The developmental origins hypothesis proposes that under-nutrition in early life is associated with an increased risk of disease in adulthood. Famine studies have provided a number of evidence supporting such a phenomena [1]. The majority of previous famine studies were conducted in developed countries [2, 3] and findings were not always consistent with those from developing countries. For example, positive associations of early life famine exposure with metabolic syndrome [4, 5] and hypertension [6, 7] were found in China and Biafran but not in a Dutch famine study [8].

The Chinese famine of 1959–1961 is the largest one in human history leading to approximately 30 million excess deaths [9, 10]. Emerging findings suggested that exposure to Chinese famine in early life was related to elevated risk of diabetes [11], metabolic syndrome [4, 5], hypertension [7], short height [12], and overweight [13] in adulthood. Anaemia is a common health problem especially in developing countries [14], with iron deficiency as one of the main causes [14]. Although the prevalence of anaemia has decreased dramatically in China, it is still as high as 15.2% on average [15]. Surprisingly, even in some economically well developed regions, the prevalence of anaemia was above 20% [16] and anaemia often coexists with metabolic syndrome in China [17]. Emerging data suggested that intrauterine exposure to malnutrition increased the risk of anaemia in children [18, 19]. However, its implication on anaemia in adulthood has not been investigated and little is known about its long term effect. Investigation of the long-term risk for anemia associated with the Chinese famine will provide a unique opportunity to evaluate fetal origin of adulthood anaemia. Based on data of adults born between 1954 and 1964 in Jiangsu province from the 2002 Chinese National Nutrition and Health Survey, we aimed to investigate whether early life exposure to famine is related to a higher risk of anaemia in adulthood.

In stratified analyses, in general, the elevated risk of adulthood anaemia associated with fetal exposure to famine persisted across different stratum of other risk factors of anaemia, such as overweight status (yes/no), hypertension (yes/no), smoking (yes/no), income (low/medium/high), and education (low/medium/high) although it became statistically insignificant in some strata due to the reduced sample size. The association seemed strongest among those with low education or overweight/obese.

Adjusted RR (95% CI) of fetal-exposed to famine in association with anaemia in adulthood was 1.75 (1.20-2.56) for overweight or obese individuals and 1.17 (0.88-1.57) for others; 2.07 (1.37-3.12) for those of low education vs 1.34 (0.93-1.93) for others.

When we limit our analyses to the rural sample, the RRs for anemia were 1.25(0.96-1.61), 1.05(0.80-1.38), 1.18(0.91-1.52), 1.49(1.13-1.96), and 1.00 for late childhood, mid childhood, early childhood, fetal-exposed and non-exposed cohorts, respectively. There was significant interaction between residence (urban/rural) and famine exposure (p=0.023). Among urban residents, there was no significant association between famine exposure and anemia (data not shown). In fully adjusted regression model using the rural sample, the regression coefficients for Hb were -0.22 (-0.48 to 0.03), -0.03(-0.28 to 0.23), -0.07(-0.32 to 0.18), -0.34(-0.63 to -0.04) and 0.00 for late childhood, mid childhood, early childhood, fetal-exposed and non-exposed cohorts, respectively.

We found that fetal exposure to the Chinese famine was associated with an increased risk of anaemia especially among men, those currently overweight or obese or with lower educational levels. Moreover, coexistence of anaemia and overweight/hypertension was more common among those exposed to famine than those non-exposed.

We are unaware of previous studies assessing early life exposure to famine and anaemia in adulthood. Several factors related to body iron levels and store may be implicated in the observed association between famine exposure in early life and adulthood anaemia, for instance dietary iron intakes. However, in a subsample with data on dietary iron intakes available, we observed that there is no significant difference in iron intake across different famine groups. Therefore, the difference in the prevalence of anemia between famine exposure groups and non-exposed cannot be explained by differences in iron intake. Similar phenomenon was found in a study in Mexico [28], which found iron deficiency in obese women and children were predicted by obesity-related inflammation rather than by differences in dietary iron intake. During famine, it would be expected that the level of stress will be high. In an animal study it was found that maternal stress during pregnancy predisposes for iron deficiency in infant monkeys [29]. In a recent study of 140 pregnant women in Israel, it was found that prenatal maternal stress (under rocket attack during a military operation) predicts lower cord ferritin concentration [30]. However, whether this effect can last until adulthood is not known.

The association between fetal-exposed to famine and anaemia appeared to be more evident among men than women. The underlying mechanisms are unclear. On one hand, famine exposure was related to a lower sex ratio leading to more female babies [31], suggesting women may be more adaptable to famine. On the other hand, menstruation, specific to women, is an important factor regulating haemoglobin levels. As we do not have information on menstruation and menopasue, we are not able to control the effect of blood loss due to menstruation. It has been reported that there was a positive association between testosterone and haemoglobin levels among men [32]. Whether a low testosterone level mediates the association between famine exposure and anemia among men warrants further investigation. In addition to differential effect by gender, in the present study, the association of fetal-exposed to famine with adulthood anaemia appeared strongest among the low socioeconomic status (SES). It is plausible that those in the low SES group suffered most from famine due to poor access to other resources. Those exposed prenatally will also have been exposed in early life since the famine lasted for 3 years rather than 9 months. The effects of exposure in pregnancy only are thus difficult to assess.

During famine, there was a significant drop in fertility. In the study area, the total fertility loss during famine was 128.3% [33]. Those who chose to give birth during famine would be those with the capacity to cope with the famine. In the study area, the excess crude death rate was 11.375 per 1000 during the famine period [33]. Because of the long duration of the famine, children born during the famine would suffer from both prenatal and postnatal under-nutrition and lead to both anemia and high mortality. Children with worse health condition would be more likely to die from famine than those with better health. Thus, both selective fertility and selective mortality would underestimate the true effect of famine in the population.

Comorbidity of anaemia and being overweight was common in famine exposed groups in this population. Among those who are anemic, the prevalence of overweight differs between fetal exposure and non-exposed cohorts (48.2% vs 33.3%). It is possible that anaemia secondary to chronic diseases is a problem in this population. Due to the lack of biomarkers of chronic diseases such as inflammation, we are not able to further explore this hypothesis. Findings from our previous studies on the association between diet and anaemia indicated that inflammation may play a role in the aetiology of anaemia in the population [26]. Future study should assess whether famine exposure affect biomarkers of inflammation such as C-reactive protein.

The study has several strengths. Firstly, it is based on a large sample composed of both urban and rural populations following standard manual of procedures. Anthropometric measurements were performed by health workers during a household visit according to a standard protocol. Other major characteristics of study participants were collected based on in person interview according to a standard protocol, which may reduce the measurement errors. Secondly, in a subsample, we have detailed information on dietary intakes. Several limitations of the study merit discussion. First we do not have information on iron metabolism biomarkers, and some other factors that may affect body iron store and levels such as menstruation status in women that are related to the development of anaemia. Secondly, regional difference in the severity of famine is highly likely in the study population, as the neighbouring province, Anhui, had an excess death rate of 474.9% during the famine period [10]. We are not able to assess the effect of the severity of famine in association with anemia due to small number of participants in the fetal- exposed group. In sensitivity analysis, when we subdivided the sample into subgroups, the power of detecting a significant association in some subgroups may be limited given the relatively small number in certain subgroups. The findings from stratified analyses are suggestive not conclusive. Selection bias is possible due to selective mortality or reduced fertility across famine exposure groups especially in the fetal exposed group.

In conclusion, we observed that early life exposure to the Chinese famine was associated with an increased risk of anaemia particularly among men, those currently overweight or obese, or with low education levels. These findings suggested that both the early life under nutrition and later life over nutrition are related to health condition in adulthood. The underlying molecular mechanisms warrant future investigation.