Area-Level Infant Mortality Exposure in Early Life and Alzheimer’s Disease Mortality: Examining Variation Based on Age,Sex,and Place of Birth

INTRODUCTION

Alzheimer’s disease and related dementias (ADRD) is one of the leading causes of death in the United States and has garnered increased attention in recent years [1–3]. ADRD is a progressive disease that is commonly believed to manifest decades before symptoms arise in individual. These symptoms typically come in the form of memory loss or problems with language, which can lead to a myriad of other issues such as mobility problems or lower quality of life as the disease progresses [4, 5]. Furthermore, given increased population aging in the United States, it is presumed that ADRD will have an increased prevalence, as there have been massive increases in longevity over time so that there will be more individuals that develop the disease [6, 7]. Therefore, identifying the specific pathways of later life ADRD is of great significance in order to minimize the burden that the condition imposes on our population health system.

Potential causes of ADRD are linked to various stages across the life course. However, researchers have increasingly sought to uncover early life factors. For instance, some studies have found associations between ADRD and childhood socioeconomic status [8], early-life advantage [9], and rural residence [10]. Developmental and origins of adult health and disease and life course frameworks stress the importance of critical periods, such as pre- or post-natal periods, in which an individual is particularly sensitive to adverse exposures. One famous example that examines these critical periods in early life was done by Barker (1986), who looked at area-level infant mortality rates (IMR) in the United Kingdom during the 1920 s and individual-level ischemic heart disease mortality in the 1970 s and found a positive association between the two [11]. Known as the “Barker hypothesis”, it posits that exposure to adverse conditions in the form of infant mortality during these critical periods increase one’s susceptibility to the development of disease in later life. This finding has been reaffirmed for heart diseases [12, 13] and other outcomes, such as cognition [14] and stroke [15], but not ADRD. Moreover, much of the setting of these studies has often been examined in the context of either the United Kingdom or Europe more broadly, but not the United States.

The linkage of adverse exposures during critical periods of development with later life ADRD is a challenging endeavor. This is due to the fact that ADRD, unlike some other forms of illness, is more concentrated towards the end of the life course, and thus greater attention is paid to potential determinants in mid and later life [16–19]. Indeed, age is often one of the most significant factors of ADRD, due to the fact that the majority of people who die from ADRD are over 65 years of age. Additionally, there are other factors outside of age that play a major role, such as genetics or family histories [20, 21]. For instance, a study by Han and colleages (2020) found a link between those with a higher polygenic risk for Alzheimer’s disease and rapid cognitive decline [22].

Outside of predetermined factors, modifiable risk factors can operate through various pathways, such as education, physical activity, and smoking status [23–25]. A study by Kivipelto and colleages (2005) found that mid-life obesity was associated with an increased risk of ADRD later in life, even when factoring in sociodemographic variables [26]. Other issues that have been identified include cardiovascular issues, such as hypertension and blood pressure, which reveal positive relationships with ADRD [27]. Diabetes, stroke prevalence, and place of death, too, in late-life have been shown to be associated with ADRD prevalence or death in later life, or poorer brain health, which sets the conditions for ADRD to potentially develop [28–31].

Another important consideration in examining adverse exposures early in the life course deals with the issue of mortality selection and sex. As previously mentioned, the majority of individuals who die from ADRD are over 65 years of age. However, the Barker hypothesis framework proposes that those exposed to worse environmental conditions may either die from ADRD at an earlier age than expected, or die from other causes of illness before they would live to see a diagnosis from the disease. In other words, it could be shown that there is a negative relationship in some instances, between infant mortality rates and later life ADRD. Thus, it is vital to investigate any potential relationship between adverse conditions in the form of IMR and ADRD by stratifying a sample between those above 65, and those below. Relatedly, it is critical to also look at the relationship between early life environments and ADRD by sex as well, given that evidence is mixed on which sex is more harmed by exposure to adversity [32, 33]. However, evidence from studies on ADRD find that women on average are more impacted by ADRD than men, and deteriorate from the condition at a quicker rate [34].

Taken together, despite all of the evidence laid out above concerning ADRD, there is scant research regarding how exposure to adverse conditions during critical periods of development may be associated with ADRD in later-life. One previous study examined this relationship and found that states with the highest rates of infant mortality was associated with an elevated dementia risk among African Americans, but not whites, though the studied individuals were drawn from California residents [35]. Apart from this example, there exists virtually no research on how early life exposure to IMR and ADRD mortality are associated. This, combined with the fact that the prospects for both the prevalence and mortality of ADRD is expected to increase [6], means that there is a fundamental need to examine this relationship. Thus, this paper asks the following questions: 1) is there an association between exposure to IMR in early life and later life ADRD mortality?; 2) are these associations different by age and sex group?; 3) how does place in the form of state of birth influence these effects?; and 4) what are the role of competing causes of death, if any?

MATERIAL AND METHODS

RESULTS

Table 1 presents descriptive statistics. Of the 397,794 individuals in our sample, 3,048 died from ADRD during the mortality follow-up period. The overall sample is 61.85% (246,046) male and 38.15% (151,748) female. Over 90% of the sample is non-Hispanic white, whereas those who are non-white account for 7.37% (29,334). The average age of the sample was 62.17 (SD = 5.38). For our independent variable of interest, infant mortality, the average rate was 56.19 deaths per 1,000 (SD = 13.58) with a range of 28.2 per 1,000 to 158.51 per 1,000; we standardize these values of IMR for our models. Regarding known risk-factors for ADRD, 14.30% (56,873), 9.22% (36,688), and 64.55% (248,201) of individuals had a self-reported history of heart disease, diabetes, and having ever smoked, respectively, and there was an average body mass index of 27.31 (SD = 4.84) at baseline interview. From this, our sample stratified by age and sex is presented in Table 2, where there are 144,958 and 101,088 males under 65 and those 65 and older, respectively. Females under 65 are composed of 94,644 individuals, whereas those 65 and older are 57,104.

Figure 1 shows the predicted probability for ADRD mortality, stratified by sex and age group. For all groups, IMR are associated with later life ADRD mortality from values of about 40 per 1,000 to 100 per 1,000. However, depending on the group, the effect is stronger at different levels of IMR, and then proceed to level off at higher values, which is consistent with a selection effect. For instance, the group of younger males sees a peak around 70 per 1,000, whereas for females 65 and older see a peak around 80 per 1,000, and a subsequent decline.

OPEN IN VIEWER

Fig. 1

Predicted Probability of ADRD Death, by Age and Sex Groups

Figure 2 then shows the same analysis, but controlling for state of birth, where we see similar trends, but only slightly. For full results for all models analyzed, Tables 3 and 4 show hazard ratios for ADRD mortality, stratified by sex and age. For the baseline findings, females see across age groups a positive association between infant mortality and later life ADRD. Likewise, males under the age of 65 see a positive association, but only one state of birth fixed effects are introduced.

OPEN IN VIEWER

Fig. 2

Predicted Probability of ADRD Death, by Age and Sex Group (State of Birth Fixed Effects)

Given the fact that the relationship between IMR and ADRD appeared to be non-linear and potentially affected by selection, we also opted to consider a competing risks approach to our analysis, to ensure the robustness of our findings. Thus, we opted to re-run the final model of analysis, but considering the competing morality risks of stroke, cardiovascular disease, and all other causes of death. We choose stroke and cardiovascular diseases given that they are major causes of death, and specifically are diseases that those suffering from ADRD tend to die from, if not from ADRD itself [43]. Even when factoring in these competing risks, our associations between early life IMR and remain relatively unchanged, with the size and the direction of the point estimates being quite similar (see Supplementary Table 1). For competing risks that include state of birth fixed effects, this is also the case (see Supplementary Table 2). For visualizations of these findings, see Supplementary Figures 1–6.

DISCUSSION

A great deal of research concerning mortality from ADRD has emphasized predetermined or modifiable risk factors specific to the individual. Likewise, much of the research on forms of health and mortality in recent years have focused on contextual factors, yet this focus has largely neglected ADRD [8, 42]. Additionally, when these contextual factors are accounted for, they tend to have a disproportionate focus on more contemporaneous factors compared to early life, likely a result of the fact that ADRD is something known to exclusively impact aging populations. Our study adds to the literature on early life factors by exploring how exposure to adverse socioeconomic and environmental conditions during critical periods of development, influence later life mortality from ADRD. We opt to use infant mortality as a proxy for these conditions, given that it has been used in the past to look into other forms of death or health conditions, initially by [11] and since then others [13, 14]. Ultimately, the findings presented throughout this research suggest several implications.

First, we further our understanding of how we may think about the how exposure to adverse conditions during critical periods of development. In this paper, we find that birth in areas of high infant mortality is associated with a higher likelihood of death from ADRD among those under 65 for females and for males under 65, when factoring in state of birth. This relationship is significant around 40 deaths per 1,000 and exists until about 100 deaths per 1,000, where it levels off. This relationship also exists for those 65 and over, though the relationship is somewhat weaker. Dying from ADRD before the age of 65 is classified as early-onset dementia, which is associated with atypical syndromes and has higher levels of mortality, and accounts for less than a tenth of all deaths from the illness [44]. Thus, given that we see an association between early-life infant mortality and later life ADRD mortality among younger individuals, particularly females, and at lower levels than one would expect, we argue that those exposed to worse adverse conditions during critical periods increase their likelihood of death from ADRD earlier than average, due to that exposure increasing their susceptibility to develop illness later on in life.

Second, we find in this paper that when factoring in controls for state of birth that the associations between IMR and later life ADRD mortality dissipates for females of both age groups, but reveals an association for males under 65 years of age. However, in looking at IMR at specific levels, it shows that infant mortality is still significant with later life ADRD, and follows a non-linear relationship, peaking around 60 or 70 deaths per 1,000 depending on the group, and declining thereafter. These findings imply that the IMR measure captures broader state-level factors, and failure to control for these may lead to misunderstandings of how specific pathways work to influence mortality across the life course. Given that different states often have different policy regimes that people are born into, and thus may have an impact through different mechanisms that may influence, it is always important to consider them as they could work to shape an individual’s health and mortality profile throughout the life course.

Third, many studies of how exposure to adverse socioeconomic and environmental conditions via infant mortality have primarily been done in the context of the United Kingdom or Europe more broadly. Our study adds to this literature in the sense that our setting is in the United States, where we find evidence that exposure to worse state-level infant mortality is linked with working-age ADRD death. This finding essentially establishes evidence for the Barker hypothesis in the United States, and should be replicated for other causes of death, to test its robustness.

There are some limitations to this study that are in need of being mentioned. One limitation is the timing and degree of exposures during the critical period. We use state-level IMR as a proxy for early life socioeconomic and environmental conditions. Given how larger units of geography may mask more heterogeneity, we have no way of knowing if the state average IMR was constant across all areas of a state. For instance, it could be that individuals living in one county had an IMR substantially lower than another, thus exposing them to a more benevolent environment. Future research should aim to consider finer units of geography (i.e., county or tract) to see if the relationships observed in this paper are robust to outcomes. Another limitation to this study is in regard to the sample composition. The data utilized in this study was primarily gathered from mostly white individuals across eight states in the United States, (California, Florida, Georgia, Louisiana, Michigan, New Jersey, North Carolina, and Pennsylvania). Given that illnesses such as Alzheimer’s disease and dementia are geographically clustered like other forms of illness, future research should consider information from individuals across all fifty states. However, despite the residence component, it is important to note that in our examination of the early life environment, this research does have information on individuals born across all fifty states.

These limitations notwithstanding, this study is novel in the sense that it allows us to explore the role of early life insults on later life ADRD mortality. The data source allows us to see if a disease that is known for predominantly impacting the oldest individuals in society is linked to events that occur earlier in life. Moreover, the data in this study allow us to consider the impact of mechanisms such as known-risk factors for ADRD, and if these differ by age and sex group. Ultimately, our findings suggest that exposure to adverse conditions in early life are related to ADRD mortality among those who die before 65, especially among females. Future studies would benefit from employing larger sample sizes, particularly for non-white populations, to see if these relationships are influenced dramatically by race. Furthermore, examinations of additional measures of early life exposures would be beneficial, such as pollution or lead. Unpacking which early life factors matter more regarding the risk for ADRD mortality would better help inform specific public health interventions in the United States, especially since the nation is bound to continue aging in the coming decades.