of the 30-km zone around the Chernobyl Nuclear Power Plant who were evacuated shortly after the accident. These settlements were the most heavily contaminated with ¹³¹I, and therefore, evacuees received the highest thyroid doses from ¹³¹I. At the first interview, 54 (5.1% of the total) mothers reported having evacuated from the 30-km zone, while at the second interview, 58 (5.5%) women did so. Among the evacuees, an almost perfect agreement was observed for naming place of residence ATA (agreement 96%; κ = 0.961); the same number of settlements of residence during 26 April–30 June 1986 was reported by 57% (r_s=0.77, p<0.001); the agreement was moderate on date of evacuation (68% agreed, κ = 0.570) and substantial on name of first settlement after evacuation (68% agreed, κ = 0.715).

Consumption of cow’s milk

It is necessary to distinguish two sources of milk consumed, i.e., milk from privately owned cows and milk from commercial trade network. This is because the ¹³¹I concentration in milk from privately owned cows was generally higher than that in milk from a commercial trade network, which was collected from neighborhood areas and processed in a local dairy. In the present study, daily consumption rates at ATA (V_26April) as well as averaged over the period 26 April–10 May 1986 (V_aver) were considered. This was the period during which most of the changes in milk consumption occurred and most of the ¹³¹I intake with milk took place. The average consumption rate was calculated as [Eq. (2)]

$$V_{\mathit{\text{aver}}}=\frac{1}{15}\cdot \sum _{i=1}^{15}{V}_i,$$ (2)

where V_aver is the consumption rate averaged over the period from 26 April to 10 May 1986 (L d⁻¹); i=15 is the number of days in the time interval from 26 April through 10 May; V_i is the consumption rate of milk on day i (L d⁻¹).

Source of milk ATA

There was a moderate degree of agreement in identifying the source of milk reported at the two personal interviews (64% agreed, κ = 0.455) (Table 2). The agreement was only fair when the time elapsed between interviews was less than 4.0 years (54%, κ = 0.295) and when mothers were in the first trimester (55%, κ = 0.364).

Consumption rate ATA

Significantly higher consumption rates (L d⁻¹) of milk from privately owned cows were reported at the first vs. second interviews, 0.68 L d⁻¹ vs. 0.59 L d⁻¹, for arithmetic mean, respectively, p_w<0.001 (Table 3). Practically the same consumption rate of cow’s milk from a commercial trade network was reported during both interviews (0.34 L d⁻¹ vs. 0.33 L d⁻¹ for arithmetic mean, respectively, p_w=0.203) (Table 3).

Table 3.

Consumption rates reported during the first and second interviews

Consumed foodstuff	N	Consumption rate reported during the first interview (L(kg) d−1)		Consumption rate reported during the second interview (L(kg) d−1)		pw valuea	r s b	Agreed (%)c
		Mean ± SD	Median (range)	Mean ± SD	Median (range)
Mother’s consumption
Privately owned cow milk ATA	380	0.68 ± 0.58	0.50 (0.04–3.0)	0.59 ± 0.50	0.50 (0.02–3.0)	< 0.001	0.66	71
Milk from a trade network ATA	197	0.34 ± 0.25	0.25 (0.02–1.3)	0.33 ± 0.32	0.25 (0.01–2.0)	0.208	0.60	72
Privately owned cow milk, averagedd	536	0.55 ± 0.55	0.43 (0.01–3.0)	0.52 ± 0.47	0.43 (0.01–3.0)	0.603	0.62	63
Milk from a trade network, averagedd	329	0.23 ± 0.21	0.15 (0.01–1.3)	0.27 ± 0.29	0.15 (0.01–2.0)	0.004	0.53	67
Milk products	947	0.18 ± 0.15	0.18 (0.04–2.0)	0.18 ± 0.12	0.19 (0.003–1.0)	0.137	0.46	61
Leafy vegetables	708	0.043 ± 0.038	0.025 (0.001–0.3)	0.038 ± 0.038	0.025 (0.001–0.3)	< 0.001	0.34	46
Child’s consumption
Privately owned cow milk (at age < 1y)	306	0.27 ± 0.18	0.25 (0.02–1.5)	0.24 ± 0.17	0.25 (0.02–2.0)	0.010	0.38	58
Privately owned cow milk (at age 1–2 years)	405	0.34 ± 0.20	0.25 (0.04–1.3)	0.30 ± 0.19	0.25 (0.02–2.0)	0.001	0.40	56
Privately owned cow milk (at age 3–5 years)	450	0.39 ± 0.22	0.28 (0.02–1.3)	0.34 ± 0.20	0.25 (0.02–1.3)	< 0.001	0.45	55
Milk from a trade network (at age < 1y)	230	0.22 ± 0.13	0.25 (0.02–1.0)	0.20 ± 0.11	0.13 (0.04–0.75)	0.056	0.32	55
Milk from a trade network (at age 1–2 years)	481	0.26 ± 0.15	0.25 (0.02–1.3)	0.25 ± 0.13	0.25 (0.04–1.0)	0.090	0.26	59
Milk from a trade network (at age 3–5 years)	644	0.30 ± 0.16	0.25 (0.02–1.0)	0.27 ± 0.15	0.25 (0.02–1.0)	0.001	0.38	51
Milk products (at age < 1y)	790	0.08 ± 0.08	0.05 (0.004–0.8)	0.10 ± 0.08	0.10 (0.002–0.75)	< 0.001	0.19	48
Milk products (at age 1–2 years)	888	0.12 ± 0.07	0.10 (0.004–0.60)	0.13 ± 0.07	0.10 (0.004–0.40)	0.002	0.25	48
Milk products (at age 3–5 years)	893	0.15 ± 0.07	0.20 (0.004–0.50)	0.15 ± 0.06	0.20 (0.004–0.50)	0.029	0.25	50

ATA at the time of the accident, SD standard deviation

^ap_w value represents the significance level of whether data sets differ according to Wilcoxon test

^bSpearman’s rank-correlation coefficient

^cPercent of agreement between two responses

^dAveraged for time period between 26 April and 10 May 1986 [see Eq. (1)]

There was a moderate agreement in the rate at which mothers reportedly consumed milk from privately owned cows ATA (71% agreed, κ = 0.498) (Table 2). The agreement was better when the time elapsed between interviews was more than 6.0 years (80%, κ = 0.661) and among mothers who had direct thyroid measurements (75%, κ = 0.586). Likewise, there was a similar degree of agreement in reporting of consumption rate of milk from a commercial trade network (72%, κ = 0.409), although there was more variability in agreement among different subsets of responders: the best agreement was in responses from mothers with direct thyroid measurements (86%, κ = 0.734), while the worst agreement was when the time elapsed between interviews was less than 4 years (60%, κ = 0.332).

Average milk consumption between 26 April and 10 May 1986

The fractions of women who reported milk consumption during 26 April–10 May 1986 at the first and second interviews were similar: 42% and 44% for milk from privately owned cows; and 24% and 27% for milk from a commercial trade network, respectively. There was no statistically significant difference in the consumption rates of privately owned cow milk averaged over this period as reported at the first vs. second interviews, 0.55 L d⁻¹ vs. 0.52 L d⁻¹, respectively, p_w=0.603 (Table 3). The consumption rate of cow’s milk from a commercial trade network reported during the second interview was higher than that reported at the first interview, 0.27 L d⁻¹ vs. 0.23 L d⁻¹ for arithmetic mean, respectively, p_w=0.004 (Table 3).

As for consumption averaged over 26 April–10 May 1986, there was a moderate agreement in reported consumption rates of privately owned cow milk (63% agreed, κ = 0.401). The agreement was better when the time elapsed between interviews was more than 6 years (68%, κ = 0.497) and for mothers who breastfed their child (68%, κ = 0.450). An agreement in reported consumption rates of cow’s milk from a commercial trade network averaged over the same period was similar (67% agreed, κ = 0.364). The best agreement was for mothers with direct thyroid measurements (75%, κ = 0.492), for those who breastfed their child (75%, κ = 0.452) or who were at third trimester of pregnancy (74%, κ = 0.443).

Consumption of milk products

There was a fair agreement in the answers on the consumption rate of milk products during the period of 26 April–30 June 1986 (61%, κ = 0.367) (Table 2). The same consumption of milk products was reported during both interviews, 0.18 kg d⁻¹ for arithmetic mean, p_w=0.137 (Table 3).

Consumption of leafy vegetables

For consumption of leafy vegetables, the overall agreement in answers from the first and second interviews was moderate in general (83%, κ = 0.501), but the fraction of mothers who reported leafy vegetables consumption during both interviews was practically the same: 76% vs. 78%. The agreement was fair for consumption rate (46%, κ = 0.250); reported consumption rates were lower at the second interview (0.038 kg d ⁻¹) than at the first (0.043 kg d⁻¹) (p_w <0.001) (Table 3). There was a slight agreement in the start date of consumption (38%, κ = 0.188).

Stable iodine administration between 26 April and 31 May 1986

During both interviews, 167 (16%) women reported that they had taken stable iodine for prophylactic purposes, with a substantial agreement observed for 90% of respondents (κ = 0.622), and a fair agreement on duration of intake (46% agreed, κ = 0.250). However, there was only a slight agreement on the start date of the stable iodine administration (κ = 0.060), with a trend towards earlier dates after the accident reported during the second interview.

Gestational age at birth and breastfeeding during 26 April–31 May 1986

Information on gestational age at birth reported during the two interviews was very consistent and with strong correlation (99% of women, r_s=0.99, p<0.001) (Table 2). In addition, a substantial agreement was observed for breastfeeding of child; the responses agreed for 98% (κ = 0.753) and 97% (κ = 0.793) of women for periods 26 April–31 May 1986 and 26 April 1986–31 March 1987, respectively. Agreements on duration of breastfeeding were also substantial, with answers agreed for 82% (κ = 0.732) and 83% (κ = 0.753) of women for these time periods.

Residential history after 30 June 1986

A reliable residential history after 1 June 1986, the end of the ¹³¹I exposure period, is important for calculating the thyroid dose due to external irradiation and ingestion of radiocaesium isotopes by a child until age 5 years. Agreement on names the settlement on 1 July 1986 was fair (κ = 0.366), but the same number of settlements of residence after 30 June 1986 was reported by 67% of women (r_s = 0.66, p < 0.001) during the two interviews (Table 2). Responses on place of residence on 31 March 1992 showed an almost perfect agreement (91% answers agreed, κ = 0.901).

Consumption rates by children

Consumption rates by children of milk from privately owned cows, milk from a commercial trade network, and milk products, were used to calculate the thyroid dose due to ingestion of radiocaesium isotopes to a child until age 5 years. There were fair agreements in reporting of the consumption rates of milk from privately owned cows by child at age less than 1 year (58% answers agreed, κ = 0.336), 1–2 years (56% agreed, κ = 0.316), and 3–5 years (55% agreed, κ = 0.352), while the agreement for the consumption rates of milk from a commercial trade network was worse, i.e., κ = 0.300, κ = 0.304, and κ = 0.277 for the same age groups, respectively (Table 2). The consumption rates of milk from privately owned cows and cow’s milk from a commercial trade network reported for children during the second interview were slightly lower than those reported during the first interview (Table 3).

Thyroid doses

Individual prenatal and postnatal thyroid doses, model-based and measurement-based, were calculated using residential history and dietary data obtained during the first and second interviews. For model-based total thyroid doses from ¹³¹I intake (Table 4, upper panel), the Jaccard similarity coefficient, J_sim, showed an arithmetic mean ± standard deviation (SD) of 0.45 ± 0.34 (median = 0.39). A better agreement was found for doses due to external irradiation, J_sim=0.82 ± 0.23 (median = 0.90), and ingestion of radiocaesium isotopes, J_sim=0.84 ± 0.24 (median = 0.96). The agreement for measurement-based doses (Table 4, lower panel) was better than for model-based doses: J_sim=0.78 ± 0.29 (median = 0.93) for total doses from ¹³¹I intake and J_sim=0.91 ± 0.19 (median = 1.0) for postnatal doses due to ingestion of ¹³⁴Cs and ¹³⁷Cs. The Jaccard similarity coefficient for prenatal thyroid dose from ¹³¹I intake (the main exposure pathway) did not correlate with dose-values calculated using the first and second interviews’ data for either model-based dose (Pearson correlation coefficient was r_p=0.032, p=0.363 and r_p=0.044, p=0.201, respectively) or measurement-based dose (r_p=0.073, p=0.320 and r_p=0.093, p=0.202, respectively) (not shown).

Table 4.

Thyroid doses for the Belarusian cohort of individuals exposed in utero calculated using individual behavior and consumption data reported by their mothers during the two interviews

Exposure pathways and time period	Thyroid doses calculated using data reported during the first interview (mGy)		Thyroid doses calculated using data reported during the second interview (mGy)		pw valuea	r s b	J sim c
	Mean ± SD	Median (range)	Mean ± SD	Median (range)
Model-based dose
Prenatal from 131I intakee	184±322	104 (2 × 10−3–5576)	178±433	80 (2 × 10−3–10,570)	< 0.001	0.83	0.46
Postnatal from 131I intakee	33 ± 92	2.4 (1 × 10−3–969)	36 ± 95	2.2 (1 × 10−3–969)	0.565	0.88	0.52
Total dose from 131I intake	191±319	116 (2 × 10−3–5576)	185±428	91 (2 × 10−3–10,570)	0.002	0.79	0.45
Prenatal due to external irradiation	1.7 ± 5.2	0.66 (3 × 10−3–101)	1.7 ± 5.1	0.66 (3 × 10−3–95)	0.187	0.94	0.72
Postnatal due to external irradiation	3.4 ± 4.2	2.1 (4.2 × 10−2–71)	3.5 ± 4.6	2.1 (4.0 × 10−2–71)	0.065	0.98	0.88
Total dose due to external irradiation	5.1 ± 7.3	3.0 (4.5 × 10−2–102)	5.1 ± 7.5	2.9 (7.1 × 10−2–101)	0.266	0.97	0.82
Prenatal from 134 and 137Cs ingestion	0.68 ± 0.96	0.36 (1 × 10−3–11)	0.68 ± 0.95	0.37 (1 × 10−3–11)	0.669	0.97	0.82
Postnatal from 134 and 137Cs ingestion	1.6 ± 2.2	1.0 (1.0 × 10−2–21)	1.6 ± 2.1	1.0 (1.2 × 10−2–20)	0.153	0.98	0.88
Total dose from 134 and 137Cs ingestion	2.3 ± 2.7	1.6 (1.4 × 10−2–25)	2.2 ± 2.6	1.6 (2.1 × 10−2–26)	0.729	0.97	0.84
Measurement-based dose
Prenatal from 131I intaked	594±1497	100 (1 × 10−3–14,754)	590 ± 1.622	98 (1 × 10−3–17,507)	0.150	0.99	0.80
Postnatal from 131I intaked	216 ± 443	7.4 (7 × 10−3–1841)	427 ± 1.280	13 (7 × 10−3–6820)	0.043	0.97	0.47
Total dose from 131I intaked	634±1500	153 (1 × 10−3–14,754)	665 ± 1.694	133 (1 × 10−3–17,507)	0.305	0.99	0.78
Postnatal from 134 and 137Cs ingestion	2.8 ± 2.8	2.1 (0.31–21)	3.0 ± 3.9	2.1 (0.32–39)	0.741	0.97	0.91

SD standard deviation

^ap_w value represents the significance level of whether data sets differ according to Wilcoxon test

^bSpearman’s rank-correlation coefficient, p<0.001 for all correlation coefficient-values

^cArithmetic mean of the Jaccard similarity coefficient between two dose-values, including zero values

^dExcluded zero dose-values

Figure 2a (left panel) compares the model-based prenatal ¹³¹I thyroid doses based on the first and second interview data, while Fig. 2b (right panel) provides a similar comparison for the measurement-based prenatal doses, calculated for 205 individuals whose mothers had measurements of ¹³¹I thyroid activity. The arithmetic mean ± SD of the ratios of the individual model-based dose calculated using data from the second to the first interview was 1.7 ± 4.0 and the corresponding median of the ratios was 1.0. These values compare with 1.1 ± 0.7 and 1.0 for measurement-based dose, respectively. In other words, the model-based thyroid doses show a wider spread in the ratios derived from different interviews than the measurement-based thyroid doses. For 76% of model-based doses, the two sets of estimates agreed within a factor of 3, compared with 95% of measurement-based doses. A difference of less than 10% in model-based doses was obtained for 22% of the subjects, compared with 63% for measurement-based doses.

Fig. 2.

Comparison of prenatal thyroid doses due to ¹³¹I calculated using information from the first and second interviews: a model-based doses; b measurement-based doses. Solid line indicates agreement between doses deduced based on data from first and second interviews, while dashed lines indicate a factor of 3 difference

Table 5 shows the overall agreement between doses from different exposure pathways calculated using data from the two interviews. A reasonable agreement was observed for model-based thyroid doses from ¹³¹I intake, here the doses agreed within a factor of 2 in 63.9%, 73.7% and 63.3% for prenatal, postnatal, and total doses, respectively (r_s=0.83, r_s=0.88, r_s=0.79, respectively, p<0.001, see Table 4). A substantial difference in total model-based thyroid doses from ¹³¹I intake (> 10 times) was found for 50 out of 854 (5.9%) individuals (Table 5). In general, a better agreement was found for measurement-based doses from ¹³¹I; they agreed within a factor of 2 in 90.5%, 69.7% and 89.9% for prenatal, postnatal, and total doses (r_s=0.99, r_s=0.97, r_s=0.99, respectively, p<0.001, see Table 4). Although measurement-based thyroid doses from 131I intake calculated for the same individual using different interviews were rather consistent, a substantial difference (> 10 times) was found for 2 out of 205 (1.0%) individuals (Table 5) for whom the different terms (prematurity or post-term) of childbirth were reported in two interviews.

Table 5.

Overall agreement between thyroid doses from different exposure pathways calculated using individual behavior and consumptions data from two interviews

Period of exposure and exposure pathway	Percent of the study subjects for whom doses calculated using two interviews differed within a factor of					Range of ratios of dose calculated using second interview to dose calculated using first interview
	< 2	2–4.99	5–9.99	10–99.9	100 +
Model-based thyroid dose due to 131I intake
Prenatal dose due to 131I intake by mother	63.9	24.5	7.1	4.4	0.1	8.3 × 10−3–51
Postnatal dose due to 131I intake	73.7	14.1	5.0	5.3	1.9	7.6 × 10−4–5480
Total dosea	63.3	23.7	7.1	5.5	0.4	2.8 × 10−3–560
Measurement-based dose due to 131I intake
Prenatal dose due to 131I intake by mother	90.5	6.9	1.6	1.0	–	0.027–6.4
Postnatal dose due to 131I intake	69.7	18.2	9.1	3.0	–	0.15–12
Total dosea	89.9	7.0	2.1	1.0	–	0.027–6.4
Model-based thyroid dose from external irradiation
Prenatal dose	92.3	6.3	0.6	0.8	–	0.012–53
Postnatal dose	98.2	1.2	0.3	0.3	–	0.15–29
Total dosea	97.5	1.9	0.5	0.1	–	0.08–8.6
Model-based thyroid dose due to Cs ingestion
Prenatal dose due to Cs ingestion by mother	94.6	3.7	1.3	0.4	–	0.025–6.8
Postnatal dose due to Cs ingestion	97.3	2.3	0.2	0.2	–	0.16–14
Total dosea	97.7	2.2	–	0.1	–	0.025–3.6
Measurement-based thyroid dose due to Cs ingestion
Postnatal dose due to Cs ingestion	99.4	0.6	–	–	–	0.58–2.9

Cs—134Cs and 137Cs

^aSum of prenatal and postnatal doses

For model-based thyroid doses due to external irradiation and ingestion of radiocaesium isotopes, agreement within a factor of 2 was observed in more than 92% (Table 5). For measurement-based thyroid doses due to ingestion of radiocaesium isotopes, agreement within a factor of 2 was observed in 99.4% (Table 5), r_s=0.97, p<0.001 (Table 4).

Discussion

For the cohort of 2965 Belarusian individuals exposed in utero and during early years of life to radioactive fallout from the Chernobyl accident in April 1986 (Yauseyenka et al. 2020), individual thyroid radiation doses were estimated using residential history and dietary consumption data obtained by personal interviews with mothers of cohort subjects, supplemented in part by measured ¹³¹I thyroid activity data (Drozdovitch et al. 2020). In the present study, the reproducibility of individual behavior and dietary data reported during repeat interviews of sampled mothers of the cohort subjects (n = 1059) was evaluated. First interviews were conducted on average 29 years after the accident, while second repeat interviews were conducted 34 years after the accident. Agreement in response from the two interviews were “substantial” or “almost perfect” for residential history during the period of exposure to ¹³¹I (26 April–30 June 1986) and mainly “substantial” for consumption of milk from privately owned cows and cow’s milk from a commercial trade network ATA. Answers to questions on consumption of milk products and leafy vegetables during 26 April–30 June 1986 were less reproducible, and the degree of agreement was only “fair”. The answers agreed better to questions on the names of settlements, number of relocations, and consumption pattern of cow’s milk rather than on dates of relocations and consumptions of milk products and leafy vegetables. When thyroid doses based on repeated interview data were compared, a moderate agreement in total (prenatal and postnatal) model-based thyroid doses due to ¹³¹I intake was found, while for measurement-based doses, a substantial agreement was observed. For external irradiation and exposure due to ingestion of ¹³⁴Cs and ¹³⁷Cs, there was an almost perfect agreement in both model- and measurement-based doses.

It was found that the reproducibility of the answers of mothers was better:

• For the mothers in the third trimester of pregnancy ATA than those for mothers in the first trimester of pregnancy ATA. According to numerous studies, the level of anxiety in the third trimester is higher compared to the level of anxiety at the beginning of the gestation period (e.g., Da Costa et al. 1999; Priyambada et al. 2017). Since the most memorable events in life are traumatic events (Babel 2017; Eysenck et al. 2007), it can be assumed that women remembered events that occurred in the third, most disturbing, semester of pregnancy better, and, therefore, provided more accurate answers.

• When the time span between interviews was longer. This result contradicts with other studies (Cui et al. 2021; Riboli et al. 1997). More research is needed to understand and explain this unexpected observation.

• For the questions related to mother’s own dietary habits during the pregnancy than for those related to the diet of her child aged up to 5 years (Tables 2 and 3). This contradicts with some other studies that reported concerns with regard to errors in autobiographical dietary recall by adults (Marshall 2005; Wu et al. 1988) and that reported that a mother could provide the most reliable information on the diet when her child was under age 10 years (Baranowski et al. 2012; Burrows et al. 2010).

• For women who resided ATA in rural settlements rather than in urban settlements.

It was further found that lower own consumption rates of cow’s milk were reported by mothers during the second interview (Table 3). In the literature it was reported that recall of past diet is strongly influenced by present dietary habits (Dwyer and Coleman 1997; Rohan and Potter 1984). Unfortunately, information on respondents’ diet at the time of the interview was not available. However, more people may recognize lactose intolerance at a later age that might cause reporting lower consumption rates during the second interview.

This study used the same questionnaire during the first and second interviews by the same trained interviewers. This is important to note, because the wording of the questionnaire, the level of detail of the questions and the interview process may influence recall ability during data collection (Friedenreich 1994). Because the individual questionnaire data were collected in the present study in the same way as before, it is believed that the respondents’ recall ability was not affected.

Recall accuracy obtained in the present study was compared with that observed in the Belarusian–American cohort of 11,732 individuals exposed in childhood and adolescence (Drozdovitch et al. 2016). Table 6 shows the consistency of answers between two personal interviews of mothers in the Belarusian cohort with those of mothers in the Belarusian–American cohort: 1065 pairs of questionnaires for 1059 mothers from the Belarusian cohort of individuals exposed in utero and 2664 pairs of questionnaires for 1994 mothers of individuals from the Belarusian–American cohort. It was found that the consistency of responses was better among the mothers in the Belarusian cohort of individuals exposed in utero than that among the mothers in the Belarusian–American cohort members. Considering that the recollection time for the Belarusian cohort of individuals exposed in utero was twice as long as that for the Belarusian–American cohort (29 years vs. 13 years and 34 years vs. 17 years for the first and second interviews, respectively), the memory recall was more accurate when women were asked about unique events in her life, which were pregnancy and childbirth around the time of nuclear reactor accident.

Table 6.

Consistency of answers between two personal interviews for the Belarusian cohort of individuals exposed in utero and Belarusian–American (BelAm) cohort

Characteristics	In utero cohort		BelAm cohort
	Agreed (%)a	κb (rsc)	Agreed (%)	κb (rsc)
Residential history
Name of settlement ATA	92	0.921	89	0.842
Number of settlements of residence	75	0.77	55	0.65
Date of first relocation	70	0.692	50	0.411
Name of settlement of first relocation	80	0.792	61	0.564
Cow milk consumption
Source of milk ATA	64	0.455	60	0.449
Consumptions of privately owned cow milk ATA	71	0.498	59	0.393
Mean consumptions of privately owned cow milkd	63	0.401	58	0.410
Consumptions of milk from a trade network ATA	72	0.409	75	0.417
Mean consumptions of milk from a trade networkd	67	0.364	61	0.331
Milk products
Mean consumption rate of milk products	61	0.327	41	0.149
Leafy vegetables consumption
Date of beginning of consumption	38	0.188	27	0.061
Date of ending of consumption	60	0.444	28	0.013
Mean consumption rate	46	0.250	60	0.241
Stable iodine administration
Yes/No	90	0.622	79	0.433
Date of beginning	57	0.060	28	0.298
Duration	46	0.250	45	0.210

ATA at the time of the accident

^aPercent of agreement between answers from two questionnaires or estimated dose-values

^bKappa statistics, shown as a plain text

^cSpearman’s rank-correlation coefficient, shown in Italic

^dAveraged for time period between 26 April and 10 May 1986 [see Eq. (1)]

The agreement between thyroid doses due to external irradiation and ingestion of radiocaesium isotopes was better than that for thyroid doses from ¹³¹I intake for both model-based and measurement-based doses. This can be explained by the fact that the dose reconstruction model for ¹³¹I intake (Drozdovitch et al. 2013) is more complex and required more detailed input data especially for the first few weeks after the accident (relocation history, consumptions that changed frequently) than those for external irradiation and ingestion of radiocaesium isotopes (Drozdovitch et al. 2022b; Minenko et al. 2006) that lasted for years.

The strength of the present study is the sufficient size of the investigated cohort including 1059 women who were pregnant and delivered child shortly after the Chernobyl accident. The use of the same questionnaire administrated during the first and second interviews by the same trained interviewers gives confidence that the data were collected in the same uniform way and that the respondents’ ability to recall information was not affected by the study instruments and design. However, unlike the other study (Drozdovitch et al. 2022a), the present study was unable to test the true validity of the responses in the absence of individual data collected shortly after the accident that can be less prone to recall error. In the present study, it was only possible to test the reproducibility of responses during repeated personal interviews.

Conclusions

The present study clearly showed that uncertainties due to human factors arising from poor memory recall in answering questions during personal interviews, are an important source of errors in questionnaire-based dose reconstruction for individuals enrolled in radiation epidemiological studies. For the model-based doses, such errors could result in variation of the possible dose-values within wide range, which, in turn, can distort radiation-related risk assessment based on these doses.

The present study also showed that estimation of reliable model-based doses requires a high quality of individual data collected through personal interviews. When individual radiation measurements (in this study, measurements of ¹³¹I thyroid activity and WBC measurements) are available, the quality of data on individual behavior and dietary habits has, in general, a relatively small influence on the quality of those measurement-based dose estimates. The lessons learned from this study are that (i) individual measurements of radionuclides in the human body represent the most valuable information for estimating radiation doses, and (ii) whenever a radiation accident occurs, at least some of the affected people should be asked to keep a diary, if at all possible.

It was found that consistency of answers between two personal interviews was better among mothers of individuals exposed in utero than among mothers in the Belarusian–American cohort who were exposed in childhood and adolescence. It is concluded that memory recall appears to be reliable if a mother was asked about unique events in her life, such as pregnancy and childbirth, around the time of nuclear reactor accident.

The approach presented in this study to quantify errors due to poor memory recall in assessment of radiation exposure that occurred long time ago can be applied to any other questionnaire-based studies in radiation or occupational epidemiology.