Chapter 4
RADIATION RISKS OF LEUKEMIA AMONG
RUSSIAN EMERGENCY WORKERS,
19861997
Introduction
In 1986 after the Chernobyl accident the AllUnion Distributed Registry
was established in the former USSR. After the disintegration of the USSR the
Russian National Medical and Dosimetric Registry (RNMDR) was set up in 1992.
The Registry places particular importance on the followup of emergency workers
with the highest doses (mean dose 0.11 Gy).
Of the radiogenic malignant neoplasms leukemia is known to have the
maximum radiation risk and minimum latent period of 2 years. Therefore, the
excess of leukemia incidence rate above the spontaneous level can serve as the
first indicator of health effects for those exposed after the Chernobyl
accident.
The principal source of
information about the doseresponse for leukemia is data about the atomic bomb
survivors in Japan (Preston et al., 1994). These data have been primarily
obtained for the medium and high doses (>0.2 Gy) and acute exposure. Results
on small doses and dose rates, on the other hand, are limited, because such
studies require large size cohorts and long followup period. Of the studies on
small doses and dose rates, mention should be made of the joint research of
leukemia incidence among nuclear industry workers (Preston et al., 1994) in
several countries, in which statistically significant radiation risks of
leukemia were obtained for prolonged exposure and accumulated radiation doses
less than 0.1 Gy. In this context, the cohort of emergency workers may be of
value to gain new information about the doseresponse in the dose range of less
than 0.2 Gy.
An analysis of radiation risks of leukemia incidence in the Russian
cohort of emergency workers was first performed in 1996. Results of the
analysis were published in papers by Cardis et al., 1995 and Ivanov et al.,
1997a. Cardis used the cohort method, with the control group being the
population of Russia of corresponding sex and age. The analysis was based on
data for 48 cases of leukemias diagnosed in males from 1986 to 1993 among
males. A statistically significant excess of leukemia incidence among the
emergency workers above the control was found. Under the assumption that this
excess was due to radiation exposure, the risk of leukemia incidence due to radiation
were estimated. The excess relative risk at dose 1 Gy was 4.3 with 95%
Confidence Intervals (0.8, 7.8).
In
the study by Ivanov et al., (1997b) the casecontrol method was used. The study
did not reveal statistically significant risks, though a positive trend
in the dose dependence of relative risk was demonstrated.
This study is a continuation of
the research on leukemia incidence among emergency workers. In the period from
1993 to 1997 major changes have been observed in the distribution of leukemia
cases: new cases were added, both registered in this period and those diagnosed
earlier. Diagnoses were made more specific and diagnosis dates were determined
more accurately. The number of leukemias diagnosed to date permits an analysis
using the most reliable medical and dosimetric data. The practice of RNMDR
shows that the information supplied by regional centers of RNMDR is not of
equal quality. The present study uses medical and dosimetric data for males
from six economic regions of Russia (NorthWest, VolgoVyatsky, Povolzhsky,
CentralChernozem, NorthCaucasus and Urals). Data supplied from these regions
are characterized by reliability and high percentage of annual checkups for
emergency workers (about 86%).
Materials and methods
General description
of the study cohort
The overall size of the study
cohort (as of December 31, 1997) was 71,217 persons. These are emergency
workers who were examined at least once in the studied period, their doses
being confirmed by documents. The analysis is based on individual information
on date of birth, date of arrival to the controlled zone and date of departure,
date of the last checkup, date of diagnosis (in the case of disease) and dose
ascertained in documents.
The number of personyears from
1986 to 1997 is 743,845. Time under risk for each emergency worker was
determined as date of the last checkup (or diagnosis in the case of disease)
minus the date of arrival in the zone.
The loss of followup personyears does not exceed 14% (ratio of
observed number of person years and the maximum attainable is 86%, given 100%
attendance of annual checkups).
Figure 1 shows dynamics of the number of emergency workers in the
studied cohort. The curve rise in the initial period of followup is due to
features of arrival of emergency workers at the zone and the following decline,
mostly due to the decrease in the cohort size because of deaths. Rapid decrease
in the cohort size by the end of the followup period is explained by 23 years
delay in data accumulation and verification. For this reason, the considered
period of the followup was limited by 1997.
Figure 1. Dynamics of the number of emergency
workers in the studied cohort.
The studied cohort does not
differ by basic dosimetric and demographic characteristics from the general
emergency workers cohort. The RNMDR
includes a total of 174,916 emergency workers as of January 1, 1999). The
entire cohort is described in detail in Ivanov et al., (1997a).
Figure 2 shows the distribution
function of the number of emergency workers by age at exposure. The mean age at
the time of arrival in the Chernobyl zone was 34.7 years for healthy persons
and 32.4 for emergency workers with diagnosed diseases. It can be seen from
Figure 2 that those with diseases have a lower mean age at exposure than the
healthy ones.
A total of 44 cases of leukemia
were diagnosed (31 cases among emergency workers with known dose) in the
considered time period (19861997) in the indicated 6 rayons (districts). The
cases, which were not fully verified by the time of diagnosis, were not
considered. Four cases were diagnosed during the latent period of 2 years and
were excluded from the study. Thus, the analysis of the doseresponse included
27 cases. The dynamics of incidence are shown in Table 1.
The structure of incidence is given in Table 2. As can be seen from
Table 2, acute leukemias account for 39% of the total number of cases.
Figure 2. Distribution of emergency workers
by age at exposure.
Table 1
Dynamics
of verified leukemia cases in time
Calendar year 
1986 
1987 
1988 
1989 
1990 
1991 
1992 
1993 
1994 
1995 
1996 
1997 
Number of
cases 
0 
0 
3 
1 
4 
3 
3 
2 
6 
4 
3 
2 
Table 2
Structure
of leukemia incidence among emergency workers
Leukemia form
(ICD9) 

Absolute 
% 

All acute leukemias 
12 
39 
Acute lymphocytic leukemia (204.0) 
2 
7 
Acute myeloid leukemia (205.0) 
6 
19 
4 
13 

All chronic leukemias 
19 
61 
Chronic lymphocytic leukemia (204.1) 
8 
26 
Chronic myeloid leukemia (205.1) 
11 
35 
0 
0 

Total (204208) 
31 
100 
All the cases in the study were verified by the algorithm adopted in
RNMDR. The verification of leukemia diagnosis was carried out at 2 levels: the
place of residence and the RNMDR. Results of verification of diagnosis made at
the residence level were sent to RNMDR where a medical expert specializing in
hemoblastoses conducted final verification. The package to be provided to RNMDR
included primary medical documents such as extracts from medical records,
hematologist conclusion, autopsy record and, if necessary, diagnostic materials
(blood smear and bone marrow samples).
Dosimetric
characteristics of the cohort
The dosimetric data on emergency workers can be divided with respect to
reliability into three groups depending on the dose estimation method:
·
radiation or absorbed dose based on individual
dosimeter;
·
group dose assigned to members of a group based on
data of individual dosimeter borne by one member of the group;
·
route dose estimated by mean dose rate in the controlled
zone and time spent there.
Doses for emergency workers were
different and depended on time of working in the exposure area. The mean doses
were maximum for emergency workers involved in the works in 1986. Table 3
contains mean doses, time spent in the zone and dose rate derived by dividing
the individual dose by the time spent in the zone.
Table 3
Mean dose characteristics for emergency workers as a function
of time spent in the accident zone
Year 
Number 
Mean dose
(Gy)* 
Mean time
spent in zone (days)* 
Mean dose
rate (mGy/day)* 
1986 
26867 
0.17 
78.3 
4.3 
1987 
28845 
0.09 
82.8 
1.6 
1988 
11918 
0.03 
111.6 
0.5 
1989 
3736 
0.03 
107.3 
0.5 
1990 
451 
0.04 
109.8 
0.5 
19861990 
71817 
0.11 
87.4 
2.3 
* Averaging was carried out with the weight of the
number of followup personyears.
Figure 3 shows distribution of
emergency workers by external radiation dose. As can be seen from the figure,
the distributions are different from each other and the mean dose among the
cases is higher the dose among the healthy persons, which may be indicative of
the effect of exposure.
Distribution of emergency workers by time spent in the zone and dose
rate is illustrated in Figures 4 and 5. It can be seen that the dose rate does
not exceed 2 mSv/day for half of the emergency workers.
Main
characteristics of the emergency workers cohort are given in Table 4.
Figure 3. Distribution of emergency workers
by external radiation dose.
Figure 4. Distribution density of the number
of emergency workers by time spent in the zone.
Figure 5. Distribution of emergency workers
by dose rate.
Table 4
Main
characteristics of the emergency workers cohort
Status 
Leukemia
cases used for risk estimation (all types)* 
Leukemia
cases used for risk estimation 
Healthy 
Total 
27 
21 
71217 
Mean dose (Gy) 
0.135 
0.153 
0.108 
Mean dose rate (Gy/day) 
0.0037 
0.0042 
0.0023 
Mean time spent in the zone (days) 
77.1 
77.7 
87.4 
* Cases with a dose for the considered period
diagnosed after the 2 years latent period.
Statistical
methods
To estimate risk coefficients we used the maximum likelihood method in
the assumption that all the cases are independent Poisson values. The criteria
of significance were determined from the asymptotic properties of the
likelihood relations. The analysis was based on individual data on external
radiation dose, number of observation personyears and spontaneous incidence
rate.
The authors believe that it is preferable to use individual information
for estimating risks to minimize the impact of subjectivity and loss of
information in grouping and stratification of data.
The logarithm of likelihood function L
for the given sample is written as
_{} (1)
where
l is the parameter to determine
(here the leukemia incidence rate); t is the period of follow up of a
cohort member (in case of disease  time from start of followup to time of
diagnosis, for a healthy member of cohort  time of monitoring the cohort); n
is the number of cases for the observation period; N is the number of
healthy members of the cohort considered in the analysis in the observation
time.
The
confidence intervals for the estimated parameters were found using the
asymptotic properties of the likelihood function from equation (Handbook of
applicable mathematics, 1984):
_{} (2)
In estimation of risk
coefficients the linear and linearquadratic functions were used:
_{} (3)
and _{}. (4)
Here l_{i}^{0} is the
spontaneous leukemia incidence rate averaged over the followup period
corresponding to the age of the ith member of the cohort. The value
of the mean rate was determined from the following equation:
_{} (5)
where l_{i}^{R} is the age dependence of the spontaneous incidence
rate; u_{i} is age at exposure for the ith member of the
cohort. As spontaneous incidence rate we use the age dependence of the
incidence rate for Russia in general (Aksel and Dvoirin, 1992); d_{i}
is the external radiation dose for the ith member of the cohort; ERR_{1Gy}
is the excess relative risks per unit dose.
Since the values of external
radiation doses and dose rates are relatively low, we use a linear model to
estimate the effect of dose rate on incidence rate under the assumption that
the influence of these factors is additive:
_{}. (6)
Here dr is dose rate [Gy/day]; k is the coefficient accounting for
the difference of the spontaneous rate in the emergency workers cohort in the
studied period from the population of Russia of corresponding age. In the model
studied this coefficient is equal to the standardized incidence ratio (SIR) for
spontaneous rate among unexposed emergency workers. The difference of
coefficient k from unity can be due to the difference in how cases are
registered (for example, emergency workers are subject to regular annual
checkups) and quality of diagnosis verification. It was assumed in
calculations that this coefficient is the same in all age groups. Thus, to
determine the risk only relative age distribution of spontaneous incidence
rates is used.
The temporal trend of risk in the studied period is considered within
the linear model (6) with the only difference in that g and dr
have the meaning of change in risk per unit time and time after exposure,
respectively.
The ratio ERR_{1Gy}/a is the estimate of dose and dose rate factor (DDREF)
in the considered dose range (0~0.25 Gy).
A question arises about the reliability of data on age distribution of
spontaneous incidence rates. It is reasonable to assume that the distribution
of spontaneous incidence rates for leukemia by age is rather a conservative
value and is weakly dependent on geographic and ethnic attributes.
To verify this hypothesis and analyze the quality of Russian data within
the above hypothesis we use the age functions of the distributions of leukemia
incidence rates from the world’s largest cancer registries (Cancer Incidence in
Five Continents, 1987). The distribution function f_{j} was
calculated from the formula below:
_{}. (7)
u_{max}u_{min} is the considered age range; l^{0}_{k} is the
age distribution of the incidence rate in the registry selected as the control;
i
is the registry index.
Figure 6 shows the distribution
function of normalized incidence rates by age at diagnosis for registries of
USA (white), UK, Belarus and Russia (Cancer Incidence in Five Continents,
1987). The distribution is normalized by corresponding rates of the cancer
registry selected as control (data for Russia in general). The distribution
function was derived for males with the attained age of 2070 years.
Figure 6. Distribution function of leukemia
incidence (males) by attained age for different countries.
The results presented in Figure 6 confirm the assumption about the
similarity of relative distributions of age incidence rates for leukemia in
different countries and reliability of this characteristic in Russian data used
as the control.
In this work we compared the
radiation risks derived for the studied cohort of emergency workers with risks
of radiogenic leukemia of all types in the cohort LSS (Life Span Study) of
atomic bomb survivors in Japan (Preston et al., 1994). For comparison to be
correct (the LSS cohort has a different age structure, other spontaneous
incidence rates) the following approach was pursued. Preston et al., (1994)
gives the following approximations of the spontaneous incidence rates and
excess absolute risk for all leukemias and individual types for the LSS:
Leukemias of all types (ICD9: 204208) (designations are borrowed from
(Preston et al., 1994)).
The spontaneous incidence rate
_{},
where a is attained age; g is the age at exposure.
Excess absolute rate EAR (10^{4}PY)
_{}
where d is dose (Sv); t is time after exposure.
To allow for the difference in the spontaneous incidence rates we passed
from estimators of absolute risk to estimators of excess relative risk using
the relation ERR^{J}(d,g,t)=EAR^{J}(d,g,t)/l^{J}(g,g+t) (l is approximation of the spontaneous incidence rate of
leukemia in the LSS cohort).
Since leukemias belong to rare diseases, the number of cases (C)
can be described with a fairly good accuracy by:
_{}. (8)
Summation is done for all members
of the cohort in question using several parameters of model (3).
For the cohort in question C=27, the number of leukemias
considered in the analysis. The expected number of leukemias can be derived by
using the risk approximation ERR^{J}(d_{i},g_{i},i)
from the LSS cohort and data for each emergency worker from formula (8) with
substitution of excess relative risk ERR by:
_{} (9)
the mean value of risk over the followup interval (tl is the latent period).
Then the excess relative risk
averaged over the studied period per unit dose for the whole cohort is
determined according to (8) as follows:
_{}. (10)
In estimation of the number of radiogenic leukemias the latent period of
2 years was taken.
Results
Based on the experience of Japanese studies and research in other
countries suggesting no radiation dependence of chronic lymphocytic leukemia
(CLL) (Preston et al., 1994) the analysis was performed both with inclusion and
exclusion of such cases.
Results of estimating the model (3) parameters are presented in Table 5.
For leukemia of all types the values of radiation risk are not statistically
significant (P=0.04), given 95% confidence intervals. If CLL is excluded,
the risk becomes statistically significant and equals 11.7 (3.3, 20.1 95%CI)
per 1 Gy (P<0.001). The values of the coefficient k less than unity are explained by the fact that the leukemia cases
considered were only those verified at the time of analysis (at the time when
analysis was carried out the list of emergency workers with diseases also
included 11 persons for whom the leukemia diagnosis had to be ascertained).
Table 5
Estimated risk of induction of
radiogenic leukemia in the emergency workers cohort
(using data on spontaneous incidence rate in Russia)
Parameter 
All leukemias
(95% CI) 
All leukemia
excluding CLL (95% CI) 
All leukemia
(arrival in 19861987) (95% CI) 
ERR/Gy 
3.5 (0.5,
7.5) 
11.7 (3.3,
20.1) 
10.2 (2.8,
17.4) 
Coefficient k 
0.7 (0.4,
0.9) 
0.4 (0.2,
0.6) 
0.4 (0.2,
0.6) 
The risk of radiogenic leukemias
becomes statistically significant for all types of leukemias, if the emergency
workers exposed in 19881990 and receiving, on the average, lower dose than
those exposed in 19861987 are excluded.
Results of calculations by the
linearquadratic model (4) indicate that in the considered range of doses and
dose rates the influence of quadratic term is not significant (the coefficient
value with linear term is 9.7 (1.69, 17.67). The level of significance of the
linearity hypothesis is P=0.89. The value of the dose and
dose rate factor DDREF=11.7/9.7=1.2.
Parameters of model (6) are presented in Table 6. Introduction of the
dose rate factor has virtually not modified the value of the excess relative
risk, which suggests weak influence of this factor on risk of induction of
radiogenic cancers for the considered range of doses and dose rates. The significance
level of the hypothesis indicates that the risk coefficient is not dependent on
dose rate P=0.79. The value of the coefficient of risk sensitivity to
dose rate is statistically insignificant.
Parameter 
All leukemias
excluding CLL (95% CI) 
ERR/Gy 
10.7 (2.2,
19.1) 
Risk per Gy/day 
43.7 (213.4,
300.8) 
Using the same model we estimated the temporal trend of excess relative
risk in the considered time period. The calculation shows that the excess
relative risk decreases with time (the hypothesis significance level suggests
that the trend is absent P=0.02).
As can be seen from Figure 6, the
age distribution of incidence rate by data of different cancer registries are
similar and therefore risk was estimated using spontaneous incidence rate for
different countries. Results of calculations are shown in Table 7. It can be
seen that results of risk estimation are practically the same, no matter data
which registry data were used in computation. The spontaneous incidence rate
for the considered range of age and sex is similar.
Table 7
Estimated leukemia risks for emergency
workers using data
on spontaneous incidence rate from different cancer registries
Registry 
Russia 
USA 
Belarus 
UK 
ERR/Gy 
3.5 (0.5,
7.5) 
3.6 (0.5,
7.7) 
3.7 (0.4,
7.8) 
3.8 (0.4,
8.0) 
Coefficient k 
0.7 (0.4,
0.9) 
0.5 (0.3,
0.7) 
0.5 (0.3,
0.7) 
0.7 (0.4,
1.0) 
What is remarkable about the obtained result is that it demonstrates
that incidence data can in principle be used from different registries which
have an established system of data collection and verification and contain a
large body of cancer data. This can be useful for analysis of data for small
populations especially with rare diseases such as leukemia.
Using the approximants derived in the LSS cohort, the excess relative
risk per unit dose averaged over the considered period for all types of
leukemia was estimated. Likewise the dynamics of excess risk for the emergency
workers cohort with time was calculated (Figure 7). The excess relative risk
per 1 Gy is equal to 12.1 and differs from the value of 3.5 calculated with the
data set under study. The expected cumulative number of cases, when using the
Japanese risk coefficients, is C1=45.6 (the observed value is 27
cases). However, considering that there are virtually no CLL cases in the LSS
cohort (Preston et al., 1994), the value 12.1 is consistent with 11.7 derived
for the cohort of emergency workers with exclusion of this type of leukemia.
As follows from Figure 7, the excess relative risk decreases in the
considered period. Over the entire period, the risk has decreased by a factor
of 4.
Figure
7. Excess relative risk
in the cohort of emergency workers as a function of time
after the accident (the dynamics is estimated using estimators (Preston et al.,
1994)).
Discussion
The present work is a logical
continuation of the Russian study of leukemia incidence rate among emergency
workers (Ivanov et al., 1997a; Ivanov et al., 1997b). Obviously, this issue
remains topical, as emergency workers, on the average, received higher
radiation doses than the population of the affected regions and radiogenic
leukemia is the first indicator of longterm radiation effects. A study by
Ivanov et al., (1997a) has been conducted for the entire cohort of emergency
workers using an external control, a corresponding age category of the Russian
population. It is known that use of an external control can lead to a shift in
risk estimates due to differences in the systems of collection, registration
and verification of diagnoses in the study and reference groups. An estimation
of risk was made under a cautious assumption that the excess of the observed
incidence rate above the control is likely to be due to the radiation factor (Ivanov
et al., 1997a). Thus, the primary goal of Ivanov’s study was to identify the
current trends in leukemia incidence rate among the emergency workers and bring
the subject to the attention of researchers, rather than to estimate leukemia
risk.
In another study by Ivanov et
al., (1997b) the casecontrol approach has been taken to analyze risks. The
analysis has not revealed statistically significant risks, but the positive
trend in leukemia incidence rate as a function of an external radiation dose was
established. It cannot be ruled out that no significant risks were found
because application of the casecontrol methodology for the cohort of emergency
workers is rather problematic, since direct dose measurements were limited and
maximum permissible levels are limited.
The information about the cohort
of emergency workers used in the presented analysis is reliable. The data used
are the verified RNMDR data and the system of data collection and verification
have been tuned up (it has taken some time to fix the operation of a large
dynamic information system of RNMDR including more than 170 thousand persons).
The analysis uses, to the extent possible, a
priori information on leukemia incidence rate.
When all types of radiogenic leukemia are included the increase is not
statistically significant for the entire cohort of emergency workers, given the
confidence intervals of 95%. This is most probably because of the relatively
small number of cases that were studied (27) and a possible shift in the
estimate because of allowing for chronic lymphocytic leukemia, for which no
dose response has ever been established. Exclusion of CLL from the analysis
results in increase in the risk and the risk becomes statistically significant.
This result confirms the modern concept that CLLs are clearly not dose dependent.
The risk of radiogenic leukemia becomes statistically significant for
all types of leukemia with the exclusion of emergency workers exposed in
19881990 who received, on the average, lower doses than those exposed in
19861987.
The risk model that was studied using a priori information on spontaneous incidence rate allows use of
individual information and immediate estimation of two main characteristics,
the risk coefficient and standardized incidence ratio ¾ the difference
in the incidence rate in the studied cohort from the national level (or the
reference group). It has been shown that information of other cancer registries
having a large volume of data and established systems of collection and
verification of leukemia data such as registries of the USA or UK can be used
for risk analysis with good accuracy. The gained experience may be useful for
analysis of incidence rates of rare diseases such as leukemia in small
populations.
The data set used in the analysis
is adequately described by a linear dose function, the significance level of
the hypothesis is P=0.89 and as a consequence the value of the dose factor and
dose rate DDREF is close to unity.
In the considered range of doses (to 0.25 Gy) and dose rates (to 0.01
Gy/day) the effect of the dose rate factor on radiogenic cancer risk is not
statistically significant. The study was carried out within the linear model.
This approximation seems justified as the values of doses and dose rates
received by emergency workers are relatively low.
The risks of radiogenic leukemias
in the emergency workers’ cohort were compared with the risks for the same
cohort calculated using the estimators derived for the LSS cohort. The risk
estimators are a function of sex, age at exposure and time since exposure,
which makes correct comparison possible. The excess relative risk per unit dose
has been estimated to be 12.1. Since the number of CLL in the LSS cohort is
small (due to geographic and ethnic features of the cohort), the approximations
presented by Preston et al., (1994) for all types of leukemia virtually
correspond to the risk derived with exclusion of CLL. Then it may be more
adequate to compare the risk value of 12.1 with the risk value of 11.7 derived
in the regression analysis with exclusion of CLL.
Considering that the above values of the risk coefficients are similar
and the temporal trends are alike, it can be argued that the risk model
presented by Preston et al., (1994) adequately describes leukemia incidence
rate in the emergency workers’ cohort.
The derived risk value of 11.7 (excluding chronic lymphocytic leukemia)
suggests that at the mean dose of 0.1 Gy the attributive risk is about 50%,
i.e. half of the diagnosed cases are radiationinduced.
The results of the above analysis lead us to conclude that the dose
factor has an effect on leukemia incidence rate among emergency workers living
in Russia. Growth of the leukemia rate among emergency workers is probably the
first indication of distant consequences of the population exposure after the
Chernobyl accident.
Main conclusions:
1.
For all types of lymphocytic leukemia the radiation
risk coefficients for the time of monitoring the cohort of emergency workers
from 1986 to 1997 were not statistically significant. The value of the excess
relative risk is 3.5 (0.5, 7.5 95% CI).
2.
Exclusion of chronic lymphocytic leukemia increases
the confidence of radiation risk values and makes it statistically significant.
The value of ERR [Gy]^{1} is 11.7 (3.3, 20.1 95% CI).
3.
The risk of radiogenic leukemia is statistically
significant for all types of leukemia and equals 10.2 (2.8, 17.4 95% CI), if we
exclude emergency workers exposed in 19881990 who received lower dose than
those exposed in 19861987.
4.
The impact of the dose rate factor for the considered
range of doses and dose rates is not statistically significant.
5.
The attributive risk of inducing radiogenic leukemia
(excluding chronic lymphocytic leukemia) is 50%. This means that radiation may
be a factor in causing disease in half of the diagnosed cases of leukemia.
6.
The excess relative risk for all types of leukemia in
the studied period is 3.5. This is in agreement with the risk estimate of 3.4
calculated using the estimators derived for the cohort LSS.
References
Aksel E.,
Dvoirin V. Statistics of Malignant Neoplasms.  Moscow: VONTS
AMN SSSR, 1992 (in Russian).
Cancer
Incidence in Five Continents. D.M.Parkin et al. (Eds.), Vol. VI, IARC Scientific
Publication.  Lyon, 1987.
Cardis
E., Gilbert E.S., Carpenter L., Howe G., Kato I., Armstrong B.K., Beral V.,
Cowper G., Douglas A., Fix J., Fry S.A., Kaldor J., Lave C., Salmon L., Smith
P.G., Voelz G.L., Wiggs L.D. Effects of low doses and low doses
rates of external ionizing radiation: Cancer mortality among nuclear workers in
three countries//Radiat. Res.  1995.  V. 143.  P. 117132.
Handbook
of Applicable Mathematics. W.Ledermann (Ed.), Vol. VI “Statistics”, Part A. 
New York: John Wiley&Sons Ltd, 1984.
Ivanov
V.(a), Tsyb A., Gorsky A., Maksyutov M., Rastopchin E., Konogorov A., Korelo
A., Biryukov A. Matyash V. Leukemia and thyroid cancer in emergency workers of
the Chernobyl accident: estimation of radiation risks (19861995)//Radiat.
Environ. Biophys.  1997.  V. 36.  P. 916.
Ivanov
V.K.(b), Tsyb A.F., Konogorov A.P., Rastopchin E.M., Khait S.E. Casecontrol
analysis of leukaemia among Chernobyl accident emergency workers residing in
the Russian Federation, 19861993//J. Radiol. Prot.  1997.  V. 17. 
P.137157.
Preston
D., Kusumi S., Tomonaga M., Izumi S., Ron E., Kuramoto A., Kamada N., Dohy H.,
Matsuo T., Nonaka H., Thompson D., Soda M., Mabuchi K. Cancer
incidence in atomic bomb survivors. Part III: Leukemia and multiple myeloma,
19501987//Radiat. Res.  1994.  V. 137.  P. S94.