Lymphoma, multiple myeloma and leukaemia incidence in regions of Belarus most heavily contaminated by the Chernobyl accident

Abstract

There is little information on non‐thyroid cancer risks, including haematological malignancies (HM), among the residents of most contaminated regions after the Chernobyl (Chornobyl) nuclear power plant accident. We studied the incidence of lymphoma, multiple myeloma and leukaemia in relation to the raion‐average age‐specific cumulative absorbed red bone marrow (RBM) dose among the residents of Gomel and Mogilev oblasts in Belarus, which were highly contaminated. The follow‐up period was 40 years (1978–2018). HM cases and population size data were received from the Belarusian national cancer registry and the state department of statistics. Our ecological study included 7328 lymphoma, 9476 leukaemia and 2003 multiple myeloma incident cases and 90.8 million person‐years in people who were born before the accident and have attained age <80 years old. The mean (median) RBM dose accumulated by December 31, 2018 was 14.2 (6.4) mGy. We found no evidence of increased risks of Hodgkin and non‐Hodgkin lymphoma, multiple myeloma or total leukaemia associated with two‐year lagged raion‐average cumulative RBM dose after adjustment for sex, attained age, urban/rural status and calendar period effects. There was a suggestion of an elevated relative risk of myeloid leukaemia per 100 mGy after exclusion of Gomel and Mogilev cities. Little evidence was found on interaction between selected factors, except sex, and RBM dose for each study outcome. Studies with individually reconstructed cumulative absorbed RBM doses are warranted to provide more insight on dose‐effect relationships between HM risk, specifically leukaemia, and protracted environmental exposure at a low dose range.

What's new?

The 1986 Chernobyl accident led to widespread radioactive contamination across parts of Europe, with elevated childhood leukaemia rates observed in heavily affected areas of Ukraine. Data on long‐term haematological and other non‐thyroid cancer risk in the general population in heavily contaminated regions, however, remain limited. Here, the authors examined rates of lymphoma, multiple myeloma and leukaemia in relation to bone marrow radiation exposure in affected regions of Belarus. Radiation exposure had little impact on blood cancer risk among people born before the accident. Males and those who were very young at the time of exposure may have an elevated risk of myeloid leukaemia.

Abbreviations

ALL

acute lymphoblastic leukaemia

APC

annual percent change

ATA

age at the time of the accident

CI

confidence interval

CLL

chronic lymphocytic leukaemia

df

degrees of freedom

ERR

excess relative risk

HL

Hodgkin lymphoma

HM

haematological malignancies

ICD‐10

the International Classification of Diseases, 10th Revision

IRR

incidence rate ratio

LL

lymphoid leukaemia

LR

likelihood ratio

ML

myeloid leukaemia

MM

multiple myeloma

NHL

non‐Hodgkin lymphoma

PYR

person‐years

RBM

red bone marrow

RR

relative risk

SIR

standardized incidence ratio

1. INTRODUCTION

The Chernobyl (Chornobyl) nuclear power plant accident on April 26, 1986, resulted in the widespread contamination of European territories with radioactive fission products. The most heavily contaminated regions were northern Ukraine, western Russia and southern Belarus. ¹ The majority of the released activity was in the form of short‐lived radionuclides, including isotopes of iodine (I) and tellurium, specifically ¹³¹I, ¹³²Te + ¹³²I, ¹³³I, ¹³⁵I, resulting in well‐documented exposure of the thyroid gland and a radiation‐related increase in thyroid cancer incidence in the exposed children and adolescents. ² , ³ , ⁴ While the deposition of relatively long‐lived radioisotopes, including ¹³⁷Cs with a half‐life of 30.2 years and ⁹⁰Sr with a half‐life of 28.8 years, was proportionally lower, these radionuclides are still present in the environment as a continuous source of radiation exposure to the local population. ⁵

Research on childhood leukaemia and lymphoma incidence after the Chernobyl fallout in Western Europe didn't reveal a statistically significant increase in each outcome compared with the background incidence rates. ⁶ , ⁷ , ⁸ , ⁹ However, a population‐based case–control study of childhood leukaemia cases diagnosed between April 26, 1986 and December 31, 2000 in Belarus, Russia and Ukraine reported a positive association between the leukaemia risk and accumulated absorbed bone marrow dose (median dose <10 mGy), most evident in Ukraine, apparent (although not statistically significant) in Belarus and not found in Russia. ¹⁰ A case–control study on acute leukaemia risk in Ukrainian residents of the most contaminated territories exposed to the fallout at age 0–20 years established a significant positive association between bone marrow dose (mean dose of 4.5 mSv, maximum dose of 101 mSv) and acute lymphoblastic leukaemia in the period of 1993–1997 as well as acute myeloid leukaemia in the period of 1987–1992. ¹¹ Moreover, elevated childhood leukaemia rates were found in the most contaminated areas of Ukraine compared with the uncontaminated territories. ¹² Comparison of observed versus expected lymphoma and leukaemia incident cases in the individuals who were less than 18 years at the time of the accident suggested some elevation in leukaemia risk both in Ukraine and Belarus, although not statistically significant, warranting further monitoring of leukaemia trends in the most contaminated areas. ¹³ , ¹⁴

Findings on post‐Chernobyl infant leukaemia are equivocal, while some studies reported the increased rates associated with in utero exposure from the fallout, ¹⁵ , ¹⁶ , ¹⁷ and other studies showed an increase that was not correlated with radiation contamination levels or found no increase at all. ¹⁸ , ¹⁹

Research on adult haematological malignancies after Chernobyl has suggested an increased risk of all leukaemia combined, non‐Hodgkin lymphoma and other malignancies of lymphoid and haematopoietic tissue in relation to bone marrow dose in a case–control study of Belarus, Russia and Baltic States liquidators. ²⁰ A significant dose–response for all leukaemia was found among the Ukrainian cleanup workers (liquidators) where both chronic lymphocytic leukaemia (CLL) and non‐CLL had positive but non‐significant dose response. Current information on the risk of haematological malignancies among the general population is limited.

Previously, we performed an ecological study on the incidence of breast cancer in Ukraine and Belarus, based on dose estimates averaged over raions that are primary unites of regional administrative division, suggesting no increased incidence between 1986 and 2016. ²¹ We are continuing to monitor cancer incidence trends in the affected population. Here, we report the results of an ecological study focusing on the incidence of haematological malignancies in the residents of radioactively contaminated areas in Belarus.

2. MATERIALS AND METHODS

2.1. Cancer incidence data

Numbers of lymphoma, multiple myeloma and leukaemia cases were received from the Belarusian national population‐based cancer registry. The national cancer registry was established in 1973 and computerized in 1979. Personal identifiers started to be recorded after 1984. The earliest data on haematological malignancy cases were available from 1978. Morphological verification was available for 97% and 98% of leukaemia and lymphoma cases in 1993–2007, increasing to 100% in 2013–2017. ²² , ²³ A detailed description of the cancer registration system in Belarus is published elsewhere. ²³

Diagnoses were coded using the International Classification of Diseases, 10th Revision (ICD‐10). ²⁴ We received case counts for each of the following outcomes: (1) Lymphoma: Hodgkin lymphoma (C81), follicular lymphoma (C82), non‐follicular lymphoma (C83), mature T/NK‐cell lymphoma (C84), other specified and unspecified types of non‐Hodgkin lymphoma (C85); (2) Myeloma: malignant immunoproliferative diseases and certain other B‐cell lymphomas (C88) and multiple myeloma and malignant plasma cell neoplasms (C90), reported as a single outcome; (3) Leukaemia: lymphoid leukaemia (C91), myeloid leukaemia (C92), monocytic leukaemia (C93), other leukaemia of specified cell type (C94), leukaemia of unspecified cell type (C95).

Cases were assigned to raions or cities based on their residence at the time of cancer diagnosis. We included Gomel and Mogilev oblasts in our study as these were the areas most heavily contaminated by Chernobyl fallout. The study oblasts consist of 42 raions and 3 major cities, namely Gomel city, Mogilev city and Bobruisk city as separate administrative units (Figure 1). The follow‐up for haematological malignancy incidence was through December 31, 2018.

FIGURE 1.

Map of the study raions based on raion‐ and age‐specific cumulative absorbed dose to the red bone marrow (RBM) in 2018 in the most contaminated oblasts of Belarus after Chernobyl fallout. The study catchment area is coloured. Raions with average 2‐year lagged cumulative absorbed RBM dose in 2018 of ≥20 mGy are the following: 1—Cherikov, 2—Slavgorod, 3—Krasnopolie, 4—Korma, 5—Chechersk, 6—Vetka, 7—Bragin, 8—Khoiniki, 9—Narovlya, 10—Elsk, 11—Lelchitsy. Cumulative absorbed RBM doses in 2018 were estimated separately for three cities, namely Gomel city, Mogilev city and Bobruisk city.

2.2. Population data

Population size data were obtained from the Belarussian state department of statistics. We used the size of population tabulated by sex, raion, calendar year, urban/rural status and attained age category as a corresponding population denominator. While raions consist of urban and rural territories depending on their industrial activity, cities are designated as urban territories. We used official administrative designation of urban/rural territories.

Annual population size data for each study raion were reconstructed for the period of 1978–2018 by linear interpolation of the population size in years between the censuses in 1979, 1989, 1999 and 2009 and taking into account major relocations in 1986 and 1990–1999. For the years between 2009 and 2016, population size was linearly extrapolated, while for the last 2 years of follow‐up we used population size from the year of 2016.

2.3. Dose estimation

Raion‐specific age‐specific annual absorbed doses to the red bone marrow (RBM) were estimated for the period of 1986–2018. Average age‐specific annual absorbed RBM doses were estimated separately for the residents of three cities in Belarus (Gomel city, Mogilev city and Bobruisk city) for the same period.

Detailed methods of dose estimation are described elsewhere. ⁵ In brief, three exposure pathways were considered: (i) external exposure to radionuclides deposited on the ground; (ii) internal exposure from ingestion of ¹³⁴Cs and ¹³⁷Cs radionuclides and (iii) internal exposure from ingestion of ⁹⁰Sr radionuclide. Doses were reconstructed using information on the gamma‐spectrometric measurements of radionuclide activity concentrations in soil samples (mainly ⁹⁵Zr, ⁹⁵Nb, ¹⁰³Ru,¹⁰⁶Ru, ¹³⁴Cs, ¹³⁷Cs, ¹⁴¹Ce, ¹⁴⁴Ce); ¹³⁷Cs ground deposition density; ⁹⁰Sr ground deposition density; ¹³⁴Cs, ¹³⁷Cs and ⁹⁰Sr activity concentrations derived from the measurements of total beta‐activity in cow's milk samples in 1986; ¹³⁷Cs activity concentration measured in cow's milk samples, potato and other foodstuffs; ⁹⁰Sr activity concentration measured in cow's milk and potato samples; radiocaesium whole‐body counter measurements.

Annual absorbed RBM doses were calculated for each settlement of the raion and then weighted by total population in each settlement in each year from 1986 to 2018 to obtain annual raion‐average absorbed doses. We calculated a raion‐specific age‐specific cumulative absorbed RBM dose for each specific follow‐up year starting from 1986 through 2018.

2.4. Statistical analysis

For each haematological malignancy (HM) outcome, we tabulated numbers of cases and population size over calendar year, sex, oblast, raion, urban/rural status and attained age categories from 0 to 80 years with 5‐year increment. Population size was used as an approximation of the person‐years at risk (PYR). For the purpose of descriptive and regression analysis we used the following outcomes: Hodgkin lymphoma (HL, ICD‐10 code: C81), non‐Hodgkin lymphoma (NHL, ICD‐10 codes: C82–C85), multiple myeloma and malignant immunoproliferative diseases reported as a single outcome by the registry (MM, ICD‐10 codes: C88 and C90), lymphoid leukaemia (LL, ICD‐10 code: C91), myeloid leukaemia (ML, ICD‐10 code: C92) and total leukaemia (ICD‐10 codes: C91–C95).

We calculated crude incidence rates using HM cases grouped as described above as the numerator and population size as the denominator by sex, attained age, urbanicity status and 10‐year calendar periods separately for Gomel and Mogilev oblasts and for both oblasts combined. We restricted our analysis to case counts and respective person‐years in those born before 1986 to avoid mixing the effects of prenatal and postnatal exposures.

We applied a 2‐year lag to raion‐average cumulative absorbed RBM doses allowing for a minimum latency period between exposure to ionizing radiation and HM occurrence. ²⁵

Baseline incidence rates for the population with raion‐average 2‐year lagged cumulative RBM dose below 5 mGy over the period from 1978 through 2018 were fitted using negative binomial regression models adjusted for effects of sex, attained age, calendar year and urban/rural status. The estimate for the calendar year parameter was used to calculate an annual percent change (APC) in HM incidence rates.

We investigated the association between raion‐average 2‐year lagged cumulative RBM dose and HM incidence using negative binomial regression models, which represent a generalization of Poisson regression models and are used if there is an overdispersion of count data. Dose–response analysis was conducted using raion‐average 2‐year lagged cumulative absorbed RBM dose both as categorical (<5, 5 to 19 and ≥20 mGy) and continuous variable for the follow‐up period of 1986–2018 and by 10‐year calendar time intervals (1986–1995, 1996–2005 and 2006–2018). For the 10‐year intervals, regression models were adjusted for the effects of sex, attained age (0–19, 20–39, 40–59, 60–79 years old) and urbanicity status (urban, rural). The model that included the follow‐up period from 1986 to 2018 was adjusted for the effect of calendar year in addition to the effects of sex, attained age and urbanicity status.

We estimated the incidence rate ratio (IRR) as an approximation of relative risk (RR) of haematological malignancies incidence. We applied the following regression equation in the model that included the follow‐up period from 1986 to 2018:

$${\mathit{IR}}_j=e^{{\beta }_0+{\beta }_1{x}_{1\_2,j}+{\beta }_2{x}_{1\_3,j}+{\beta }_3{x}_{2\_2,j}+{\beta }_4{x}_{2\_3,j}+{\beta }_5{x}_{2\_4,j}+{\beta }_6{x}_{3\_2,j}+{\beta }_7{x}_{4\_2,j}+{\beta }_8{x}_5}$$

where IR _j is the incidence rate for the jth observation, x ₁ is 2‐year lagged raion‐average cumulative bone marrow dose (as categorical variable), x ₂ is attained age (as categorical variable), x ₃ is sex (as categorical variable), x ₄ is urbanicity status (as categorical variable), x ₅ is calendar year (as continuous variable) and ${\beta }_i$ (i = 0,8) represents estimated coefficients. To test for the trend over radiation dose categories, categories of 2‐year lagged cumulative RBM dose were denoted as integer numbers from 1 to 3 with degrees of freedom (df) equal to 1.

We tested an association between RBM dose and HM outcomes as the main effect. Then, all the above‐listed variables were sequentially tested for potential confounding effects by adding them into the initial model. The variables remained in the model if the regression coefficient for the dose changed by 10% or more, or if the variable inclusion statistically significantly improved the model fit.

We studied an effect modification of sex, attained age, age at the time of the accident and time since the accident on dose‐related HM risks by fitting the dose‐including model adjusted for the effects of sex, attained age, urbanicity status and calendar year, and then comparing the deviance of the fitted model with the deviance of the model including an interaction term between the dose and one of the above mentioned factors using likelihood ratio (LR) test. A significant p‐value at n – 1 df, where n is a number of categories of potential effect modification factor, indicates that the effect of dose is not homogeneous across the categories of the studied factor. The models with interaction term between the dose and age at the time of accident or the dose and time since the accident were adjusted for age at the time of the accident or time since the accident instead of attained age.

We conducted the following sensitivity analyses: (i) excluding the population of Gomel and Mogilev cities from the RR analysis for all HM outcomes to test for the potential impact of migration that could result in dose misclassification; (ii) excluding the population older than 60 years from the RR analysis of lymphocytic leukaemia (LL) to avoid inclusion of chronic LL cases, rarely observed at age below 60 years; (iii) excluding the population younger than 60 years from the RR analysis of lymphocytic leukaemia (LL) to limit inclusion of acute LL cases.

We used the statistical software STATA 17 ²⁶ for model parameter estimation. Goodness of fit of the study models was based on the LR test. We used z‐test statistics as STATA's standard output for coefficients estimated based on the negative binomial regression model. All statistical tests were two‐sided and we considered p < .05 significant.

3. RESULTS

A total of 20,048 HM cases, including 7907 lymphomas (2952 HL and 4955 NHL), 10,138 leukaemias and 2003 multiple myelomas, were reported in the study population of 108.7 million PYR with attained age less than 80 years in the period of 1978–2018. Twelve percent of HM patients were younger than 20 years at the time of the accident. We excluded from further analyses 1241 HM cases together with the respective 17.9 million PYR of people who were born during or after 1986 due to potential in utero exposure. After these exclusions, we analysed 18,807 HM cases, among which 7328 lymphoma, 9476 leukaemia and 2003 multiple myeloma cases in the population of 90.8 million PYR. Table 1 presents numbers of HM cases, PYR and crude incidence rates for Gomel and Mogilev oblasts separately and for the two oblasts combined, for those who were born before the accident and were younger than 80 years old at the time of diagnosis. Detailed description of the crude incidence rates by sex, attained age, urbanicity status and RBM dose categories is presented in Tables S1–S5.

TABLE 1.

Characteristics of the study population: Haematological malignancy crude incidence rates (cases per 100,000 PYR) in the people born before the accident and younger than 80 years old in Gomel and Mogilev oblasts, followed up during the period of 1978–2018.

Outcome	Gomel			Mogilev			Total
	No. of cases	PYR/105	Crude rate/105	No. of cases	PYR/105	Crude rate/105	No. of cases	PYR/105	Crude rate/105
Lymphoma
Hodgkin (C81)	1430	511.7	2.8	1135	396.6	2.9	2565	908.4	2.8
Non‐Hodgkin (C82–C85)	2742	511.7	5.4	2021	396.6	5.1	4763	908.4	5.2
Multiple myeloma (C88, C90)	1150	511.7	2.3	853	396.6	2.2	2003	908.4	2.2
Leukaemia
Lymphoid (C91)	2515	511.7	4.9	2461	396.6	6.2	4976	908.4	5.5
Myeloid (C92)	1776	511.7	3.5	1231	396.6	3.1	3007	908.4	3.3
Total leukaemia (C91–C95)	5004	511.7	9.8	4472	396.6	11.3	9476	908.4	10.4

Table 2 presents cumulative raion‐specific RBM doses from all exposure pathways for the residents born before the accident. The mean (median) age‐averaged raion‐specific RBM dose accumulated by December 31, 2018 was 14.2 (6.4) mGy ranging from 0.7 to 60 mGy, with the higher levels in Gomel compared with Mogilev oblast. About 65%–75% of the total RBM dose accumulated by the end of 2018 was received in the first 10 years after the accident. Exactly 17 out of 42 study raions together with Gomel city, Mogilev city and Bobruisk city had cumulative age‐averaged raion‐specific RBM doses <5 mGy (n = 14,088 HM incident cases), while 11 raions were in the range from 20 to 60 mGy in 2018 (n = 1119 cases) (Figure 1).

TABLE 2.

Cumulative raion‐specific age‐specific red bone marrow doses from all exposure pathways in the residents born before the accident (mGy).

	Min	Median	Mean	Max
Total a	0.7	6.4	14.2	60.0
ATA, years b
<9	0.5	5.1	11.5	50.9
10–19	0.6	6.0	13.4	58.2
20+	0.8	6.6	14.7	61.6
Oblast c
Gomel	1.8	12.3	19.9	60.0
Mogilev	0.7	2.8	8.5	36.7
Calendar year d
1986	0.5	2.8	5.0	22.7
1996	0.6	5.0	10.7	45.6
2006	0.7	5.7	12.6	53.5
2018	0.7	6.4	14.2	60.0

^a Age‐averaged raion‐specific doses accumulated by December 31, 2018.

^b Age at the time of the accident (ATA)‐averaged raion‐specific doses accumulated by December 31, 2018.

^c Age‐averaged raion‐specific doses accumulated by December 31, 2018.

^d Age‐averaged raion‐specific doses accumulated between April 26, 1986 (the date of the accident) and December 31, 1986, 1996, 2006 and 2018.

3.1. HM baseline incidence rates

Annual Hodgkin lymphoma (HL) baseline (i.e., observed in the absence of ionising radiation exposure) incidence rates did not statistically significantly increase (p for trend = .15) in the period from 1978 to 2018. There was no statistically significant difference between urban and rural population (p = .29). Females had 12% lower HL rates than males (p = .004). Those who were 0–19 years old had the lowest HL rates followed by age groups 40–59, 60–79, with the highest incidence rate registered in the people aged 20–39 years.

Annual increase of non‐HL (NHL) baseline incidence rates was 3% (p for trend <.001). The population living in rural areas of Gomel and Mogilev oblasts had lower NHL rates compared with urban residents (p < .001). The rates were lower in females than in males (p < .001), and they increased with attained age (p for trend <.001).

Multiple myeloma (MM) incidence baseline rates increased by 3% per follow‐up year (p for time trend <.001). The rural population had lower MM incidence rates compared with residents of urban areas (p < .001). There was no difference in MM incidence rates between men and women (p = .1). The rates were increasing with attained age (p for trend < .001).

Baseline incidence rates of lymphoid (LL), myeloid (ML) and total leukaemia were increasing by 1%, 2% and 1% per year, respectively (p for time trend <.001 in each case) during the period of 1978–2018. Leukaemia rates were lower in rural compared with the urban population (p < .001), and in females compared with males (p < .001). LL and total leukaemia baseline rates were the lowest in the age group of 30–39 years, followed by the rates in the age group of 0–19 years. The LL and total leukaemia incidence rates started to significantly increase after the age of 40 years. ML incidence rates were increasing with attained age (p for trend <.001).

Detailed description of HM baseline incidence rates is presented in Table S6.

3.2. Doses‐response analysis

After adjustment for sex, attained age, urban/rural status and calendar period effects, there was no evidence of dose‐related increase in HL, NHL, LL, ML, total leukaemia or MM neither during the total follow‐up period from 1986 through 2018 (Table 3) nor by follow‐up decades (Tables S7 and S8). For NHL and MM, we observed a decrease in RR estimates with raion‐average 2‐year lagged cumulative RBM dose used as a categorical or continuous variable (Table 3). There was no clear trend for LL, ML and total leukaemia with the RBM dose used as a categorical variable (Table 3). In the model including RBM dose as a continuous variable, ML incidence risk was positively associated with the dose, with RR/ 100 mGy estimate equal to 1.29 (95% CI: 0.81, 2.05) (Table 3).

TABLE 3.

Hodgkin lymphoma (HL, ICD‐10 code: C81), Non‐Hodgkin lymphoma (NHL, ICD‐10 code: C82–C85), Multiple Myeloma (MM, ICD‐10 code: C88, C90), Lymphoid leukaemia (LL, ICD‐10 code: C91), myeloid leukaemia (ML, ICD‐10 code: C92) and total leukaemia (ICD‐10 code: C91–C95) relative risk (RR) estimates in relation to the 2‐year‐lagged cumulative bone marrow dose as categorical and continuous variable in the Belarusian study population born before the accident with attained age less than 80 years old and follow‐up period from 1986 to 2018.

	HL			NHL			MM
	No. of cases	RR a	95% CI	No. of cases	RR a	95% CI	No. of cases	RR a	95% CI
2‐year‐lagged cumulative bone marrow dose, mGy
<5	1557	1.00		3099	1.00		1336	1.00
5–19	371	1.03	0.91, 1.16	925	0.99	0.92, 1.08	369	0.87	0.77, 0.99
20+	94	0.83	0.67, 1.03	241	0.85	0.74, 0.97	119	0.93	0.77, 1.12
P for trend b			.32			.06			.09
Dose as cont. variable c	2022	0.63	0.32, 1.22	4265	0.60	0.40, 0.90	1824	0.83	0.46, 1.49
Dose as cont. variable c , d	1332	0.69	0.35, 1.34	2795	0.69	0.45, 1.04	1106	1.28	0.70, 2.31

	LL			ML			Total leukaemia
	No. of cases	RR a	95% CI	No. of cases	RR a	95% CI	No. of cases	RR a	95% CI
2‐year‐lagged cumulative bone marrow dose, mGy
<5	3145	1.00		1901	1.00		5915	1.00
5–19	897	0.93	0.85, 1.00	603	1.01	0.92, 1.12	1699	0.94	0.88, 1.00
20+	291	0.97	0.95, 1.09	190	1.03	0.89, 1.21	559	0.99	0.91, 1.09
P for trend b			.17			.64			.25
Dose as cont. variable c	4333	0.90	0.61, 1.33	2694	1.29	0.81, 2.05	8173	0.98	0.74, 1.30
Dose as cont. variable c , d	2899	1.19	0.81, 1.76	1798	1.56	0.97, 2.51	5405	1.27	0.96, 1.69

^a Adjusted for the effects of attained age, sex and urban/rural status and additionally adjusted for calendar year effect in the analysis of the total follow‐up period from 1986 to 2018.

^b The p value for trend across 2‐year‐lagged cumulative absorbed bone marrow dose categories (df = 1).

^c Model included 2‐year‐lagged cumulative absorbed bone marrow dose as a continuous variable. Risk is expressed per 100 mGy.

^d Population of Gomel and Mogilev cities was excluded from the analysis. Model included 2‐year‐lagged cumulative absorbed bone marrow dose as a continuous variable. Risk is expressed per 100 mGy.

Because people from the most contaminated territories were mainly resettled to large urban areas with low contamination levels, we performed a sensitivity analysis by excluding the population from Gomel and Mogilev cities. The sensitivity analysis using HL and NHL cases resulted in some elevation in RR estimates for each outcome, with a 9% increase in the HL risk estimate and a 13% increase in the NHL radiation risk estimate (Table 3). A negative trend of multiple myeloma RRs with RBM dose categories in the second decade after the accident was less pronounced when Gomel and Mogilev cities are excluded (p for trend = .2) (results not shown), and there was a positive although non‐significant association with the RBM dose used as a continuous variable (RR/100 mGy = 1.28, 95%: 0.70, 2.31).

After exclusion of Gomel and Mogilev cities from the analysis, positive but statistically non‐significant risks associated with RBM dose for lymphoid leukaemia (RR/100 mGy = 1.19) and total leukaemia (RR/100 mGy = 1.27) were observed (Table 3). Most noticeable changes in risk estimates in the sensitivity analysis were for myeloid leukaemia with RR/100 mGy = 1.56 (95% CI: 0.97, 2.51) using RBM dose as a continuous variable as compared with RR/100 mGy = 1.29 (95% CI: 0.81, 2.05) in the full data analysis (Table 3).

Using aggregated data, we could not delineate between acute lymphoblastic leukaemia (ICD‐10 code: C91.0) and chronic lymphocytic leukaemia (CLL) (ICD‐10 code: C91.1) cases. To test for association between dose and non‐CLL and CLL incidence, two types of sensitivity analysis were performed. Because CLL is usually rare under the age of 60 years, meaning lymphoid leukaemia cases in the population younger than 60 years old are likely to be almost entirely a non‐CLL, we run the analysis excluding people older than 60 years. An estimate of RR per 100 mGy changed from 0.90 (95% CI: 0.61, 1.33; n = 4333 cases) for all lymphoid leukaemia to 0.70 (95% CI: 0.33, 1.45; n = 1577 cases) for presumably non‐CLL cases. Fitting the regression model in the population of 60–79 years old, the RR per 100 mGy was 0.99 (95% CI: 0.63, 1.56) based on 2756 cases presumably with high proportion of CLL.

3.3. Dose response effect modification

Table 4 shows radiation‐associated RR estimates by the categories of selected effect modifying factors. RR/100 mGy was higher in men compared with women for HL (p‐value for homogeneity = .04), myeloid leukaemia (p‐value = .02) and all leukaemia combined (p‐value = .03). For myeloid leukaemia, RR/100 mGy was 2.09 (95% CI: 1.13, 3.84) in men compared with 0.75 (95% CI: 0.38, 1.49) in women. For non‐Hodgkin lymphoma, RR/100 mGy estimates significantly vary across the categories of attained age, with the lowest risk estimate observed at attained age of 0–19 years, although based only on 108 cases (p‐value = .02). For myeloid leukaemia, the highest but rather uncertain radiation‐related risk estimate was observed in the attained age group of 0–19 years based on 63 cases. The risk estimates decreased with attained age, with no evidence of statistically significant heterogeneity between the estimates (p‐value = .14). Radiation‐related risk estimates for myeloid leukaemia and all leukaemia combined vary with age at the time of the accident (ATA), with the estimates to be the highest in those who were less than 10 years ATA for myeloid leukaemia, and younger than 20 years ATA for all leukaemia combined but not statistically significant (p‐value > .5 in each case). Radiation‐related risk estimates of all study outcomes do not significantly vary with time since the accident except for non‐Hodgkin lymphoma (p‐value for heterogeneity = .01), and suggestively for multiple myeloma (p‐value for heterogeneity = .08). The highest RR/100 mGy of 1.27 (95% CI: 0.64, 2.51) was observed 10–20 years since the accident for non‐Hodgkin lymphoma. For multiple myeloma, the highest RR/100 mGy of 1.39 (95% CI: 0.70, 2.77) was observed 20 or more years after the accident (Table 4).

TABLE 4.

Effect modification role of sex, attained age, age at the time of the accident (ATA) and time since the accident on the radiation‐related risk of Hodgkin lymphoma, non‐Hodgkin lymphoma, multiple myeloma, lymphoid and myeloid leukaemia and all leukaemia combined.

Factors	Hodgkin lymphoma		Non‐Hodgkin lymphoma
	No. of cases	RR/100 mGy (95% CI)	No. of cases	RR/100 mGy (95% CI)
Sex a
Males	992	1.17 (0.49, 2.78)	2157	0.65 (0.37, 1.13)
Females	1030	0.31 (0.11, 0.83)	2108	0.55 (0.31, 0.98)
p value b		.04		.68
Attained age a , years
<20	202	1.11 (0.05, 25.20)	108	0.001 (<10−4, 0.62)
20 to <40	937	0.29 (0.10, 0.89)	460	0.14 (0.03, 0.76)
40 to <60	483	1.54 (0.52, 4.60)	1457	0.97 (0.50, 1.87)
60 to <80	400	0.59 (0.17, 1.99)	2240	0.65 (0.33, 0.95)
p value b		.21		.02
ATA, years c
<10	423	0.82 (0.15, 4.57)	219	0.11 (0.01, 1.66)
10 to <20	484	0.47 (0.13, 1.79)	317	0.69 (0.16, 2.93)
≥20	1115	0.77 (0.34, 1.74)	3729	0.63 (0.41, 0.97)
p value b		.80		.42
Time since the accident d , years
<10	735	0.37 (0.09, 1.51)	1017	0.17 (0.05, 0.53)
10 to <20	683	0.72 (0.25, 2.04)	1222	1.27 (0.64, 2.51)
≥20	604	0.62 (0.22, 1.73)	2026	0.59 (0.34, 1.01)
p value b		.76		.01

	Multiple myeloma		Myeloid leukaemia
Sex
Males	787	0.82 (0.34, 1.95)	1336	2.09 (1.13, 3.84)
Females	1037	0.84 (0.39, 1.81)	1358	0.75 (0.38, 1.49)
p value b		.96		.02
Attained age, years
<20	0	—	63	129.74 (3.19, >1000)
20 to <40	37	18.07 (0.63, 518.67)	323	1.94 (0.44, 8.59)
40 to <60	614	0.69 (0.24, 2.00)	809	1.46 (0.64, 3.37)
60 to <80	1173	0.82 (0.41, 1.65)	1499	1.06 (0.59, 1.91)
p value b		.26		.14
ATA, years c
<10	8	N.E.	131	3.23 (0.28, 37.38)
10 to <20	61	3.82 (0.30, 48.86)	189	2.02 (0.40, 10.25)
≥20	1755	0.82 (0.45, 1.50)	2374	1.22 (0.74, 2.00)
p value b		1		.65
Time since the accident d , years
<10	346	0.66 (0.12, 3.61)	667	2.17 (0.71, 6.61)
10 to <20	484	0.29 (0.08, 1.00)	883	0.83 (0.36, 1.88)
≥20	994	1.39 (0.70, 2.77)	1144	1.37 (0.73, 2.57)
p value b		.08		.37

	Lymphoid leukaemia		Total leukaemia
Sex
Males	2321	1.21 (0.73, 2.01)	4228	1.30 (0.90, 1.89)
Females	2012	0.63 (0.35, 1.12)	3945	0.71 (0.47, 1.07)
p value b		.08		.03
Attained age, years
<20	130	0.45 (0.01, 28.61)	213	10.68 (0.86, 132.75)
20 to <40	203	2.42 (0.38, 15.50)	610	2.58 (0.89, 7.45)
40 to <60	1244	0.57 (0.26, 1.24)	2428	0.88 (0.52, 1.49)
60 to <80	2756	1.02 (0.65, 1.60)	4922	0.91 (0.64, 1.28)
p value b		.41		.09
ATA, years c
<10	149	0.23 (0.01, 4.84)	314	1.50 (0.26, 8.54)
10 to <20	156	0.25 (0.02, 2.73)	397	1.55 (0.47, 5.12)
≥20	4028	0.98 (0.66, 1.47)	7462	0.94 (0.69, 1.27)
p value b		.32		.65
Time since the accident d , years
<10	1372	0.66 (0.27, 1.58)	2375	0.86 (4.45, 1.67)
10 to <20	1232	0.92 (0.47, 1.82)	2442	0.94 (0.58, 1.54)
≥20	1729	1.07 (0.63, 1.83)	3356	1.08 (0.73, 1.59)
p value b		.64		.82

^a Model included 2‐year lagged cumulative bone marrow dose as main effect after adjustment for the effects of sex, attained age, urbanicity status and calendar period, with interaction term between the dose and potential effect modifier.

^b The p‐value from the likelihood ratio test of homogeneity across categories (degrees of freedom (df) = N − 1 where N is number of categories in effect modification factor).

^c Model included 2‐year lagged cumulative bone marrow dose as main effect after adjustment for the effects of sex, age at the time of the accident, urbanicity status and calendar period, with interaction term between the dose and age at the time of the accident.

^d Model included 2‐year lagged cumulative bone marrow dose as main effect after adjustment for the effects of sex, attained age, urbanicity status and time since the accident, with interaction term between the dose and time since the accident.

Table 5 shows the results of further testing of potential age ATA modification effect on radiation‐related risk of myeloid leukaemia based on fitted model with three‐way interaction term between the cumulative RBM dose, sex and age ATA. RR/100 mGy estimates by categories of age ATA were highest in females <10 years ATA but there was no evidence of statistically significant difference in radiation risk between men and women across the age ATA categories (p‐value for homogeneity = .22).

TABLE 5.

Radiation‐related RR estimates of myeloid leukaemia (ML, ICD‐10 code: C92) by sex and age at the time of the accident (ATA).

Effect modifier	No. of cases	RR/100 mGy a
Males
ATA, years
1–9	74	1.43 (0.05, 44.18)
10–19	96	1.64 (0.20, 13.47)
20+	1166	1.99 (1.04, 3.80)
Females
ATA, years
1–9	57	6.46 (0.34, 122.35)
10–19	93	2.53 (0.31, 20.85)
20+	1208	0.69 (0.33, 1.43)
p value b		.22

^a Model included 2‐year lagged cumulative bone marrow dose as main effect after adjustment for the effects of sex, attained age, urbanicity status and calendar period, with three‐way interaction term between the dose, sex and age at the time of the accident.

^b The p‐value from the likelihood ratio test of homogeneity across categories (df = 5).

4. DISCUSSION

This is the first comprehensive follow‐up (1978–2018) of lymphoma, multiple myeloma and leukaemia incidence in relation to raion‐ and age‐specific cumulative absorbed red bone marrow dose in the population living in the regions of Belarus heavily contaminated by Chernobyl nuclear accident radioactive fallout.

We observed 1% annual increase in total leukaemia and 3% annual increase in non‐Hodgkin lymphoma and multiple myeloma baseline incident rates. All study outcomes, except Hodgkin lymphoma, had higher baseline incidence rates in men compared with women, and in urban compared with rural areas. Sex‐, age‐ and time‐related patterns observed in our study correspond to the global trends of haematological malignancy incidence. ²⁷ , ²⁸

Findings for the risk of haematological malignancies in relation to childhood exposure from the Chernobyl radioactive fallout are ambiguous. A large collaborative study of 23 European countries – ECLIS, included 23,756 leukaemia cases in children younger than 15 years diagnosed in the follow‐up period of 1987–1991. ⁶ Although a small increase in leukaemia incidence was reported, no association with the radiation exposure resulting from the nuclear accident (expressed as effective dose equivalents and cumulative excess dose in mSv) was found. However, the authors estimated the number of expected childhood leukaemia cases as a result of the first‐year exposure from the fallout using respective BEIR V risk estimates suggesting that only in Belarus the number of expected cases is big enough to distinguish from the background incidence rate (the incidence increased by 6% in relation to the average first‐year effective dose equivalent of 2 mSv). ⁶

There have been several reports on post‐Chernobyl childhood leukaemia incidence in Belarus with little evidence of an association between the childhood leukaemia incidence trends and levels of radioactive contamination. ¹⁷ , ²⁹ , ³⁰ , ³¹ In a case–control study, a significantly increased risk of childhood leukaemia (acute lymphoid, acute myeloid and unspecified cases of leukaemia diagnosed in 1986–2000) in Belarus, Ukraine and Russia combined was reported with an estimated excess relative risk [ERR] per Gy of 32.4 (95% CI: 8.78, 84.0). ¹⁰ Mean values of bone marrow dose in the Belarusian study participants were 12.8 mGy in leukaemia cases and 11.7 mGy in controls, with a country‐specific estimate to be positive but not statistically significant (ERR/Gy = 4.09, 95% CI: not estimable, 37.7). ¹⁰ Due to potential sampling‐derived bias in Ukrainian data and lack of statistically significant dose–response in Belarus and Russia, the study did not provide convincing evidence of an increased risk of childhood leukaemia as a result of the Chernobyl accident. Comparison of observed versus expected leukaemia incident cases in the individuals who were <18 years at the time of the accident suggested some elevation in leukaemia risk both in Ukraine and Belarus, although not statistically significant. ¹³ , ¹⁴ Findings of statistically significant associations between childhood leukaemia and Chernobyl radiation exposure in Ukraine and Germany were published by. ¹¹ , ¹² , ¹⁶ , ¹⁹ , ³²

Evidence on elevated risks of haematological malignancies, specifically leukaemia, following adulthood exposure from Chernobyl fallout, is largely obtained from case–control studies on clean‐up workers (liquidators) from Ukraine, Belarus, Russia and the Baltic States. ²⁰ , ³³ , ³⁴ , ³⁵ Median bone marrow doses in Belarusian clean‐up workers were in a low dose range, 6.83 mGy in cases and 5.49 mGy in controls. The estimate of ERR/mGy for all haematological malignancies combined was 0.64 (90% CI: −0.57, 4.44). ²⁰

Cumulative raion‐specific absorbed RBM doses reconstructed in our study were in the low dose range (<100 mGy) with the mean (median) raion‐specific age‐averaged bone marrow doses accumulated by 31/12/2018 of 14.2 (6.4) mGy and maximum of 60 mGy. We found little evidence of increased leukaemia risks in relation to bone marrow dose in the study. However, there was a suggestion of a positive association between bone marrow dose used as a continuous variable and leukaemia risk, more pronounced for myeloid leukaemia and after exclusion of Gomel and Mogilev cities (a sensitivity analysis to address possible dose misclassification) with RR/100 mGy = 1.56 (95% CI: 0.97, 2.51). In the sensitivity analysis we also observed a positive but non‐significant association with the bone marrow dose for lymphoid leukaemia (RR/100 mGy = 1.19) and all leukaemia combined (RR/100 mGy = 1.27). Estimates of leukaemia radiation‐related risk tended to be higher in males than in females, and in those who were younger at the time of the accident and at diagnosis.

There is strong evidence on radiation‐associated risk for acute lymphoblastic leukaemia (ALL) both protracted ³⁶ and acute ³⁷ exposures, both in adults ³⁶ and children, ³⁸ while association between CLL risk and ionizing radiation exposure remains poorly understood and warrants further research. ³⁹ , ⁴⁰ , ⁴¹ Although we were not able to separate acute from chronic lymphoid leukaemia cases due to lack of the information on leukaemia subtype, we performed sensitivity analyses to investigate the radiation‐related risks of lymphoid leukaemia in people below and above 60 years old assuming a larger proportion of CLL leukaemia in the age group of above 60 years. The RR/100 mGy estimates for lymphoid leukaemia in all ages, below and above 60 years old were not statistically significantly different with widely overlapping confidence intervals.

We found no evidence of increased radiation‐related risk of Hodgkin lymphoma, non‐Hodgkin lymphoma and multiple myeloma. Our findings on lymphoma are in line with the results from the cohorts of individuals from Belarus exposed to Chernobyl radioactive fallout under age of 18 years. Comparison of observed versus expected total lymphoma incident cases (ICD‐10 code: C81–C90) provided a standardized incidence ratio (SIR) of 0.66 (95% CI: 0.26, 1.33). ¹⁴ A record‐based case–control study in the United Kingdom ⁴² and a census‐based nationwide cohort study in Switzerland on natural background radiation and the incidence of childhood cancer found no evidence of increased risk of lymphoma in relation to cumulative bone marrow dose from gamma‐radiation. ⁴³ No evidence of radiation‐associated excess risk of Hodgkin lymphoma and multiple myeloma incidence was found in the atomic bomb survivors followed up from late 1950 through the end of 2001. ³⁷

In our study, there was a suggestion of different radiation‐related effects between males and females for Hodgkin lymphoma. In the atomic bomb survivors, there was a weak suggestion of positive dose–response for non‐Hodgkin lymphoma in males but not in females. ³⁷

The high quality of the population‐based cancer registry, validated methods of exposure assessment and a long follow‐up period represent the major strengths of our study. A follow‐up of 40 years, during which a total of 18,807 cases of lymphoma, multiple myeloma and leukaemia were diagnosed gave a 90% statistical power to detect a relative risk of 2.0. Cases of haematological malignancies were obtained from the Belarusian national population‐based cancer registry that applies international standards of cancer registration and reporting. ²² Both internal and external exposure pathways were considered for the estimation of raion‐average bone marrow absorbed doses. Numerous measurements of radioactivity in samples collected from the environment, foodstuffs and whole‐body counter (WBC) measurements were used as input data in the dose reconstruction model. Validation of the estimated doses was based on comparison with directly measured external doses using thermoluminescence dosimeters, internal doses derived from WBC measurements and doses from ⁹⁰Sr intake calculated using ⁹⁰Sr activity measured in ribs bone ash of adults in 11 raions in Gomel Oblast. A recent validation exercise showed a high degree of reliability. ⁵

The study has several limitations. Our dose–response analyses were based on raion‐specific age‐specific bone marrow dose estimates, where the same dose estimate is assigned to the population group of a certain age living in a particular raion and calendar year. The inability to account for individual residence history in dose reconstruction could result in potential dose misclassification. Because people could migrate from highly contaminated raions to less contaminated raions and cities seeking better life conditions or medical care, migrants from highly contaminated areas could be assigned low doses.

One limitation is the lack of information on individual environmental, lifestyle and occupational exposures to investigate the role of non‐radiation risk factors such as smoking or exposure to chemicals such as benzene. ⁴⁴ , ⁴⁵

In our study, we could only divide the leukaemia cases into lymphoid and myeloid leukaemia without further subdivision (acute, subacute, chronic) and could not investigate an association between acute forms of leukaemia and bone marrow exposure.

In summary, we found little evidence of increased risk of lymphoma, multiple myeloma, lymphoid and total leukaemia associated with two‐year lagged raion‐specific cumulative RBM dose in the study population born before the accident and whose attained age was below 80 years during the follow‐up period from 1986 to 2018. There was a suggestion of an elevated radiation‐related risk of myeloid leukaemia, more pronounced in males compared with females, and at a young age at the time of the accident (<10 years old) and at diagnoses (<20 years old). These findings should be interpreted with caution due to the limitations of ecological study design. Studies with individually reconstructed bone marrow doses are required for more precise assessment of the radiation‐related risk of leukaemia and other haematological malignancies in the affected populations of Belarus.

AUTHOR CONTRIBUTIONS

Ljubica Zupunski: Writing – original draft; conceptualization; formal analysis; data curation; visualization. Alesia Yaumenenka: Conceptualization; data curation; methodology; writing – review and editing. Ilya Veyalkin: Conceptualization; writing – review and editing; data curation; methodology. Victor Minenko: Conceptualization; methodology; data curation; writing – review and editing. Pavel Moiseyev: Methodology; data curation; investigation; writing – review and editing. Joachim Schüz: Conceptualization; writing – review and editing; funding acquisition; resources; project administration. Tatiana Kukhta: Data curation; methodology; validation; writing – review and editing. Sergey Trofimik: Methodology; validation; writing – review and editing; data curation. Richard Harbron: Writing – review and editing; methodology. Vladimir Drozdovitch: Conceptualization; methodology; writing – review and editing; data curation. Evgenia Ostroumova: Conceptualization; writing – original draft; methodology; validation; supervision.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest.

Supporting information

Data S1. Supporting Information.

Zupunski L, Yaumenenka A, Veyalkin I, et al. Lymphoma, multiple myeloma and leukaemia incidence in regions of Belarus most heavily contaminated by the Chernobyl accident. Int J Cancer. 2026;158(12):3161‐3172. doi: 10.1002/ijc.70346

DATA AVAILABILITY STATEMENT

Datasets generated and analysed in this study are available from the corresponding author upon request conditional on approval from Belarusian national population‐based cancer registries as data custodians.