- Email: [email protected]
Association between type 2 diabetes and prenatal exposure to the Ukraine famine of 1932–33: a retrospective cohort study
Articles
Association between type 2 diabetes and prenatal exposure to the Ukraine famine of 1932–33: a retrospective cohort study L H Lumey, Mykola D Khalangot, Alexander M Vaiserman
Summary Background The effect of fetal and early childhood living conditions on adult health has long been debated, but empirical assessment in human beings remains a challenge. We used data from during the man-made Ukrainian famine of 1932–33 to examine the association between restricted nutrition in early gestation and type 2 diabetes in offspring in later life. Methods We included all patients with type 2 diabetes diagnosed at age 40 years or older in the Ukraine national diabetes register 2000–08, and used all individuals born between 1930 and 1938 from the 2001 Ukraine national census as the reference population. This study population includes individuals born before and after the famine period as controls, and those from regions that experienced extreme, severe, or no famine. We used prevalence odds ratios (ORs) as the measure of association between type 2 diabetes and early famine exposure, with stratification by region, date of birth, and sex for comparisons of diabetes prevalence in specific subgroups. Findings Using these two datasets, we compared the odds of type 2 diabetes by date and region of birth in 43 150 patients with diabetes and 1 421 024 individuals born between 1930 and 1938. With adjustment for season of birth, the OR for developing type 2 diabetes was 1·47 (95% CI 1·37–1·58) in individuals born in the first half of 1934 in regions with extreme famine, 1·26 (1·14–1·39) in individuals born in regions with severe famine, and there was no increase (OR 1·00, 0·91–1·09) in individuals born in regions with no famine, compared with births in other time periods. Multivariable analyses confirmed these results. The associations between type 2 diabetes and famine around the time of birth were similar in men and women. Interpretation These results show a dose–response relation between famine severity during prenatal development and odds of type 2 diabetes in later life. Our findings suggest that early gestation is a critical time window of development; therefore, further studies of biological mechanisms should include this period. Funding Ukraine State Diabetes Mellitus Program, US National Institutes of Health.
Introduction Poor living conditions around the time of birth have long been thought to have lasting effects on health.1 Nutrition during pregnancy has been proposed as an explanation for the occurrence of type 2 diabetes later in life,2, 3 but direct measures are hard to obtain. Birthweight is inversely associated with type 2 diabetes4 and has been used as a marker of fetal nutrition despite its limitations;5 however, birthweight is not affected by nutrition changes in early gestation.6 Fetal growth and type 2 diabetes could also have genetic links.7 Causal inference in the assessment of fetal antecedents of adult disease is therefore still a challenge, and reported associations could be biased rather than causal.8 Here we present a different approach to examine these questions that does not focus on birthweight alone. To better identify any effects of early nutrition, studies of man-made famines can offer distinct opportunities for the unbiased comparison of individuals born under such conditions with unexposed controls.9 These approaches can provide information about specific pregnancy exposures rather than size at birth alone, especially if the population at risk, the timing and degree of exposure, and relevant health outcomes can be accurately defined.
For type 2 diabetes in later life, studies of the Dutch famine of 1944–45,10,11 the Chinese famine of 1959–62,12 and three famines in 20th century Austria13 all suggest a relation with early life nutrition. By contrast, a study of the Siege of Leningrad of 1941–44 does not show an association.14 All studies have a particular strength: the timing of the famine in relation to the stage of gestation was particularly well defined in the Dutch study, as was the risk of adult hyperglycaemia in the Chinese study, and a large number of patients receiving antidiabetic drugs over a long time period was included in the nationwide Austrian study. However, no study combines all these strengths. To overcome the limitations of previous approaches, we used the setting of the Ukraine famine of 1932–33 to study the association between early-life nutrition and late-life type 2 diabetes in a large population, in cohorts well defined with respect to the timing of the famine in relation the stage of pregnancy and the severity of the famine around the time of birth. The Great Ukrainian Famine (also known as the Holodomor) was caused by the Soviet Union Government’s forced collectivisation of agriculture in the early 1930s. Although severe famines occurred in
www.thelancet.com/diabetes-endocrinology Published online September 3, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00279-X
Lancet Diabetes Endocrinol 2015 Published Online September 3, 2015 http://dx.doi.org/10.1016/ S2213-8587(15)00279-X See Online/Comments http://dx.doi.org/10.1016/ S2213-8587(15)00323-X Department of Epidemiology, Mailman School of Public Health, Columbia University, New York, NY, USA (L H Lumey MD); Molecular Epidemiology, Leiden University Medical Center, Leiden, Netherlands (L H Lumey); Komisarenko Institute of Endocrinology and Metabolism (M D Khalangot MD) and Chebotarev Institute of Gerontology (A M Vaiserman PhD), National Academy of Medical Sciences, Kiev, Ukraine; and Shupyk National Medical Academy of Postgraduate Education, Kiev, Ukraine (M D Khalangot) Correspondence to: Dr L H Lumey, Department of Epidemiology, Mailman School of Public Health, Columbia University, New York, NY 10032, USA [email protected]
1
Articles
Panel: Research In context Evidence before this study Studies of the Dutch famine of 1944–45, the Chinese famine of 1959–62, and three famines in 20th century Austria suggest that risk of type 2 diabetes might be affected by prenatal nutrition. Each of these studies has specific strengths, but unanswered questions remain because of limitations of sample size, uncertainty about the timing of the exposure, and short duration of follow-up. Added value of this study To overcome these limitations we collected evidence for the Ukraine famine of 1932–33. These data provide a large study population, well defined with respect to the timing of the famine in relation the stage of pregnancy. Regional analysis also provides estimates of the potentially differential effect of severe versus extreme prenatal famine exposure on type 2 diabetes in
Volyn
Rivne
Chernihiv
Kiev Khmelnytsky Vinnytsia
Kharkiv
Cherkasy
Kherson 0 0
100 miles 100 km Non-famine regions Severe famine regions Extreme famine regions
Luhansk
2
Implications of all the available evidence Our findings identify early gestation as a sensitive period and show a dose–response relation between the severity of famine and the odds of type 2 diabetes in later life. Our findings are consistent with those from epigenetic studies that point to effects of famine in early gestation on DNA methylation of genes related to metabolic disease. As a next step, specific epigenetic changes and morbidity of type 2 diabetes should be compared in populations with and without early famine exposure. Such comparison might establish whether prenatal and postnatal determinants of type 2 diabetes have specific epigenetic pathways in common.
the Holodomor as a crime against the Ukrainian people and against humanity.17 Estimates of population decline due to the famine range from 2·7 million to 3·9 million.18,19 The regional severity of the famine was determined by the prevailing political situation. In eastern Ukraine, under Soviet control, some regions suffered extreme population losses of more than 25% and others had severe losses of about 10–20%,18,20 with a ten-times mortality increase between April and July, 1933, compared with prefamine times.15 There were no population losses in western Ukraine, a region that was still under Polish rule during the famine years and not exposed to famine (figure 1).
Azov Sea
Methods
Black Sea Crimea
Figure 1: Regional geography of regions affected by the Ukrainian famine of 1932–33 Map of Ukraine showing the locations of the nine regions included in this study. Severely exposed regions are defined by a rate of population decline of up to 15% between 1929 and 1933, and extremely exposed regions by a rate of between 20% and 25%.18,20 See Online for appendix
later life. Our study combines the strengths of all existing studies.
several Soviet regions, including Kazakhstan, where no grain was produced, an especially strict food policy was applied to Ukraine.15 Ukraine had two successive bad harvests in 1931 and 1932. By the spring of 1932, there was an absolute shortage of grain, which became more severe in the ensuing 12 months. Peasants were also required to return not only grain they had earned by meeting production targets, but also other food such as potatoes and poultry. These events resulted in a catastrophic mortality increase between April and July, 1933.15 In 2007, the Parliament of Ukraine passed a law characterising the famine as a genocide of the Ukrainian people.16 In 2008, the European Parliament recognised
Data sources The reference population is derived from region-specific population counts from the 2001 Ukraine national census of all individuals born between 1930 and 1938, categorised by sex and month of birth. This information was collected through the census for all 25 Ukrainian regions, but is not in the public domain. After formal applications to the Ukraine State Statistics Committee, we obtained permissions from nine regions for the use of selected census data under strict privacy conditions. The regions represent all levels of famine exposure as defined by the magnitude of population losses. From eastern Ukraine, we included four regions with exposure to extreme famine in 1932–33 (Kharkiv, Kherson, Cherkasy, and Luhansk) and three with exposure to severe famine (Chernihiv, Vinnytsia, Khmelnytskyi). From western Ukraine, with no exposure to famine, we included two regions (Rivne and Volyn; figure 1). The reference population was stratified by region, date of birth, and sex for comparisons of type 2 diabetes prevalence in specific subgroups.
www.thelancet.com/diabetes-endocrinology Published online September 3, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00279-X
Table 1: Cases of type 2 diabetes and current census counts by year and season of birth (1930–38), and sex, by severity of Ukraine famine (1932–33) in the region of birth
Registered cases of type 2 diabetes were from the national Ukraine diabetes register 2000–08, and census counts were from the 2001 Ukraine national census. Regions with no famine were Rivne and Volyn; with severe famine were Chernihiv, Vinnytsia, and Khmelnytskyi; and with extreme famine were Kharkiv, Kherson, Cherkasy, and Luhansk.
242 078
6860 7932
224 496 40 615
1108 1282
38 713 40 022
1155 1382
41 093 34 125
999 1064
29 436 28 349
790 901
23 073 22 696
655 611
11 855 13 621
389 524
12 853 16 975
522 691
20 218 20 083
563 665
21 435 25 592
679 812
25 820 Census count
Type 2 diabetes cases
Extreme famine
3937
139 413 134 116
4234 626
20 369 19 793
709 648
20 062 21 499
728 526
18 335 14 362
580 480
15 059 12 103
421 298
11 955 7558
285 274
10 147 9859
285 347
13 005 15 557
384 288
13 096 14 674
397 450
17 385 18 711
445 Type 2 diabetes cases
Severe famine
Census count
57 910
2987 3547
60 223 6230
379 429
6678 6462
411 408
6761 6728
364 431
6933 6589
342 421
6629 6350
322 349
6201 6014
291 353
6040 6753
317 415
7155 5696
267 335
6307 Census count
294 406
7519
Type 2 diabetes cases
No famine
Women
www.thelancet.com/diabetes-endocrinology Published online September 3, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00279-X
7088
159 308
2843 3367
153 179 27 683
486 563
27 506 27 032
521 649
29 046 22 908
427 454
20 825 19 465
320 397
16 530 15 657
287 258
8922 9111
150 186
8821 10 488
197 258
12 711 12 168
204 273
13 462 14 796
251 329
15 356 Census count
Type 2 diabetes cases
Extreme famine
2084
88 223 87 968
2182 341
13 911 13 957
358 357
13 820 15 543
408 300
12 355 10 348
274 250
10 077 8528
228 173
8050 5017
162 108
6030 6219
140 179
7086 9041
225 160
7621 8810
158 216
9273 10 505
229 Type 2 diabetes cases
Census count
36 031 Severe famine
1730
38 079 4076
157 203
4469 4172
202 219
4569 4184
175 214
4602 4157
190 209
4214 3863
144 198
4023 3926
141 163
3811 3924
143 183
4075 3583
150 136
3939 Census count
145 205
4377
Type 2 diabetes cases
No famine
Jan–Jun Jan–Jun Jul–Dec
4146
Jul–Dec Jan–Jun Jul–Dec Jan–Jun Jul–Dec Jan–Jun Jan–Jun Jul–Dec Jan–Jun Jul–Dec Jan–Jun Jul–Dec Jan–Jun Jul–Dec Men
1931 1930
Jul–Dec
1932
1933
Statistical analysis We prepared contingency tables stratified by birth region, year and month of birth, and sex of individuals with type 2 diabetes diagnosed in 2000–08 among births between 1930 and 1938 in the selected study regions in the 2001 Ukraine census. We used prevalence odds ratios (ORs), with 95% CIs, as the measure of association between type 2 diabetes and famine exposure during gestation. The OR in this setting approximates the risk of famine exposure on type 2 diabetes.22 Because of differences in prevalence of adult type 2 diabetes by season of birth in Ukraine,23 we plotted the ORs for type 2 diabetes by year and season of birth (January to June vs July to December) and famine severity in the region of birth in 1932–33 (no famine, severe famine, extreme famine) relative to the odds of type 2 diabetes for region-specific births in July to December, 1938, taken as the reference group. After recording the highest prevalence of type 2 diabetes for births in the first half-year of 1934, we calculated stratified Mantel-Haenszel ORs for type 2 diabetes within each region of birth associated with births in the first half of 1934 relative to births in all other first half-years separately for men and women and combined (adjusted for sex). We
Jan–Jun Jul–Dec
All years 1938 1937 1936 1935 1934
We confirmed the exposure status of the selected regions in the population age pyramids of the 2001 Ukrainian census. These pyramids show a marked under-representation of births in the famine years in the eastern Ukraine, but not the western Ukraine, regions. In both geographical areas, the demographic losses of World War II were of a similar degree (appendix). To ascertain the outcome of interest in our study population, we identified all prevalent cases of type 2 diabetes from the national Ukraine diabetes register (Komisarenko Institute of Endocrinology and Metabolism, Kiev, Ukraine) in the selected regions among individuals born between 1930 and 1938. The register was created in 2000, and was last updated in 2008. Reports by primary care physicians for people with diabetes are the primary data source for the register. Diagnosis of type 2 diabetes was based on WHO 1999 criteria.21 To minimise bias due to misclassification of the type of diabetes, we restricted cases to individuals diagnosed with type 2 diabetes aged 40 years or older. We further identified a subgroup of patients with type 2 diabetes who received insulin treatment to confirm the consistency of our findings in this subgroup. This study is one of historical cohorts, and no individual linkages can be established between cohorts defined at birth and follow-up data. The use of anonymised data was approved by the institutional review board of the Komisarenko Institute of Endocrinology and Metabolism, Kiev, Ukraine, and the need for informed consent for the purpose of this study was waived.
1447
Articles
3
Articles
Odds ratio
2·0
Role of the funding source
Extreme famine regions Severe famine regions Non-famine regions
The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.
1·0
Results
De
ly– Ju
Jan
ua
ry –Ju
ne c Jan em ua ber ry Ju –Ju ly– De ne ce Jan m ua ber ry Ju –Ju ly– De ne c Jan em ua ber ry Ju –Ju ly– De ne ce Jan m ua ber ry Ju –Ju ly– De ne ce Jan m ua ber ry Ju –Ju ly– De ne c Jan em ua ber ry Ju –Ju ly– De ne ce Jan m ua ber ry Ju –Ju ly– De ne ce Jan m ua ber ry Ju –Ju ly– De ne ce m be r
0·5
1930
1931
1932
1933
1934
1935
1936
1937
1938
Time (month and year)
Figure 2: Odds for registered cases of type 2 diabetes (2000–08) by month and year of birth, and famine severity in the region of birth (1932–33) Odds ratios comparing the odds of type 2 diabetes among births by half-year to the odds for births in July to December, 1938 (reference group). Error bars show 95% CIs. Men
Women
Combined (adjusted for sex)
Stratified analysis within birth region* Births in first half of 1934 (January to June) versus births in first half of all other years Number of individuals
279 226
No famine region (n=98 302)
418 835
1·10 (0·94–1·28)
0·95 (0·85–1·06)
·· 1·00 (0·91–1·09)
Severe famine region (n=222 084)
1·34 (1·14–1·57)
1·22 (1·08–1·38)
1·26 (1·14–1·39)
Extreme famine region (n=377 675)
1·35 (1·19–1·54)
1·52 (1·40–1·66)
1·47 (1·37–1·58)
Births in second half of 1934 (July to December) versus births in second half of all other years Number of individuals
283 562
439 401
··
No famine region (n=93 941)
0·92 (0·77–1·09)
0·98 (0·87–1·10)
0·96 (0·87–1·06)
Severe famine region (n=227 636)
0·90 (0·77–1·05)
0·87 (0·77–0·98)
0·88 (0·80–0·97)
Extreme famine region (n=401 386)
1·03 (0·91–1·17)
1·02 (0·94–1·11)
1·02 (0·96–1·10)
Multivariable analysis including all birth regions and birth half-years† Births in first half of 1934 versus all other births Number of individuals No famine (n=192 243)
562 788
858 236
1·0 (reference)
1·0 (reference)
·· 1·0 (reference)
Severe famine (n=449 720)
1·16 (0·94–1·45)
1·26 (1·07–1·48)
1·23 (1·07–1·40)
Extreme famine (n=779 061)
1·28 (1·05–1·55)
1·65 (1·44–1·90)
1·51 (1·35–1·69)
Data are odds ratio (95% CI), unless otherwise indicated. *Mantel-Haenszel odds ratios within birth region, with and without adjustment for sex. †Odds ratios for January to June, 1934, for interactions between date of birth and region of birth from logistic regression analysis, adjusted for birth season, region of birth, and sex where applicable.
Table 2: Odds of type 2 diabetes by regional severity of Ukraine famine 1932–33, comparing men and women still alive in 2000 who were born in 1934, with all other births in 1930–38, by region of birth and sex
did the same for births in the second half-year. We then did multivariable logistic regression analyses combining all regions to calculate cross-product ORs for interactions between date of birth and region of birth. These models included indicators for the main effects of season, region of birth, their interaction terms, and an indicator for sex when applicable. We did analyses with Stata (version 11). 4
We determined the odds of type 2 diabetes by date and region of birth by using data from 43 150 patients with diabetes and 1 421 024 individuals born in 1930–38. The study population included 562 788 (40%) men and 858 236 (60%) women. 55% of the population were born in regions with extreme population losses, 32% in regions with severe losses, and 14% in regions with no losses (table 1). The mean age at type 2 diabetes diagnosis was 61·7 years (SD 8·9) in men and 62·1 years (8·6) in women. Individuals born in regions with extreme, severe, and no population loss were diagnosed at mean ages of 61·0 years (SD 8·4), 61·6 years (8·4), and 64·6 years (8·5), respectively. Type 2 diabetes was more common in individuals born in January to June compared with those born in July to December, 1930–38 (table 1). We noted a seasonal pattern for odds of type 2 diabetes, with a peak for births in the first half-year of 1934. Famine mortality peaked in the second quarter of 1933 (appendix). Stratification by region of birth showed that the seasonal pattern for type 2 diabetes was present in all regions (figure 2). The further increase in type 2 diabetes in individuals born between January and June, 1934, shows a gradient with the largest increase in regions exposed to the most severe famine (figure 2). In stratified analyses, we calculated the odds of type 2 diabetes for people born in 1934 relative to births in any of the remaining years between 1930 and 1938, first for births in the first half of all years and then for births in the second half. In sex-adjusted stratified analyses by region of birth, individuals born in January to June, 1934, in regions with extreme famine had a 1·5 times increase in odds of type 2 diabetes compared with other January to June births between 1930 and 1938 (table 2); births in regions with severe famine showed a 1·3 times increase and births in regions without famine showed no increase (table 2). Odds of type 2 diabetes for individuals born in July to December, 1934, did not differ from the other birth years (table 2). We further analysed these findings by multivariable sexspecific and sex-adjusted logistic regression analyses combining all regions of birth, with adjustment for date of birth by 6 month intervals during 1930–38, and for region of birth with indicator variables to denote births in regions of extreme famine or severe famine. Results of the multivariable analysis confirmed those of the stratified analysis, with similar patterns in men and women (table 2). The overall findings were similar for the subgroup of 12 566 patients with type 2 diabetes receiving
www.thelancet.com/diabetes-endocrinology Published online September 3, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00279-X
Articles
6
Births per month Deaths per month Prevalence of type 2 diabetes
4
4
3
3
2
2
1
1
1932
1933
0
ril M ay Ju ne Ju Au ly Se g pt us em t Oc ber No tob ve er De mb ce er m be r
ch
ar
M
Ap
ch
ar
M
Ap
ch
ar
M
Ap
Jan Fe uary br ua ry
0
ril M ay Ju ne J Au uly Se g pt us em t O ber No ctob ve er De mb ce er m Jan ber Fe uary br ua ry
5
ril M ay Ju ne Au July Se gu pt st em Oc ber No tob ve er De mb ce er m Jan ber Fe uary br ua ry
5
Prevalence of type 2 diabetes (%)
Births and deaths (monthly number/mean)
6
1934
Time (month and year)
Figure 3: Monthly births and deaths in eastern Ukraine in 1932–34, and prevalence of type 2 diabetes between 2000 and 2008, by year and month of birth Ratio of monthly values to overall mean. Deaths rates are per 1000 population,15 birth counts are from the Ukraine 2001 census, and prevalence of type 2 diabetes is from the Ukraine national diabetes register.
insulin treatment, although with smaller numbers of cases the estimates were less precise (appendix). In analysis of the timing of the famine in relation to mortality and month of birth in 1932–34, and type 2 diabetes diagnosed in 2000–08, diabetes prevalence was most increased in individuals born between February and April, 1934. These births took place 9 months after famine mortality peaked between May and July, 1933 (figure 3).18 The birth-count deficit was likewise most pronounced 9 months after famine mortality peaked (figure 3).
Discussion Our findings show a 1·5 times increase in odds of developing type 2 diabetes in adulthood in individuals born during the famine in eastern Ukraine in the first 6 months of 1934. This finding is in broad agreement with the 40% increase in diabetes reported in people born in eastern Austria after the 1918–19 famine following the collapse of the Austro-Hungarian Empire.13 The data from Ukraine additionally point to early gestation as a critical timing window for determining risk of type 2 diabetes. This question could not be adequately addressed in studies of the Dutch famine of 1944–45,10, 11 because of inadequate sample size, or in studies of the Chinese famine of 1959–61,12 because the timing of the exposure
was less well defined. An important new finding in our study is the dose–response relation between the severity of famine exposure and the likelihood of later type 2 diabetes. Our monthly data for the years 1932–34 show that the odds of type 2 diabetes were highest, and the birth counts lowest, among people born between February and April, 1934, 9 months after famine mortality peaked between May and July, 1933.15 This finding suggests that early gestation is a period of particular vulnerability for the developing fetus. These results are in agreement with the increase in all-cause mortality in military recruits with prenatal exposure to the Dutch famine of 1944–45, which was limited to men exposed in early gestation.24 We previously used births from four regions in the Ukraine to report on predisposition to type 2 diabetes in relation to prenatal famine.25,26 These findings showed differences in diabetes prevalence depending on the season of birth, and increases in prevalence in individuals born in early 1934, but were deficient in several respects. These studies, for example, did not include regions without famine exposure, the birth data were limited to 6 month time windows, and no statistical modelling was done. In the present study, we added two regions with no famine exposure, one with severe exposure, and two with
www.thelancet.com/diabetes-endocrinology Published online September 3, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00279-X
5
Articles
extreme exposure, for a total of nine regions overall. We were thus able to establish a dose–response relation between famine severity in the region of birth and prevalence of type 2 diabetes in adulthood. With use of month of birth, we were also able to identify early gestation as a time window that seems to be particularly sensitive to long-term effects of prenatal famine exposure. Additionally, we could quantify the effect of prenatal famine on type 2 diabetes in men and women, accounting for region of birth, seasonal effects, and their possible interactions. The association between season of birth and type 2 diabetes as reported initially by us (AMV and MDK) for three Ukrainian regions23 is related to seasonal differences in food availability. Deficiencies of a diet chronically marginal in calories, and most probably in several nutrients, are then exacerbated during the winter and spring when food stocks from the previous harvest are depleted. In the present study, we show that the effect of the existing season of birth on late-life type 2 diabetes in individuals born between 1930 and 1938 was substantially larger for births in the first half-year of 1934 after famine in early gestation. Regarding possible mechanisms, epigenetic changes in relation to early nutrition could provide a causal link with long-term health outcomes,27 although specific effects on gene expression need further investigation.28 Suboptimum maternal nutrition during gestation in animal models induces beta-cell dysfunction in offspring, which might involve the dysregulation of gene expression through epigenetic modification.29 Furthermore, a study of DNA methylation profiling in type-2-diabetic and nondiabetic cadaveric donors showed epigenetic changes in human pancreatic islet cells.30 These findings need further exploration. In survivors of the Dutch famine exposed in early gestation, we showed DNA methylation changes of the imprinted IGF2 gene and of candidate genes involved in metabolic disease.31,32 Further studies of biological materials in survivors of the Ukraine famine might therefore be valuable. The strengths of our study are the large population size, availability of individual birth dates, inclusion of regions with different levels of famine exposure, and the long follow-up period. However, there are also several limitations. There could be misclassification of an individual’s exposure to famine or biases related to diabetes diagnosis. Additionally, the study findings could be confounded by unmeasured risk factors for type 2 diabetes that are also related to famine exposure, or could be biased by differences in survival or loss to follow-up in specific exposure groups. Migrations within Ukraine and between-region comparisons could also lead to biased estimates. We classified an individual’s famine exposure by place and date of birth on the basis of monthly mortality at the group level in the region of birth between 1932 and 1934. It is possible that individual variations in access to food 6
could have existed within famine-exposed groups because of special individual privileges or other circumstances. Inferences are therefore limited to aggregated data that represent the group experience. We obtained data about diabetes outcomes from the Ukraine national diabetes register. Because insulin treatment for registered patients with type 2 diabetes is free of charge in Ukraine, most of those in need of such treatment were likely to have registered and be included in the register. In 2008, the coverage of insulin-treated cases by the register was about 90%, based on estimates from Health Ministry surveys, and the coverage of noninsulin-treated cases was about 40–60%, depending on the region. To minimise bias from type 1 diabetes contamination, we restricted the cases to individuals diagnosed at age 40 years or older. The type 2 diabetes cases in the registry were ascertained among individuals born between 1930 and 1938 who were still alive in 2000. The outcomes indicate two different tendencies. Early famine exposure might lead to an increased risk of type 2 diabetes in later life and to increased mortality such that these individuals are no longer alive to be diagnosed at older ages. Alternatively, survivors of the famine might represent more resilient individuals who are less likely to have type 2 diabetes in old age.33 Both mechanisms could lead to an underestimate of true famine effects. In this study, women in all regions were more likely to be diagnosed with type 2 diabetes than men, and were older at diagnosis. This difference could be because women have greater longevity than men. However, the effect of the famine on adult type 2 diabetes is no different between men and women, as our sexspecific analyses show. Migrations could be a potential issue if men and women born in non-famine regions in the first half year of 1934 were more likely to be diagnosed with type 2 diabetes in later life in former famine regions or vice versa. This situation might lead to biased estimates of famine effects. However, internal migrations in the former Soviet Union were subject to severe administrative restrictions, especially in the rural areas. At the time of the collapse of the Soviet Union in 1991, the study participants were also close to retirement age, after which migrations are even less common than in younger people. We do not think that the small potential for migrations can explain the difference in odds of type 2 diabetes related to early famine exposure. A final but important limitation is that we cannot exclude potential confounding by past and more recent lifestyles. For example, we have no information about socioeconomic status or present dietary habits, and information about BMI, which could be an important intermediary factor, is not yet available for all individuals. We note, however, that the region-specific analyses of the effect of famine on type 2 diabetes are in close agreement with the multivariable analyses across regions. This finding suggests that the study results are not greatly
www.thelancet.com/diabetes-endocrinology Published online September 3, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00279-X
Articles
biased by potential lifestyle differences between eastern and western Ukrainian populations. Our findings have several implications for understanding the rapidly increasing risk of type 2 diabetes in both developed and developing countries around the world. The emerging pandemic is partly driven by the postnatal effects of population ageing, rising levels of obesity and physical inactivity, and greater longevity in patients with type 2 diabetes attributable to improved management.34 Our findings contribute to emerging suggestions from studies of famine in early life10–13 showing that the prenatal environment during critical windows of development might also play an important part. Although we know from genome-wide DNA analysis that prenatal famine exposure in early gestation can affect DNA methylation at specific CpG dinucleotides related to metabolic disease,31 comparisons of specific epigenetic changes related to type 2 diabetes in populations with and without early famine exposure have not yet been possible. Important to establish is whether the prenatal and postnatal determinants of type 2 diabetes risk might have specific epigenetic pathways in common. In summary, we noted a dose–response association in men and women between famine severity in early gestation and type 2 diabetes in later life. Further studies of biological mechanisms underlying this association should focus on changes in DNA methylation in this critical period, which might also be relevant for determinants of type 2 diabetes risk in later life. Contributors MDK and AMV collected data. LHL, MDK, and AMV analysed data. LHL, MDK, and AMV prepared the manuscript and approved the final version. Declaration of interests We declare no competing interests. Acknowledgments The Ukraine State Diabetes Mellitus Program provided funding for this study (project number 0106U000844). LHL received support from a NIAS-Lorentz Fellowship of the Royal Netherlands Academy of Sciences and a grant from the National Institute of Aging, National Institutes of Health (number RO1 AG028593). We thank S G Wheatcroft (University of Melbourne, Parkville, VIC, Australia, and Nazarbayev University, Astana, Kazakhstan) for information about Ukraine deaths 1932–38, and Volodymir Kovtun (Komisarenko Institute of Endocrinology and Metabolism, Kiev, Ukraine) and Lyudmila Mechova (Chebotarev Institute of Gerontology, Kiev, Ukraine) for technical assistance. References 1 Kermack WO, McKendrick AG, McKinlay PL. Death-rates in Great Britain and Sweden. Some general regularities and their significance. Lancet 1934; 223: 698–703. 2 Hales CN, Barker DJ. Type 2 (non-insulin-dependent) diabetes mellitus: the thrifty phenotype hypothesis. Diabetologia 1992; 35: 595–601. 3 Jones RH, Ozanne SE. Fetal programming of glucose-insulin metabolism. Mol Cell Endocrinol 2009; 297: 4–9. 4 Whincup PH, Kaye SJ, Owen CG, et al. Birth weight and risk of type 2 diabetes: a systematic review. JAMA 2008; 300: 2886–97. 5 Gillman MW. Developmental origins of health and disease. N Engl J Med 2005; 353: 1848–50. 6 Stein AD, Zybert PA, van de Bor M, Lumey LH. Intrauterine famine exposure and body proportions at birth: the Dutch Hunger Winter. Int J Epidemiol 2004; 33: 831–86.
7
8
9 10 11
12
13
14
15
16
17
18
19 20
21
22
23
24
25
26
27
28 29
Horikoshi M, Yaghootkar H, Mook-Kanamori DO, et al. New loci associated with birth weight identify genetic links between intrauterine growth and adult height and metabolism. Nat Genet 2013; 45: 76–82. Joseph KS, Kramer MS. Review of the evidence on fetal and early childhood antecedents of adult chronic disease. Epidemiol Rev 1996; 18: 158–74. Lumey LH, Stein AD, Susser E. Prenatal famine and adult health. Annu Rev Public Health 2011; 32: 237–62. Ravelli AC, van der Meulen JH, Michels RP, et al. Glucose tolerance in adults after prenatal exposure to famine. Lancet 1998; 351: 173–77. Lumey LH, Stein AD, Kahn HS. Food restriction during gestation and impaired fasting glucose or glucose tolerance and type 2 diabetes mellitus in adulthood: evidence from the Dutch Hunger Winter Families Study. J Dev Orig Health Dis 2009; 1: S164. Li Y, He Y, Qi L, et al. Exposure to the Chinese famine in early life and the risk of hyperglycemia and type 2 diabetes in adulthood. Diabetes 2010; 59: 2400–06. Thurner S, Klimek P, Szell M, et al. Quantification of excess risk for diabetes for those born in times of hunger, in an entire population of a nation, across a century. Proc Natl Acad Sci USA 2013; 110: 4703–07. Stanner SA, Bulmer K, Andres C, et al. Does malnutrition in utero determine diabetes and coronary heart disease in adulthood? Results from the Leningrad siege study, a cross sectional study. BMJ 1997; 315: 1342–48. Rudnytskyi O, Levchuk N, Wolowyna O, et al. Demography of a man-made human catastrophe: the case of massive famine in Ukraine 1932–1933. Can Stud Popul 2015; 42: 53–80. Carynnyk M, Luciuk LY, Kordan BS. The Foreign Office and the famine. British documents on Ukraine and the great famine of 1932–1933. Kingston, ON: The Limestone Press, 1988. Davies RW, Wheatcroft SG. The years of hunger: Soviet agriculture, 1931–1933. New York, NY: Palgrave Macmillan, 2010. http://www. europarl.europa.eu/sides/getDoc.do?type=TA&reference=P6-TA-20080523&language=EN&ring=P6-RC-2008-0571 (accessed Dec 10, 2013). Ukrainian Parliament. Law of Ukraine No 376–V. On the Holodomor of 1932–1933 in Ukraine. 2006. http://www.president. gov.ua/documents/5280.html (accessed Dec 10, 2013). European Parliament. Commemoration of the Holodomor, the artificial famine in Ukraine (1932–1933). 2008. Vallin J, Mesle F, Adamets S, Pyrozhkov S. A new estimate of Ukrainian population losses during the crises of the 1930s and 1940s. Popul Stud 2002; 56: 249–63. WHO. Definition, diagnosis, and classification of diabetes mellitus and its complications. Part 1: diagnosis and classification of diabetes mellitus. Geneva: World Health Organization, 1999. Rothman KJ, Greenland S, Lash TL. Modern epidemiology, 3rd edn. Philadelphia, PA: Wolters Kluwer/Lippincott Williams & Wilkins, 2008. Vaiserman AM, Khalangot MD, Carstensen B, et al. Seasonality of birth in adult type 2 diabetic patients in three Ukrainian regions. Diabetologia 2009; 52: 2665–67. Ekamper P, van Poppel F, Stein AD, Lumey LH. Independent and additive association of prenatal famine exposure and intermediary life conditions with adult mortality between age 18–63 years. Soc Sci Med 2014; 119: 232–39. Vaiserman AM, Khalangot ND, Pisaruk AV, et al. Predisposition to type II diabetes in Ukraine residents exposed to famine 1932–1933 during prenatal development. Adv Gerontol 2010; 23: 588–92 (in Russian). Vaiserman AM, Khalangot MD, Strele I, Lumey LH. Early-life exposure to the Ukraine famine of 1933 and type 2 diabetes in adulthood. In: Lumey LH, Vaiserman AM, eds. Early life nutrition and adult health and development. New York, NY: Nova Science Publishers, 2013. pp 145–60. Portha B, Fournier A, Ah Kioon MD, Mezger V, Movassat J. Early environmental factors, alteration of epigenetic marks and metabolic disease susceptibility. Biochimie 2014; 97: 1–15. Hochberg Z, Feil R, Constancia M, et al. Child health, developmental plasticity, and epigenetic programming. Endocr Rev 2011; 32: 159–224. Sandovici I, Hammerle CM, Ozanne SE, Constancia M. Developmental and environmental epigenetic programming of the endocrine pancreas: consequences for type 2 diabetes. Cell Mol Life Sci 2013; 70: 1575–95.
www.thelancet.com/diabetes-endocrinology Published online September 3, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00279-X
7
Articles
30
31
32
8
Volkmar M, Dedeurwaerder S, Cunha DA, et al. DNA methylation profiling identifies epigenetic dysregulation in pancreatic islets from type 2 diabetic patients. EMBO J 2012; 31: 1405–26. Tobi EW, Slieker RC, Stein AD, et al. Early gestation as the critical time-window for changes in the prenatal environment to affect the adult human blood methylome. Int J Epidemiol 2015; published online May 5. DOI:101093/ije/dyv043. Heijmans BT, Tobi EW, Stein AD, et al. Persistent epigenetic differences associated with prenatal exposure to famine in humans. Proc Natl Acad Sci USA 2008; 105: 17046–49.
33
34
Song S, Wang W, Hu P. Famine, death, and madness: schizophrenia in early adulthood after prenatal exposure to the Chinese Great Leap Forward Famine. Soc Sci Med 2009; 68: 1315–21. Vaag A, Brons C, Gillberg L, et al. Genetic, nongenetic and epigenetic risk determinants in developmental programming of type 2 diabetes. Acta Obstet Gynecol Scand 2014; 93: 1099–108.
www.thelancet.com/diabetes-endocrinology Published online September 3, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00279-X
Association between type 2 diabetes and prenatal exposure to the Ukraine famine of 1932–33: a retrospective cohort study L H Lumey, Mykola D Khalangot, Alexander M Vaiserman
Summary Background The effect of fetal and early childhood living conditions on adult health has long been debated, but empirical assessment in human beings remains a challenge. We used data from during the man-made Ukrainian famine of 1932–33 to examine the association between restricted nutrition in early gestation and type 2 diabetes in offspring in later life. Methods We included all patients with type 2 diabetes diagnosed at age 40 years or older in the Ukraine national diabetes register 2000–08, and used all individuals born between 1930 and 1938 from the 2001 Ukraine national census as the reference population. This study population includes individuals born before and after the famine period as controls, and those from regions that experienced extreme, severe, or no famine. We used prevalence odds ratios (ORs) as the measure of association between type 2 diabetes and early famine exposure, with stratification by region, date of birth, and sex for comparisons of diabetes prevalence in specific subgroups. Findings Using these two datasets, we compared the odds of type 2 diabetes by date and region of birth in 43 150 patients with diabetes and 1 421 024 individuals born between 1930 and 1938. With adjustment for season of birth, the OR for developing type 2 diabetes was 1·47 (95% CI 1·37–1·58) in individuals born in the first half of 1934 in regions with extreme famine, 1·26 (1·14–1·39) in individuals born in regions with severe famine, and there was no increase (OR 1·00, 0·91–1·09) in individuals born in regions with no famine, compared with births in other time periods. Multivariable analyses confirmed these results. The associations between type 2 diabetes and famine around the time of birth were similar in men and women. Interpretation These results show a dose–response relation between famine severity during prenatal development and odds of type 2 diabetes in later life. Our findings suggest that early gestation is a critical time window of development; therefore, further studies of biological mechanisms should include this period. Funding Ukraine State Diabetes Mellitus Program, US National Institutes of Health.
Introduction Poor living conditions around the time of birth have long been thought to have lasting effects on health.1 Nutrition during pregnancy has been proposed as an explanation for the occurrence of type 2 diabetes later in life,2, 3 but direct measures are hard to obtain. Birthweight is inversely associated with type 2 diabetes4 and has been used as a marker of fetal nutrition despite its limitations;5 however, birthweight is not affected by nutrition changes in early gestation.6 Fetal growth and type 2 diabetes could also have genetic links.7 Causal inference in the assessment of fetal antecedents of adult disease is therefore still a challenge, and reported associations could be biased rather than causal.8 Here we present a different approach to examine these questions that does not focus on birthweight alone. To better identify any effects of early nutrition, studies of man-made famines can offer distinct opportunities for the unbiased comparison of individuals born under such conditions with unexposed controls.9 These approaches can provide information about specific pregnancy exposures rather than size at birth alone, especially if the population at risk, the timing and degree of exposure, and relevant health outcomes can be accurately defined.
For type 2 diabetes in later life, studies of the Dutch famine of 1944–45,10,11 the Chinese famine of 1959–62,12 and three famines in 20th century Austria13 all suggest a relation with early life nutrition. By contrast, a study of the Siege of Leningrad of 1941–44 does not show an association.14 All studies have a particular strength: the timing of the famine in relation to the stage of gestation was particularly well defined in the Dutch study, as was the risk of adult hyperglycaemia in the Chinese study, and a large number of patients receiving antidiabetic drugs over a long time period was included in the nationwide Austrian study. However, no study combines all these strengths. To overcome the limitations of previous approaches, we used the setting of the Ukraine famine of 1932–33 to study the association between early-life nutrition and late-life type 2 diabetes in a large population, in cohorts well defined with respect to the timing of the famine in relation the stage of pregnancy and the severity of the famine around the time of birth. The Great Ukrainian Famine (also known as the Holodomor) was caused by the Soviet Union Government’s forced collectivisation of agriculture in the early 1930s. Although severe famines occurred in
www.thelancet.com/diabetes-endocrinology Published online September 3, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00279-X
Lancet Diabetes Endocrinol 2015 Published Online September 3, 2015 http://dx.doi.org/10.1016/ S2213-8587(15)00279-X See Online/Comments http://dx.doi.org/10.1016/ S2213-8587(15)00323-X Department of Epidemiology, Mailman School of Public Health, Columbia University, New York, NY, USA (L H Lumey MD); Molecular Epidemiology, Leiden University Medical Center, Leiden, Netherlands (L H Lumey); Komisarenko Institute of Endocrinology and Metabolism (M D Khalangot MD) and Chebotarev Institute of Gerontology (A M Vaiserman PhD), National Academy of Medical Sciences, Kiev, Ukraine; and Shupyk National Medical Academy of Postgraduate Education, Kiev, Ukraine (M D Khalangot) Correspondence to: Dr L H Lumey, Department of Epidemiology, Mailman School of Public Health, Columbia University, New York, NY 10032, USA [email protected]
1
Articles
Panel: Research In context Evidence before this study Studies of the Dutch famine of 1944–45, the Chinese famine of 1959–62, and three famines in 20th century Austria suggest that risk of type 2 diabetes might be affected by prenatal nutrition. Each of these studies has specific strengths, but unanswered questions remain because of limitations of sample size, uncertainty about the timing of the exposure, and short duration of follow-up. Added value of this study To overcome these limitations we collected evidence for the Ukraine famine of 1932–33. These data provide a large study population, well defined with respect to the timing of the famine in relation the stage of pregnancy. Regional analysis also provides estimates of the potentially differential effect of severe versus extreme prenatal famine exposure on type 2 diabetes in
Volyn
Rivne
Chernihiv
Kiev Khmelnytsky Vinnytsia
Kharkiv
Cherkasy
Kherson 0 0
100 miles 100 km Non-famine regions Severe famine regions Extreme famine regions
Luhansk
2
Implications of all the available evidence Our findings identify early gestation as a sensitive period and show a dose–response relation between the severity of famine and the odds of type 2 diabetes in later life. Our findings are consistent with those from epigenetic studies that point to effects of famine in early gestation on DNA methylation of genes related to metabolic disease. As a next step, specific epigenetic changes and morbidity of type 2 diabetes should be compared in populations with and without early famine exposure. Such comparison might establish whether prenatal and postnatal determinants of type 2 diabetes have specific epigenetic pathways in common.
the Holodomor as a crime against the Ukrainian people and against humanity.17 Estimates of population decline due to the famine range from 2·7 million to 3·9 million.18,19 The regional severity of the famine was determined by the prevailing political situation. In eastern Ukraine, under Soviet control, some regions suffered extreme population losses of more than 25% and others had severe losses of about 10–20%,18,20 with a ten-times mortality increase between April and July, 1933, compared with prefamine times.15 There were no population losses in western Ukraine, a region that was still under Polish rule during the famine years and not exposed to famine (figure 1).
Azov Sea
Methods
Black Sea Crimea
Figure 1: Regional geography of regions affected by the Ukrainian famine of 1932–33 Map of Ukraine showing the locations of the nine regions included in this study. Severely exposed regions are defined by a rate of population decline of up to 15% between 1929 and 1933, and extremely exposed regions by a rate of between 20% and 25%.18,20 See Online for appendix
later life. Our study combines the strengths of all existing studies.
several Soviet regions, including Kazakhstan, where no grain was produced, an especially strict food policy was applied to Ukraine.15 Ukraine had two successive bad harvests in 1931 and 1932. By the spring of 1932, there was an absolute shortage of grain, which became more severe in the ensuing 12 months. Peasants were also required to return not only grain they had earned by meeting production targets, but also other food such as potatoes and poultry. These events resulted in a catastrophic mortality increase between April and July, 1933.15 In 2007, the Parliament of Ukraine passed a law characterising the famine as a genocide of the Ukrainian people.16 In 2008, the European Parliament recognised
Data sources The reference population is derived from region-specific population counts from the 2001 Ukraine national census of all individuals born between 1930 and 1938, categorised by sex and month of birth. This information was collected through the census for all 25 Ukrainian regions, but is not in the public domain. After formal applications to the Ukraine State Statistics Committee, we obtained permissions from nine regions for the use of selected census data under strict privacy conditions. The regions represent all levels of famine exposure as defined by the magnitude of population losses. From eastern Ukraine, we included four regions with exposure to extreme famine in 1932–33 (Kharkiv, Kherson, Cherkasy, and Luhansk) and three with exposure to severe famine (Chernihiv, Vinnytsia, Khmelnytskyi). From western Ukraine, with no exposure to famine, we included two regions (Rivne and Volyn; figure 1). The reference population was stratified by region, date of birth, and sex for comparisons of type 2 diabetes prevalence in specific subgroups.
www.thelancet.com/diabetes-endocrinology Published online September 3, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00279-X
Table 1: Cases of type 2 diabetes and current census counts by year and season of birth (1930–38), and sex, by severity of Ukraine famine (1932–33) in the region of birth
Registered cases of type 2 diabetes were from the national Ukraine diabetes register 2000–08, and census counts were from the 2001 Ukraine national census. Regions with no famine were Rivne and Volyn; with severe famine were Chernihiv, Vinnytsia, and Khmelnytskyi; and with extreme famine were Kharkiv, Kherson, Cherkasy, and Luhansk.
242 078
6860 7932
224 496 40 615
1108 1282
38 713 40 022
1155 1382
41 093 34 125
999 1064
29 436 28 349
790 901
23 073 22 696
655 611
11 855 13 621
389 524
12 853 16 975
522 691
20 218 20 083
563 665
21 435 25 592
679 812
25 820 Census count
Type 2 diabetes cases
Extreme famine
3937
139 413 134 116
4234 626
20 369 19 793
709 648
20 062 21 499
728 526
18 335 14 362
580 480
15 059 12 103
421 298
11 955 7558
285 274
10 147 9859
285 347
13 005 15 557
384 288
13 096 14 674
397 450
17 385 18 711
445 Type 2 diabetes cases
Severe famine
Census count
57 910
2987 3547
60 223 6230
379 429
6678 6462
411 408
6761 6728
364 431
6933 6589
342 421
6629 6350
322 349
6201 6014
291 353
6040 6753
317 415
7155 5696
267 335
6307 Census count
294 406
7519
Type 2 diabetes cases
No famine
Women
www.thelancet.com/diabetes-endocrinology Published online September 3, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00279-X
7088
159 308
2843 3367
153 179 27 683
486 563
27 506 27 032
521 649
29 046 22 908
427 454
20 825 19 465
320 397
16 530 15 657
287 258
8922 9111
150 186
8821 10 488
197 258
12 711 12 168
204 273
13 462 14 796
251 329
15 356 Census count
Type 2 diabetes cases
Extreme famine
2084
88 223 87 968
2182 341
13 911 13 957
358 357
13 820 15 543
408 300
12 355 10 348
274 250
10 077 8528
228 173
8050 5017
162 108
6030 6219
140 179
7086 9041
225 160
7621 8810
158 216
9273 10 505
229 Type 2 diabetes cases
Census count
36 031 Severe famine
1730
38 079 4076
157 203
4469 4172
202 219
4569 4184
175 214
4602 4157
190 209
4214 3863
144 198
4023 3926
141 163
3811 3924
143 183
4075 3583
150 136
3939 Census count
145 205
4377
Type 2 diabetes cases
No famine
Jan–Jun Jan–Jun Jul–Dec
4146
Jul–Dec Jan–Jun Jul–Dec Jan–Jun Jul–Dec Jan–Jun Jan–Jun Jul–Dec Jan–Jun Jul–Dec Jan–Jun Jul–Dec Jan–Jun Jul–Dec Men
1931 1930
Jul–Dec
1932
1933
Statistical analysis We prepared contingency tables stratified by birth region, year and month of birth, and sex of individuals with type 2 diabetes diagnosed in 2000–08 among births between 1930 and 1938 in the selected study regions in the 2001 Ukraine census. We used prevalence odds ratios (ORs), with 95% CIs, as the measure of association between type 2 diabetes and famine exposure during gestation. The OR in this setting approximates the risk of famine exposure on type 2 diabetes.22 Because of differences in prevalence of adult type 2 diabetes by season of birth in Ukraine,23 we plotted the ORs for type 2 diabetes by year and season of birth (January to June vs July to December) and famine severity in the region of birth in 1932–33 (no famine, severe famine, extreme famine) relative to the odds of type 2 diabetes for region-specific births in July to December, 1938, taken as the reference group. After recording the highest prevalence of type 2 diabetes for births in the first half-year of 1934, we calculated stratified Mantel-Haenszel ORs for type 2 diabetes within each region of birth associated with births in the first half of 1934 relative to births in all other first half-years separately for men and women and combined (adjusted for sex). We
Jan–Jun Jul–Dec
All years 1938 1937 1936 1935 1934
We confirmed the exposure status of the selected regions in the population age pyramids of the 2001 Ukrainian census. These pyramids show a marked under-representation of births in the famine years in the eastern Ukraine, but not the western Ukraine, regions. In both geographical areas, the demographic losses of World War II were of a similar degree (appendix). To ascertain the outcome of interest in our study population, we identified all prevalent cases of type 2 diabetes from the national Ukraine diabetes register (Komisarenko Institute of Endocrinology and Metabolism, Kiev, Ukraine) in the selected regions among individuals born between 1930 and 1938. The register was created in 2000, and was last updated in 2008. Reports by primary care physicians for people with diabetes are the primary data source for the register. Diagnosis of type 2 diabetes was based on WHO 1999 criteria.21 To minimise bias due to misclassification of the type of diabetes, we restricted cases to individuals diagnosed with type 2 diabetes aged 40 years or older. We further identified a subgroup of patients with type 2 diabetes who received insulin treatment to confirm the consistency of our findings in this subgroup. This study is one of historical cohorts, and no individual linkages can be established between cohorts defined at birth and follow-up data. The use of anonymised data was approved by the institutional review board of the Komisarenko Institute of Endocrinology and Metabolism, Kiev, Ukraine, and the need for informed consent for the purpose of this study was waived.
1447
Articles
3
Articles
Odds ratio
2·0
Role of the funding source
Extreme famine regions Severe famine regions Non-famine regions
The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.
1·0
Results
De
ly– Ju
Jan
ua
ry –Ju
ne c Jan em ua ber ry Ju –Ju ly– De ne ce Jan m ua ber ry Ju –Ju ly– De ne c Jan em ua ber ry Ju –Ju ly– De ne ce Jan m ua ber ry Ju –Ju ly– De ne ce Jan m ua ber ry Ju –Ju ly– De ne c Jan em ua ber ry Ju –Ju ly– De ne ce Jan m ua ber ry Ju –Ju ly– De ne ce Jan m ua ber ry Ju –Ju ly– De ne ce m be r
0·5
1930
1931
1932
1933
1934
1935
1936
1937
1938
Time (month and year)
Figure 2: Odds for registered cases of type 2 diabetes (2000–08) by month and year of birth, and famine severity in the region of birth (1932–33) Odds ratios comparing the odds of type 2 diabetes among births by half-year to the odds for births in July to December, 1938 (reference group). Error bars show 95% CIs. Men
Women
Combined (adjusted for sex)
Stratified analysis within birth region* Births in first half of 1934 (January to June) versus births in first half of all other years Number of individuals
279 226
No famine region (n=98 302)
418 835
1·10 (0·94–1·28)
0·95 (0·85–1·06)
·· 1·00 (0·91–1·09)
Severe famine region (n=222 084)
1·34 (1·14–1·57)
1·22 (1·08–1·38)
1·26 (1·14–1·39)
Extreme famine region (n=377 675)
1·35 (1·19–1·54)
1·52 (1·40–1·66)
1·47 (1·37–1·58)
Births in second half of 1934 (July to December) versus births in second half of all other years Number of individuals
283 562
439 401
··
No famine region (n=93 941)
0·92 (0·77–1·09)
0·98 (0·87–1·10)
0·96 (0·87–1·06)
Severe famine region (n=227 636)
0·90 (0·77–1·05)
0·87 (0·77–0·98)
0·88 (0·80–0·97)
Extreme famine region (n=401 386)
1·03 (0·91–1·17)
1·02 (0·94–1·11)
1·02 (0·96–1·10)
Multivariable analysis including all birth regions and birth half-years† Births in first half of 1934 versus all other births Number of individuals No famine (n=192 243)
562 788
858 236
1·0 (reference)
1·0 (reference)
·· 1·0 (reference)
Severe famine (n=449 720)
1·16 (0·94–1·45)
1·26 (1·07–1·48)
1·23 (1·07–1·40)
Extreme famine (n=779 061)
1·28 (1·05–1·55)
1·65 (1·44–1·90)
1·51 (1·35–1·69)
Data are odds ratio (95% CI), unless otherwise indicated. *Mantel-Haenszel odds ratios within birth region, with and without adjustment for sex. †Odds ratios for January to June, 1934, for interactions between date of birth and region of birth from logistic regression analysis, adjusted for birth season, region of birth, and sex where applicable.
Table 2: Odds of type 2 diabetes by regional severity of Ukraine famine 1932–33, comparing men and women still alive in 2000 who were born in 1934, with all other births in 1930–38, by region of birth and sex
did the same for births in the second half-year. We then did multivariable logistic regression analyses combining all regions to calculate cross-product ORs for interactions between date of birth and region of birth. These models included indicators for the main effects of season, region of birth, their interaction terms, and an indicator for sex when applicable. We did analyses with Stata (version 11). 4
We determined the odds of type 2 diabetes by date and region of birth by using data from 43 150 patients with diabetes and 1 421 024 individuals born in 1930–38. The study population included 562 788 (40%) men and 858 236 (60%) women. 55% of the population were born in regions with extreme population losses, 32% in regions with severe losses, and 14% in regions with no losses (table 1). The mean age at type 2 diabetes diagnosis was 61·7 years (SD 8·9) in men and 62·1 years (8·6) in women. Individuals born in regions with extreme, severe, and no population loss were diagnosed at mean ages of 61·0 years (SD 8·4), 61·6 years (8·4), and 64·6 years (8·5), respectively. Type 2 diabetes was more common in individuals born in January to June compared with those born in July to December, 1930–38 (table 1). We noted a seasonal pattern for odds of type 2 diabetes, with a peak for births in the first half-year of 1934. Famine mortality peaked in the second quarter of 1933 (appendix). Stratification by region of birth showed that the seasonal pattern for type 2 diabetes was present in all regions (figure 2). The further increase in type 2 diabetes in individuals born between January and June, 1934, shows a gradient with the largest increase in regions exposed to the most severe famine (figure 2). In stratified analyses, we calculated the odds of type 2 diabetes for people born in 1934 relative to births in any of the remaining years between 1930 and 1938, first for births in the first half of all years and then for births in the second half. In sex-adjusted stratified analyses by region of birth, individuals born in January to June, 1934, in regions with extreme famine had a 1·5 times increase in odds of type 2 diabetes compared with other January to June births between 1930 and 1938 (table 2); births in regions with severe famine showed a 1·3 times increase and births in regions without famine showed no increase (table 2). Odds of type 2 diabetes for individuals born in July to December, 1934, did not differ from the other birth years (table 2). We further analysed these findings by multivariable sexspecific and sex-adjusted logistic regression analyses combining all regions of birth, with adjustment for date of birth by 6 month intervals during 1930–38, and for region of birth with indicator variables to denote births in regions of extreme famine or severe famine. Results of the multivariable analysis confirmed those of the stratified analysis, with similar patterns in men and women (table 2). The overall findings were similar for the subgroup of 12 566 patients with type 2 diabetes receiving
www.thelancet.com/diabetes-endocrinology Published online September 3, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00279-X
Articles
6
Births per month Deaths per month Prevalence of type 2 diabetes
4
4
3
3
2
2
1
1
1932
1933
0
ril M ay Ju ne Ju Au ly Se g pt us em t Oc ber No tob ve er De mb ce er m be r
ch
ar
M
Ap
ch
ar
M
Ap
ch
ar
M
Ap
Jan Fe uary br ua ry
0
ril M ay Ju ne J Au uly Se g pt us em t O ber No ctob ve er De mb ce er m Jan ber Fe uary br ua ry
5
ril M ay Ju ne Au July Se gu pt st em Oc ber No tob ve er De mb ce er m Jan ber Fe uary br ua ry
5
Prevalence of type 2 diabetes (%)
Births and deaths (monthly number/mean)
6
1934
Time (month and year)
Figure 3: Monthly births and deaths in eastern Ukraine in 1932–34, and prevalence of type 2 diabetes between 2000 and 2008, by year and month of birth Ratio of monthly values to overall mean. Deaths rates are per 1000 population,15 birth counts are from the Ukraine 2001 census, and prevalence of type 2 diabetes is from the Ukraine national diabetes register.
insulin treatment, although with smaller numbers of cases the estimates were less precise (appendix). In analysis of the timing of the famine in relation to mortality and month of birth in 1932–34, and type 2 diabetes diagnosed in 2000–08, diabetes prevalence was most increased in individuals born between February and April, 1934. These births took place 9 months after famine mortality peaked between May and July, 1933 (figure 3).18 The birth-count deficit was likewise most pronounced 9 months after famine mortality peaked (figure 3).
Discussion Our findings show a 1·5 times increase in odds of developing type 2 diabetes in adulthood in individuals born during the famine in eastern Ukraine in the first 6 months of 1934. This finding is in broad agreement with the 40% increase in diabetes reported in people born in eastern Austria after the 1918–19 famine following the collapse of the Austro-Hungarian Empire.13 The data from Ukraine additionally point to early gestation as a critical timing window for determining risk of type 2 diabetes. This question could not be adequately addressed in studies of the Dutch famine of 1944–45,10, 11 because of inadequate sample size, or in studies of the Chinese famine of 1959–61,12 because the timing of the exposure
was less well defined. An important new finding in our study is the dose–response relation between the severity of famine exposure and the likelihood of later type 2 diabetes. Our monthly data for the years 1932–34 show that the odds of type 2 diabetes were highest, and the birth counts lowest, among people born between February and April, 1934, 9 months after famine mortality peaked between May and July, 1933.15 This finding suggests that early gestation is a period of particular vulnerability for the developing fetus. These results are in agreement with the increase in all-cause mortality in military recruits with prenatal exposure to the Dutch famine of 1944–45, which was limited to men exposed in early gestation.24 We previously used births from four regions in the Ukraine to report on predisposition to type 2 diabetes in relation to prenatal famine.25,26 These findings showed differences in diabetes prevalence depending on the season of birth, and increases in prevalence in individuals born in early 1934, but were deficient in several respects. These studies, for example, did not include regions without famine exposure, the birth data were limited to 6 month time windows, and no statistical modelling was done. In the present study, we added two regions with no famine exposure, one with severe exposure, and two with
www.thelancet.com/diabetes-endocrinology Published online September 3, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00279-X
5
Articles
extreme exposure, for a total of nine regions overall. We were thus able to establish a dose–response relation between famine severity in the region of birth and prevalence of type 2 diabetes in adulthood. With use of month of birth, we were also able to identify early gestation as a time window that seems to be particularly sensitive to long-term effects of prenatal famine exposure. Additionally, we could quantify the effect of prenatal famine on type 2 diabetes in men and women, accounting for region of birth, seasonal effects, and their possible interactions. The association between season of birth and type 2 diabetes as reported initially by us (AMV and MDK) for three Ukrainian regions23 is related to seasonal differences in food availability. Deficiencies of a diet chronically marginal in calories, and most probably in several nutrients, are then exacerbated during the winter and spring when food stocks from the previous harvest are depleted. In the present study, we show that the effect of the existing season of birth on late-life type 2 diabetes in individuals born between 1930 and 1938 was substantially larger for births in the first half-year of 1934 after famine in early gestation. Regarding possible mechanisms, epigenetic changes in relation to early nutrition could provide a causal link with long-term health outcomes,27 although specific effects on gene expression need further investigation.28 Suboptimum maternal nutrition during gestation in animal models induces beta-cell dysfunction in offspring, which might involve the dysregulation of gene expression through epigenetic modification.29 Furthermore, a study of DNA methylation profiling in type-2-diabetic and nondiabetic cadaveric donors showed epigenetic changes in human pancreatic islet cells.30 These findings need further exploration. In survivors of the Dutch famine exposed in early gestation, we showed DNA methylation changes of the imprinted IGF2 gene and of candidate genes involved in metabolic disease.31,32 Further studies of biological materials in survivors of the Ukraine famine might therefore be valuable. The strengths of our study are the large population size, availability of individual birth dates, inclusion of regions with different levels of famine exposure, and the long follow-up period. However, there are also several limitations. There could be misclassification of an individual’s exposure to famine or biases related to diabetes diagnosis. Additionally, the study findings could be confounded by unmeasured risk factors for type 2 diabetes that are also related to famine exposure, or could be biased by differences in survival or loss to follow-up in specific exposure groups. Migrations within Ukraine and between-region comparisons could also lead to biased estimates. We classified an individual’s famine exposure by place and date of birth on the basis of monthly mortality at the group level in the region of birth between 1932 and 1934. It is possible that individual variations in access to food 6
could have existed within famine-exposed groups because of special individual privileges or other circumstances. Inferences are therefore limited to aggregated data that represent the group experience. We obtained data about diabetes outcomes from the Ukraine national diabetes register. Because insulin treatment for registered patients with type 2 diabetes is free of charge in Ukraine, most of those in need of such treatment were likely to have registered and be included in the register. In 2008, the coverage of insulin-treated cases by the register was about 90%, based on estimates from Health Ministry surveys, and the coverage of noninsulin-treated cases was about 40–60%, depending on the region. To minimise bias from type 1 diabetes contamination, we restricted the cases to individuals diagnosed at age 40 years or older. The type 2 diabetes cases in the registry were ascertained among individuals born between 1930 and 1938 who were still alive in 2000. The outcomes indicate two different tendencies. Early famine exposure might lead to an increased risk of type 2 diabetes in later life and to increased mortality such that these individuals are no longer alive to be diagnosed at older ages. Alternatively, survivors of the famine might represent more resilient individuals who are less likely to have type 2 diabetes in old age.33 Both mechanisms could lead to an underestimate of true famine effects. In this study, women in all regions were more likely to be diagnosed with type 2 diabetes than men, and were older at diagnosis. This difference could be because women have greater longevity than men. However, the effect of the famine on adult type 2 diabetes is no different between men and women, as our sexspecific analyses show. Migrations could be a potential issue if men and women born in non-famine regions in the first half year of 1934 were more likely to be diagnosed with type 2 diabetes in later life in former famine regions or vice versa. This situation might lead to biased estimates of famine effects. However, internal migrations in the former Soviet Union were subject to severe administrative restrictions, especially in the rural areas. At the time of the collapse of the Soviet Union in 1991, the study participants were also close to retirement age, after which migrations are even less common than in younger people. We do not think that the small potential for migrations can explain the difference in odds of type 2 diabetes related to early famine exposure. A final but important limitation is that we cannot exclude potential confounding by past and more recent lifestyles. For example, we have no information about socioeconomic status or present dietary habits, and information about BMI, which could be an important intermediary factor, is not yet available for all individuals. We note, however, that the region-specific analyses of the effect of famine on type 2 diabetes are in close agreement with the multivariable analyses across regions. This finding suggests that the study results are not greatly
www.thelancet.com/diabetes-endocrinology Published online September 3, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00279-X
Articles
biased by potential lifestyle differences between eastern and western Ukrainian populations. Our findings have several implications for understanding the rapidly increasing risk of type 2 diabetes in both developed and developing countries around the world. The emerging pandemic is partly driven by the postnatal effects of population ageing, rising levels of obesity and physical inactivity, and greater longevity in patients with type 2 diabetes attributable to improved management.34 Our findings contribute to emerging suggestions from studies of famine in early life10–13 showing that the prenatal environment during critical windows of development might also play an important part. Although we know from genome-wide DNA analysis that prenatal famine exposure in early gestation can affect DNA methylation at specific CpG dinucleotides related to metabolic disease,31 comparisons of specific epigenetic changes related to type 2 diabetes in populations with and without early famine exposure have not yet been possible. Important to establish is whether the prenatal and postnatal determinants of type 2 diabetes risk might have specific epigenetic pathways in common. In summary, we noted a dose–response association in men and women between famine severity in early gestation and type 2 diabetes in later life. Further studies of biological mechanisms underlying this association should focus on changes in DNA methylation in this critical period, which might also be relevant for determinants of type 2 diabetes risk in later life. Contributors MDK and AMV collected data. LHL, MDK, and AMV analysed data. LHL, MDK, and AMV prepared the manuscript and approved the final version. Declaration of interests We declare no competing interests. Acknowledgments The Ukraine State Diabetes Mellitus Program provided funding for this study (project number 0106U000844). LHL received support from a NIAS-Lorentz Fellowship of the Royal Netherlands Academy of Sciences and a grant from the National Institute of Aging, National Institutes of Health (number RO1 AG028593). We thank S G Wheatcroft (University of Melbourne, Parkville, VIC, Australia, and Nazarbayev University, Astana, Kazakhstan) for information about Ukraine deaths 1932–38, and Volodymir Kovtun (Komisarenko Institute of Endocrinology and Metabolism, Kiev, Ukraine) and Lyudmila Mechova (Chebotarev Institute of Gerontology, Kiev, Ukraine) for technical assistance. References 1 Kermack WO, McKendrick AG, McKinlay PL. Death-rates in Great Britain and Sweden. Some general regularities and their significance. Lancet 1934; 223: 698–703. 2 Hales CN, Barker DJ. Type 2 (non-insulin-dependent) diabetes mellitus: the thrifty phenotype hypothesis. Diabetologia 1992; 35: 595–601. 3 Jones RH, Ozanne SE. Fetal programming of glucose-insulin metabolism. Mol Cell Endocrinol 2009; 297: 4–9. 4 Whincup PH, Kaye SJ, Owen CG, et al. Birth weight and risk of type 2 diabetes: a systematic review. JAMA 2008; 300: 2886–97. 5 Gillman MW. Developmental origins of health and disease. N Engl J Med 2005; 353: 1848–50. 6 Stein AD, Zybert PA, van de Bor M, Lumey LH. Intrauterine famine exposure and body proportions at birth: the Dutch Hunger Winter. Int J Epidemiol 2004; 33: 831–86.
7
8
9 10 11
12
13
14
15
16
17
18
19 20
21
22
23
24
25
26
27
28 29
Horikoshi M, Yaghootkar H, Mook-Kanamori DO, et al. New loci associated with birth weight identify genetic links between intrauterine growth and adult height and metabolism. Nat Genet 2013; 45: 76–82. Joseph KS, Kramer MS. Review of the evidence on fetal and early childhood antecedents of adult chronic disease. Epidemiol Rev 1996; 18: 158–74. Lumey LH, Stein AD, Susser E. Prenatal famine and adult health. Annu Rev Public Health 2011; 32: 237–62. Ravelli AC, van der Meulen JH, Michels RP, et al. Glucose tolerance in adults after prenatal exposure to famine. Lancet 1998; 351: 173–77. Lumey LH, Stein AD, Kahn HS. Food restriction during gestation and impaired fasting glucose or glucose tolerance and type 2 diabetes mellitus in adulthood: evidence from the Dutch Hunger Winter Families Study. J Dev Orig Health Dis 2009; 1: S164. Li Y, He Y, Qi L, et al. Exposure to the Chinese famine in early life and the risk of hyperglycemia and type 2 diabetes in adulthood. Diabetes 2010; 59: 2400–06. Thurner S, Klimek P, Szell M, et al. Quantification of excess risk for diabetes for those born in times of hunger, in an entire population of a nation, across a century. Proc Natl Acad Sci USA 2013; 110: 4703–07. Stanner SA, Bulmer K, Andres C, et al. Does malnutrition in utero determine diabetes and coronary heart disease in adulthood? Results from the Leningrad siege study, a cross sectional study. BMJ 1997; 315: 1342–48. Rudnytskyi O, Levchuk N, Wolowyna O, et al. Demography of a man-made human catastrophe: the case of massive famine in Ukraine 1932–1933. Can Stud Popul 2015; 42: 53–80. Carynnyk M, Luciuk LY, Kordan BS. The Foreign Office and the famine. British documents on Ukraine and the great famine of 1932–1933. Kingston, ON: The Limestone Press, 1988. Davies RW, Wheatcroft SG. The years of hunger: Soviet agriculture, 1931–1933. New York, NY: Palgrave Macmillan, 2010. http://www. europarl.europa.eu/sides/getDoc.do?type=TA&reference=P6-TA-20080523&language=EN&ring=P6-RC-2008-0571 (accessed Dec 10, 2013). Ukrainian Parliament. Law of Ukraine No 376–V. On the Holodomor of 1932–1933 in Ukraine. 2006. http://www.president. gov.ua/documents/5280.html (accessed Dec 10, 2013). European Parliament. Commemoration of the Holodomor, the artificial famine in Ukraine (1932–1933). 2008. Vallin J, Mesle F, Adamets S, Pyrozhkov S. A new estimate of Ukrainian population losses during the crises of the 1930s and 1940s. Popul Stud 2002; 56: 249–63. WHO. Definition, diagnosis, and classification of diabetes mellitus and its complications. Part 1: diagnosis and classification of diabetes mellitus. Geneva: World Health Organization, 1999. Rothman KJ, Greenland S, Lash TL. Modern epidemiology, 3rd edn. Philadelphia, PA: Wolters Kluwer/Lippincott Williams & Wilkins, 2008. Vaiserman AM, Khalangot MD, Carstensen B, et al. Seasonality of birth in adult type 2 diabetic patients in three Ukrainian regions. Diabetologia 2009; 52: 2665–67. Ekamper P, van Poppel F, Stein AD, Lumey LH. Independent and additive association of prenatal famine exposure and intermediary life conditions with adult mortality between age 18–63 years. Soc Sci Med 2014; 119: 232–39. Vaiserman AM, Khalangot ND, Pisaruk AV, et al. Predisposition to type II diabetes in Ukraine residents exposed to famine 1932–1933 during prenatal development. Adv Gerontol 2010; 23: 588–92 (in Russian). Vaiserman AM, Khalangot MD, Strele I, Lumey LH. Early-life exposure to the Ukraine famine of 1933 and type 2 diabetes in adulthood. In: Lumey LH, Vaiserman AM, eds. Early life nutrition and adult health and development. New York, NY: Nova Science Publishers, 2013. pp 145–60. Portha B, Fournier A, Ah Kioon MD, Mezger V, Movassat J. Early environmental factors, alteration of epigenetic marks and metabolic disease susceptibility. Biochimie 2014; 97: 1–15. Hochberg Z, Feil R, Constancia M, et al. Child health, developmental plasticity, and epigenetic programming. Endocr Rev 2011; 32: 159–224. Sandovici I, Hammerle CM, Ozanne SE, Constancia M. Developmental and environmental epigenetic programming of the endocrine pancreas: consequences for type 2 diabetes. Cell Mol Life Sci 2013; 70: 1575–95.
www.thelancet.com/diabetes-endocrinology Published online September 3, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00279-X
7
Articles
30
31
32
8
Volkmar M, Dedeurwaerder S, Cunha DA, et al. DNA methylation profiling identifies epigenetic dysregulation in pancreatic islets from type 2 diabetic patients. EMBO J 2012; 31: 1405–26. Tobi EW, Slieker RC, Stein AD, et al. Early gestation as the critical time-window for changes in the prenatal environment to affect the adult human blood methylome. Int J Epidemiol 2015; published online May 5. DOI:101093/ije/dyv043. Heijmans BT, Tobi EW, Stein AD, et al. Persistent epigenetic differences associated with prenatal exposure to famine in humans. Proc Natl Acad Sci USA 2008; 105: 17046–49.
33
34
Song S, Wang W, Hu P. Famine, death, and madness: schizophrenia in early adulthood after prenatal exposure to the Chinese Great Leap Forward Famine. Soc Sci Med 2009; 68: 1315–21. Vaag A, Brons C, Gillberg L, et al. Genetic, nongenetic and epigenetic risk determinants in developmental programming of type 2 diabetes. Acta Obstet Gynecol Scand 2014; 93: 1099–108.
www.thelancet.com/diabetes-endocrinology Published online September 3, 2015 http://dx.doi.org/10.1016/S2213-8587(15)00279-X