Lead Exposure and Alzheimer’s Disease Mortality Among US Adults: NHANES Analysis from 1988-2008
=================================================================================================

* Aaron Grossman

## Abstract

**Objective** To investigate the association between blood lead level (BLL) and Alzheimer’s disease (AD) mortality.

**Design** Prospective cohort study.

**Setting** US National Health and Nutrition Examination Survey (NHANES) 1988-94 and 1999-2008.

**Participants** 21,308 subjects aged 40 years and over with a BLL ≥1.0 µg/dL.

**Main outcome measures** AD mortality from baseline until December 2019.

**Results** AD was the underlying cause of death for 350 participants. In the fully adjusted model with a calendar effect variable, log-transformed BLL was inversely associated with AD mortality among all subjects (HR: 0.57 [95% CI, 0.46, 0.70]). A similar finding was evident when BLL was modeled as a categorical variable, with hazard ratios of 0.69 (95% CI: 0.54, 0.88), 0.56 (0.39, 0.80), 0.36 (0.19, 0.70), and 0.31 (0.14, 0.72) for 2.5-4.9 µg/dL, 5.0-7.4 µg/dL, 7.5-9.9 µg/dL, and ≥10 µg/dL, when compared to 1-2.4 µg/dL (*p*-trend < 0.0001). Restricted cubic spline analysis confirmed an inverse dose-response relationship between BLL and AD mortality.

**Conclusions** Our study suggests an inverse association between lead exposure and AD mortality among US adults; however, the results should be interpreted with caution due to the observational nature of the study.

**What is already known on this topic**

*   Studies investigating the association of blood lead and Alzheimer’s disease are rare.

*   Prior publications reported increased risk of Alzheimer’s disease mortality with exposure to lead, however, due to potential biases in these prior studies, it was critical to reassess the association between blood lead and Alzheimer’s disease mortality.

**What this study adds**

*   In contrast to much of the existing literature on dementia, the findings from this study suggest that lead exposure reduces the risk of Alzheimer’s disease mortality and likely mitigates the development of Alzheimer’s disease itself.

*   Prospective studies focused on blood lead and its association with the development or progression of Alzheimer’s disease are necessary to confirm the results.

## Introduction

Alzheimer’s disease (AD) is a devastating neurodegenerative disorder characterized by progressive and severe dementia1. Though the ultimate causes of AD pathogenesis remain elusive, metals have been a major focus of research. For instance, altered concentrations of metals have been detected within brain tissue, cerebrospinal fluid, blood serum, and urine of subjects with AD; however, the causal relationship of these metal abnormalities is not yet known2–7.

Lead (Pb) exposure is of particular interest due to its relatively recent decline in the built environment. In the 1970s, many developed countries initiated the phase-out of leaded gasoline as a consequence of its detrimental effects on children’s health, while the transition to unleaded gasoline in many developing nations did not begin until much later 8,9. Thus, lead exposure remains high in some of these developing regions 8,10. For instance, Nigeria first began its phase-out in 2002, and as late as 2014 leaded gasoline was still present at 0.49-1.90 mg/L 11. Consequently, mean blood lead level (BLL) in non-occupationally exposed adults in Nigeria in 2005 was found to be 33 µg/dl, whereas geometric mean BLL in US adults decreased from 1.68 µg/dL in 1999 to 0.82 µg/dL in 2016 12,13.

Intriguingly, investigations have found great disparities in AD prevalence among developed and developing nations. For example, a 1992 study from Ibadan, Nigeria found an AD prevalence of 1.41% among those aged ≥65 years, while a significantly higher rate of 6.24% was found among Blacks in Indianapolis in a corresponding study14. Similarly, a study in northern India from 1998 found an AD incidence of 4.7 per 1000 person-years versus a rate of 17.5 per 1000 person-years in Pennsylvania 15.

If lead were able to mitigate the risk of AD, its recent decline may also explain the increased the rate of AD within the US. It is common to think of AD as increasing in prevalence due to an aging population, however, aged-adjusted metrics have also been rising. For instance, according to data from the Centers for Disease Control and Prevention (CDC), age-standardized AD mortality almost doubled in the US from 18.1 deaths per 100,000 in 2000 to 32.4 deaths per 100,000 in 2020 16. Further, there is evidence that AD was increasing even before 2000. Among Medicare beneficiaries, age-standardized rates of AD increased by 55% between the periods 1984-1990 and 1991-200017. Moreover, this increase in AD prevalence has also been observed outside US borders 18,19.

Thus, it is conceivable that the steep decline in exposure to lead over the past few decades may have inadvertently increased risk of AD. In contrast to this hypothesis, a longitudinal investigation by Horton et al. (2019), which linked participants from 1999-2008 in the National Health and Nutrition Examination Survey (NHANES) study to the National Death Index (NDI), found BLL to be increased in those who eventually died from AD, though this finding was not statistically significant20. However, it is crucial to reexamine this study for potential bias and confounding. First, due to declining levels of lead exposure in the built environment, it is likely that subjects recruited for earlier NHANES cycles had higher overall BLL than those in subsequent cycles. Additionally, earlier cycles may have had a longer duration of follow-up, potentially introducing bias. Second, those with higher BLLs are typically older as well as tobacco smokers and would be expected to have a shorter lifespan than individuals of lower BLL21,22. Third, outcomes may lack relevance within time intervals proximal to the time of BLL assessment for two reasons: 1) those with early AD typically have several years of life expectancy ahead of them so that BLL proximate to time of death may not accurately depict the deeper causes of AD, and 2) individuals who are near death may exhibit higher BLL due to physiological processes that lead to increased BLL (i.e., reverse causality) 23. Fourth, a calendar effect may have developed wherein BLLs measured within different years, even if identical, may signify differences due to the ever-changing context of the population at large24. Finally, in conjunction with the latter observation, a previously unexamined calendar effect, namely, year of study exit, should be explored, to control for further decreases in BLL and changes in medical practice up to the time of end of follow-up.

Therefore, in the present study, we attempt investigate to mitigate some of these possible biases and reassess the association between lead and AD mortality at the individual level.

### 2.0 Methods

We conducted the following study in two phases. In Phase 1, we reproduced and modified the prior Continuous NHANES (1999-2008) study by Horton et al. examining the association of BLL and AD mortality. In Phase 2, we tested the robustness of the findings from Phase 1 by integrating the prior dataset with the older NHANES III (1988-1994) dataset.

NHANES is a series of cross-sectional studies of US adults and children conducted periodically between 1970 and 1994 and then each year since 1999 with data released in two-year cycles 25. The studies, conducted by the National Center for Health Statistics (NCHS), are designed to collect health data from a representative sample of the civilian noninstitutionalized population. Survey instruments, physical examination, and laboratory measurements of participants in NHANES have been described previously 26,27. Study protocols for NHANES were approved by the NCHS Ethics Review Board. All adult participants provided signed informed consent.

### 2.1 Phase 1, Continuous NHANES 1999-2008

To replicate the study by Horton et al., we obtained Continuous NHANES datasets for the five two-year cycles from 1999-2008 (hereafter, Cycles 1-5) 20. Rather than applying the 2014 based Linked Mortality Dataset that the authors originally used to link with National Death Index (NDI), we instead linked with updated datasets through 2019 which used an enhanced linkage algorithm 28. Subjects who were not linked to these death records were assumed alive at end of follow-up. The International Classification of Diseases, 10th Revision (ICD-10, 1999-present), was used to identify participants for whom AD was listed as the underlying cause of death. Subjects who died from causes other than AD were censored at the time of death.

We initially followed the study participant criteria by Horton et al. which included only subjects aged ≥60 years. Confounding variables included age, sex, race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, and other), and socioeconomic status (assigned by poverty-to-income ratio [PIR] <2, ≥2, or unknown). Smoking status was altered from the original study; participants were deemed current smokers if found to have a cotinine level ≥10 ng/ml or, if not available, by self-reporting as a recent smoker of cigarettes, pipes, or cigars. Blood samples collected during the NHANES mobile examination center visit had a lower limit of 0.3 µg/dL; those below this threshold assumed a BLL of 0.2 µg/dL. A continuous association between natural log-transformed BLL and AD mortality was tested and results were compared to the original study by Horten et al.

Due to the considerations mentioned earlier in terms of likely biases, we next employed modifications to the analysis. To limit the potential calendar effect produced by differing examination dates and to mitigate bias in duration of follow-up, analyses were stratified by cycle using the allotted follow-up periods. Adjustments to the selection of subjects as well as covariates were made in a stepwise fashion, as follows: 

*   We included only subjects with a BLL of ≥0.3 µg/dL.

*   Participants were aged ≥50 years at baseline.

*   We excluded subjects with a top-coded age of 85 years in Cycles 1-4 and 80 years in Cycle 5.

*   We considered covariates related to medical history that might impact mortality risk, including cardiovascular disease (CVD, defined as a history of heart attack, stroke, coronary heart disease, or congestive heart failure), cancer, hypertension, diabetes, and osteoporosis.

*   We excluded participants who died within 5.0 or 10.0 years from the date of examination, as deaths proximal to this date may confound the relationship of BLL.

*   We limited participants who were aged <75 years at baseline due to the increased potential for death to be multifactorial within later life.

*   As those who smoke have higher BLL and increased risk of death, we excluded subjects who were current smokers.

Finally, we reintegrated cycles and regressions analyses were evaluated with and without use of the calendar effect variable. The calendar effect variable adjusted for the final month of follow-up (due to death or end of follow-up period). This covariate simply added the difference between the midpoints of each cycle after the midpoint of Cycle 1 (i.e., 24, 48, 72, and 96 months for Cycles 2-5, respectively) to the total follow-up time (i.e., number of months from examination until end of follow-up). Thus, the overall calendar adjustment was greater in those surveyed in later cycles and in subjects with a longer follow-up duration. Analyses were again modified in a stepwise manner beginning with a base model of subjects aged ≥50 years and BLL ≥0.3 µg/dL. We employed sensitivity analyses by varying the inclusion criteria. These analyses involved excluding participants who died within 5.0-year and 10-year time-windows from examination date, excluding subjects who remained alive at end of follow-up, and examining AD cases only. These latter analyses were conducted for two distinct groups: 1) the entire study population meeting the specific inclusion criteria and 2) a subset of participants under 75 years of age.

### 2.2 Phase 2, Combined NHANES datasets (1988-2008)

We then integrated NHANES 1999-2008 with the NHANES III dataset. As with the prior study, participants not matched with a death record were considered alive through the end of the follow-up period. Coding for all deaths occurring prior to 1999 was in accordance with the 9th revision of the ICD (ICD-9), while coding for all deaths afterwards followed ICD-10 guidelines 28. Given the probabilistic nature of mortality ascertainment, participants whose follow-up extended beyond the age of 100 years were excluded from the analysis due to the low likelihood that these participants were correctly matched with death records. BLL measurements within NHANES III had a minimum level of detection of 1.0 µg/dL, thus all individuals in which BLL was undetected were excluded. Only nonpregnant participants aged ≥40 years were selected.

Study covariates included sex, race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, and other), education (<12 or ≥12 years), and PIR (<1.30, 1.30-3.49, ≥3.50, and missing). Current smokers were identified as above. In addition, relevant medical history which may be associated with earlier death were included. These comprised CVD, which was defined as a history of heart attack, stroke, or congestive heart failure. Additionally, for Continuous NHANES participants, a history of coronary heart disease was also considered. Other conditions assessed were cancer (excluding melanoma), hypertension, diabetes, and osteoporosis.

As above, a covariate for year of study exit was derived by adding the difference between the midpoint of each period after the midpoint of phase 1 (March 1990) of NHANES III (i.e., 24 and 42 months, for years 4 and 5-6 of NHANES III and 117, 141, 165, 189, and 213 months for Cycles 1-5 of Continuous NHANES, respectively) to the total follow-up time from examination29.

We evaluated the dose-effect between BLL and AD mortality by several means. First, we tested a continuous association using natural log-transformed BLL and examined the association of BLL and AD mortality by first stratifying for time of study exit based on 5-year intervals. In a sensitivity analysis, the study exit-time algorithm was applied solely to subjects who died during the follow-up period, while those who remained alive at the end of follow-up were assigned the final month of the follow-up period as their study exit time. Next, we evaluated a continuous association without stratification by year of study exit, but with and without the calendar effect covariate. As with Phases 1 and 2 above, three groups of participants were evaluated among: 1) all subjects and 2) subjects aged <75 years.

We also modeled BLL categorized by intervals (1.0-2.4, 2.5-4.9, 5.0-7.4, 7.5-9.9, and ≥10 µg/dL) across the five BLL categories and tested for trend. This was achieved by using a linear term consisting of the median values for each BLL group in place of the dummy variables for each group, and then analyzing the linear term using a Wald test.

Finally, we employed restricted cubic spline (RCS) analysis using five knots and a referent value set at the 12.5th percentile of BLL. Knots were placed at the 0.05, 0.275, 0.5, 0.725, and 0.95 percentiles in alignment with recommendations from Harrel30. Survival time was examined from the examination date. RCS analyses included: 1) all subjects, 2) exclusion of participants who died within a 10.0-year time-window from examination date, and 3) AD mortality cases only. All RCS analyses were conducted with and without a calendar effect covariate.

### 2.4 Statistical analysis

Descriptive statistics were provided for Continuous NHANES subjects, stratified by Cycles 1-5, and NHANES III and integrated datasets, stratified by BLL interval (1.0-2.4, 2.5-4.9, 5.0-7.4, 7.5-9.9, and ≥10 µg/dL). We computed means (standard deviations) for continuous variables and frequency (weights) for categorical variables; BLL, which had a right-skewed distribution, was reported with geometric mean and standard error. Hazard ratios (HRs) and 95% confidence intervals (CIs) of AD mortality were estimated. Categorical variables were compared using χ2 tests and the Kruskal-Wallis test for continuous variables.

All models were conducted using multivariate-adjusted Cox proportional hazard regression. Survival time was defined as the difference in months between the examination date and the end of follow-up. In all phases, tests of statistical significance were 2-tailed, and a probability value <0.05 was considered significant. Interactions between BLL and other sociodemographic and clinical characteristics were considered. The proportional hazards assumption was investigated by checking Martingale residuals. All analyses were conducted in SAS® OnDemand for Academics.

### Patient and public involvement

No participants were involved in setting the research questions or outcome measures, or in the design and implementation of the study. No plans exist to involve patients in dissemination.

## Results

### Phase 1, NHANES 1999-2008

Patient characteristics for the overall study population and stratified by NHANES Cycles 1-5 are found in Table 1. The participants included in this replication study did not differ from the original study by Horton et al. (n = 8080)20. However, due to the switching of the linked mortality files, in which the follow-up period was extended from 2014 to 2019, and consequent to the change in algorithm that links participants with the NDI, the status of participants at the end of the follow-up period was altered. We found AD deaths (n = 231, 2.86%) and deaths from all other causes (n = 3452, 42.72%) to be higher in comparison to the original study (n = 81 [1.00%] and n = 2227 [27.56%], respectively). BLL decreased from 2.51 µg/dL in the first cycle to 1.92 µg/dL in the final cycle, while the mean follow-up period also decreased from 9.9 to 4.3 years over the same timeframe. It is noteworthy that the maximum BLL among participants who died from AD and among all subjects was 8.4 µg/dL and 54 µg/dL, respectively. The number of AD deaths peaked within Cycle 2 participants (n = 35, 2.26%) and reached a nadir with Cycle 5 participants (n = 6, 0.31%).

View this table:
[Table 1.](http://medrxiv.org/content/early/2024/07/24/2024.07.20.24310751/T1)

Table 1. Demographic and clinical characteristics of NHANES study participants by cycle (1999-2008).

The base model analysis revealed an adjusted hazard ratio of 1.00 (95% CI: 0.78, 1.28) for the association between natural log-transformed BLL and AD mortality, which is lower than the hazard ratio reported in the original study (Supplement, Table S1). We then stratified analyses by cycle and modified the models in sequence (Table 2). Initially, Cycle 1 showed the greatest inverse association between BLL and AD mortality, Cycle 5, in contrast, showed the greatest positive association between these variables. The inclusion of other health covariates and younger participants into the analysis had marginal effects on the estimated hazard ratio. However, after excluding participants who died within 5.0 or 10.0 years of examination, hazard ratios exhibited an increasingly inverse association. Further, restricting subjects to non-smokers aged under 75 years resulted in even lower hazard ratios across all cycles.

View this table:
[Table 2.](http://medrxiv.org/content/early/2024/07/24/2024.07.20.24310751/T2)

Table 2. Comparison of HRs (95% CI) of natural log-transformed BLL predicting AD mortality by NHANES cycles (1999-2008).

Table 3 shows results after reintegration of all cycles. All models fitted with the calendar effect covariate displayed greater inverse risk in the association of BLL with AD mortality than models without the calendar effect. While the overall hazard ratio incorporating the calendar effect covariate was not statistically significant (HR: 0.84 [95% CI: 0.63, 1.10]), a significantly reduced hazard ratio was observed in the subset of subjects under 75 (HR: 0.56 [95% CI: 0.33, 0.93]). A significantly reduced risk of AD mortality was also observed in analyses which included only deceased subjects (without calendar effect: 0.70 [95% CI: 0.49, 1.00] vs. with calendar effect: 0.68 [95% CI: 0.48, 0.97]).

View this table:
[Table 3.](http://medrxiv.org/content/early/2024/07/24/2024.07.20.24310751/T3)

Table 3. HRs (95% CI) of natural log-transformed BLL predicting AD mortality among Continuous NHANES cycles (1999-2008).

### Phase 2, NHANES 1988-2008

Next, we combined NHANES III with Continuous NHANES datasets. Of 27,029 adults ≥40 years, we excluded 5281 subjects with BLL >1.0 μg/dl, 313 subjects without follow-up time from date of examination, and 127 subjects who were aged >100 years, resulting in 21,308 subjects eligible for analysis. Patient characteristics stratified by BLL category are presented in Table 4. Highlighting the skewed distribution of BLL, over 80% of subjects were found in the lowest BLL groups.

View this table:
[Table 4.](http://medrxiv.org/content/early/2024/07/24/2024.07.20.24310751/T4)

Table 4. Demographic and clinical characteristics of NHANES III (1988-1994) and Continuous NHANES (1999-2008) study participants

Table 5 exhibits the results of log-transformed BLL predicting AD mortality. Subjects were first stratified by time of study exit (i.e., due to death or end of follow-up period) in 5-year intervals. Participant BLL decreased with each 5-year interval, from 5.01 µg/dL in 1988-2002 to 2.36 µg/dL in the final two years of follow-up. Hazards generally decreased with each 5-year interval, with a high of 1.25 (95% CI: 0.45, 3.47) in 1993-1997 to a low of 0.54 (95% CI: 0.10, 3.01) in 2018-2019. In sensitivity analyses (Supplement, Table S2), in which all surviving subjects were assigned a final month of follow-up in 2019, the hazards were further reduced for years 2013-2017 (HR: 0.44; 95% CI: 0.30, 0.65) and 2018-2019 (HR: 0.51; 95% CI: 0.09, 2.85).

View this table:
[Table 5.](http://medrxiv.org/content/early/2024/07/24/2024.07.20.24310751/T5)

Table 5. Natural log-transformed BLL as continuous predictor of AD mortality in NHANES III (1988-1994) and NHANES 1999-2008 combined datasets

Among all subjects meeting inclusion criteria, and without adding a calendar effect to the model, risk was non-significantly decreased with increasing BLL (HR: 0.86, 95% CI: 0.70, 1.05). However, introducing a calendar effect variable led to a significant negative association (HR: 0.57, 95% CI: 0.46, 0.70). The sensitivity analysis incorporating the alternative algorithm for year of study exit yielded similar results (HR: 0.55, 95% CI: 0.44, 0.67). Excluding subjects <75 years further reduced the risk of AD (HR: 0.51, 95% CI, 0.39, 0.66). The addition of a 5- or 10-year time-window only marginally decreased risk further. Among AD cases only, inverse associations were also apparent both with (HR: 0.56, 95% CI: 0.44, 0.71) and without (HR: 0.78, 95% CI: 0.62, 0.98) the calendar effect covariate.

Table 6 displays results of the categorical analyses of BLL. In the fully adjusted model incorporating the calendar effect covariate, higher BLL categories exhibited an increasingly strong inverse association with AD mortality. Compared to the reference group with BLL between 1.0-2.5, the hazard ratios ranged from 0.69 (95% CI: 0.54, 0.88) for the 2.5-4.9 µg/dL group to as low as 0.31 (95% CI: 0.14, 0.72) for the group with BLL greater than 10 µg/dL (*P*-trend, <0.0001). Excluding participants who died within 5.0 or 10 years of examination date reduced risk of AD mortality, particularly within the 2.5-4.9 µg/dL group. A trend for decreased risk of mortality was also found among the subset of cases who died from AD (*P* = 0.002).

View this table:
[Table 6.](http://medrxiv.org/content/early/2024/07/24/2024.07.20.24310751/T6)

Table 6. Adjusted HRs (95% CI) of categorical BLL predicting AD mortality in NHANES III (1988-1994) and NHANES 1999-2008 combined datasets

Finally, we constructed restricted cubic spline models. Panels A and B illustrate the diminishing risk of AD mortality as BLL increases among all participants, both without and with the calendar effect covariate, respectively. Panels C and D show the same analysis, but with exclusion of subjects who less than 10 years of follow-up. Lastly, panels E and F depict the risk of AD mortality specifically among those subjects who eventually succumbed to the disease, again without and with the calendar effect covariate, respectively.

## Discussion

In the present study, we introduced several modifications to the previous Continuous NHANES (1999-2008) study by Horton et al. on AD mortality 20. These modifications included extending the follow-up period, incorporating a calendar effect covariate to account for year of study exit, and other adjustments, which led to findings that contested the prior study’s conclusions. In contrast to Horton et al., we found an inverse relationship between BLL and AD mortality. These findings not only contradict the previous NHANES study, but also challenge the prevailing discourse within the literature on the effect of lead on cognitive decline.

The protective feature of the calendar effect covariate observed across all phases of the study was instrumental in providing a more comprehensive context for interpreting BLL. Past environmental exposures from contaminated water, paint, and air pollutants can contribute to an increased burden of bone lead which over time can be released back into the bloodstream through demineralization 31,32. BLL depends on the balance between bone resorption and recent environmental exposures. It is conceivable that if two individuals obtained the same BLL estimate but were examined in different years, with varying mean BLL in the overall population, the estimates may reflect future differences. More specifically, the individual tested later, when mean BLL had already declined, is more likely to be exposed to higher levels of lead or may have higher bone deposits than the other individual, potentially leading to a comparatively elevated BLL throughout the follow-up period. Consistent with the findings from this study, such ongoing exposure would ultimately translate to decreased AD mortality. Alternatively, the protective feature of the calendar effect covariate may be attributable to advancements in medical practices and technology over time. Furthermore, due to the frailty of persons with AD, years in which infectious diseases were more rampant and severe could have influenced their time of death. Thus, the incorporation of the calendar effect covariate enabled us to account for sustained and intermittent temporal variations, not only across the NHANES survey periods but also throughout the entire follow-up period, thereby indirectly capturing the influence of other potential exposures and factors.

Due to AD’s slow progression, which may confound the association between BLL and short-term AD mortality, and due to the possibility of reverse causality, wherein BLL increases due to a physiological response to declining health (e.g., bone demineralization), we applied a time-window of varying length to filter out subjects who had brief follow-up periods. This technique proved more successful for Continuous NHANES than NHANES III datasets, possibly due to differing age structures of the participants, but may also be due to the rapid decline in BLL over the study years in the latter compared to that of the former. Similarly, the strength of the inverse association among the subset of participants aged less than 75 years was greater with Continuous NHANES than NHANES III, possibly because of the differing age structure of participants.

We examined if selecting participants based on their final status (i.e., all subjects, deceased only, or AD mortality only) affected the association between BLL and AD mortality. We found that by excluding those still alive at follow-up, the inverse association between BLL and risk of AD mortality was strengthened. The rationale behind conducting this analysis was to address the potential bias from having an extended duration of follow-up for patients surveyed earlier, who were likely to have higher BLL, than those surveyed later, with presumably lower BLL. The observation of delayed mortality with increasing BLL in analyses selecting only cases of AD mortality refutes the hypothesis that elevated BLL merely transfers the risk of death to another cause and does not have a genuine protective effect. However, we cannot verify whether the participants already had onset of disease at the time of examination to know with certainty if the disease process can be slowed or halted once underway.

The literature is particularly deficient in studies examining the effects of lead exposure on risk of AD. Other than the study by Horton et al., an investigation from Iran tested BLL among 27 subjects with AD and 54 controls. Mean BLL was found at higher levels in those with AD than controls (22.2 µg/dL vs. 7.9 µg/dL)20,33. However, this study had a significant confounder in that a large portion of each group was abusing opium which generally contains high levels of lead—similar to tobacco smokers in the current study. This is clearly demonstrated by the observation that opium abusers had substantially higher BLL (AD: 54.5 vs. controls: 7.2 µg/dL) than non-abusers (AD: 3.8 vs. controls: 5.8 µg/dL) 34.

Investigations in adults have showed varied results in terms of cognitive or neuropsychological symptoms from excess lead exposure. Studies within the occupational setting, where lead exposures may be extreme, have often found positive associations between BLL and cognitive effects 35,36. In contrast, studies from the non-occupational setting have typically not found such effects between BLL and cognition 37,38. In terms of non-blood matrices, Weisskopf et al. (2004) observed an inverse association between bone lead and decrements in cognition, while Prada et al. (2016) only found such an association in individuals with apolipoprotein E (*APOE*) ε4ε4 genotypes 37,39. It should be emphasized that while cognitive impairment may sometimes be a precursor of AD, this is not invariably so 40. In individuals exposed to extreme levels of lead, particularly in the occupational setting, cognitive impairment may arise from an alternative mechanism.

The protective effect of lead may be mediated by its capacity to outcompete other metals for transportation via the ubiquitous divalent metal transporter-1 (DMT-1) or a similar transporter. The hierarchical behavior of DMT-1 has been documented, with iron (Fe) being prioritized over other metals 41. If lead were to occupy a higher position in this hierarchy, it could potentially interfere with the transport of other metals implicated in AD. This intricate interplay between lead and other metals warrants further investigation to elucidate its role in AD pathogenesis.

This study has several limitations. First, though linkage of NHANES participants with the NDI has been enhanced since prior versions, type I and II errors remain a possibility. Second, AD is underreported due to only the leading cause of death being accessible on the linked mortality file, which may obscure instances in which AD was one of other attributable causes of death, often recorded on death certificates 42. This is particularly relevant when AD it still early in its progression43. Third, BLL is time-varying and was only captured at baseline (i.e. examination date), therefore pre- and post-examination BLL are unaccounted for. Fourth, there is the remote possibility that BLL is a surrogate for another correlated factor that is associated with AD mortality. However, these analyses were adjusted for several confounders, including socioeconomic and health indicators. It is also unlikely that lead correlated with other metal exposures across time, as lead exposure was continuously decreasing.

## Conclusion

From the present study we can conclude, based on a single blood lead measurement, that an inverse dose-response relationship likely exists between lead exposure and AD mortality in the U.S. adult population. Longitudinal clinical studies with multiple BLL measurements could shed more light on the progression of AD and lead’s putative role in mitigating risk. Though the implications of this study may be far-reaching, it would be prudent to apply caution due to the limitations associated with time-varying covariates in observational studies.

![Figure 1.](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2024/07/24/2024.07.20.24310751/F1.medium.gif)

[Figure 1.](http://medrxiv.org/content/early/2024/07/24/2024.07.20.24310751/F1)

Figure 1. 
Flow chart, Phase 2.

![Figure 2.](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2024/07/24/2024.07.20.24310751/F2.medium.gif)

[Figure 2.](http://medrxiv.org/content/early/2024/07/24/2024.07.20.24310751/F2)

Figure 2. 
Restricted cubic spline of the association of blood lead level and subsequent Alzheimer’s disease (AD) mortality, including lower 95% confidence interval.

Adjusted for age, gender, race, education levels, smoking status, medical history (osteoporosis, cardiovascular disease, hypertension, cancer, and diabetes), and calendar effect (i.e., year of end of follow-up) derived by adding the time from examination and the difference in time from midpoint of first cycle to midpoint of current cycle.

Panel A: All adults ≥40 years without adjustment for calendar effect. Panel B: All adults ≥40 years with adjustment for calendar effect.

Panel C: All adults ≥40 years without adjustment for calendar effect and 10-year time window of exclusion.

Panel D: All adults ≥40 years with adjustment for calendar effect and 10-year time window of exclusion.

Panel E: Adults ≥40 years with AD mortality without adjustment for calendar effect. Panel F: Adults ≥40 years with AD mortality with adjustment for calendar effect.

## Conflict of interest

None declared.

## Funding

This work did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

## Data availability

NHANES datasets are publicly available.

## Footnotes

*   **Conflict of interest:** None.

*   **Funding:** This work did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

*   Received July 20, 2024.
*   Revision received July 20, 2024.
*   Accepted July 24, 2024.


*   © 2024, Posted by Cold Spring Harbor Laboratory

This pre-print is available under a Creative Commons License (Attribution 4.0 International), CC BY 4.0, as described at [http://creativecommons.org/licenses/by/4.0/](http://creativecommons.org/licenses/by/4.0/)

## Work Cited

1.  1.Takahashi RH, Nagao T, Gouras GK. Plaque formation and the intraneuronal accumulation of β-amyloid in Alzheimer’s disease. Pathol Int. 2017;67:185–193.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1111/pin.12520&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=28261941&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F07%2F24%2F2024.07.20.24310751.atom) 

2.  2.Prakash A, Dhaliwal GK, Kumar P, Majeed ABA. Brain biometals and Alzheimer’s disease - boon or bane? Int J Neurosci. 2017;127:99–108.
    
    

3.  3.Tong Y, Yang H, Tian X, et al. High manganese, a risk for Alzheimer’s disease: high manganese induces amyloid-β related cognitive impairment. J Alzheimers Dis. 2014;42:865–878.
    
    

4.  4.Yang Y-W, Liou S-H, Hsueh Y-M, et al. Risk of Alzheimer’s disease with metal concentrations in whole blood and urine: A case-control study using propensity score matching. Toxicol Appl Pharmacol. 2018;356:8–14.
    
    

5.  5.Ehmann WD, Markesbery WR, Alauddin M, Hossain TI, Brubaker EH. Brain trace elements in Alzheimer’s disease. Neurotoxicology. 1986;7:195–206.
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=3714121&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F07%2F24%2F2024.07.20.24310751.atom) 

6.  6.Squitti R, Ghidoni R, Siotto M, et al. Value of serum nonceruloplasmin copper for prediction of mild cognitive impairment conversion to Alzheimer disease. Ann Neurol. 2014;75:574–580.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1002/ana.24136&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=24623259&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F07%2F24%2F2024.07.20.24310751.atom) 

7.  7.Gerhardsson L, Lundh T, Minthon L, Londos E. Metal concentrations in plasma and cerebrospinal fluid in patients with Alzheimer’s disease. Dement Geriatr Cogn Disord. 2008;25:508–515.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1159/000129365&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=18463412&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F07%2F24%2F2024.07.20.24310751.atom) 

8.  8.Ericson B, Hu H, Nash E, Ferraro G, Sinitsky J, Taylor MP. Blood lead levels in low-income and middle-income countries: a systematic review. The Lancet Planetary Health. 2021;5:e145–e153.
    
    

9.  9.Dignam T, Kaufmann RB, LeStourgeon L, Brown MJ. Control of Lead Sources in the United States, 1970-2017: Public Health Progress and Current Challenges to Eliminating Lead Exposure. J Public Health Manag Pract. 2019;25 Suppl 1, Lead Poisoning Prevention:S13–S22.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1097/PHH.0000000000000889&link_type=DOI) 

10. 10.Guo Y, Deng Y-H, Ke H-J, Wu J-L. Iron Status in Relation to Low-Level Lead Exposure in a Large Population of Children Aged 0-5 Years. Biol Trace Elem Res. 2021;199:1253– 1258.
    
    

11. 11.Jimoda LA, Olatunji SO, Adeniran JA, Fakinle BS, Sonibare JA. Atmospheric Loadings of Lead From Refined Petroleum Products Consumption in Southwestern Nigeria. Petroleum Science and Technology. 2014;32:2921–2929.
    
    

12. 12.Wang T, Zhou YP, Sun Y, Zheng YX. Trends in Blood Lead Levels in the U.S. From 1999 to 2016. American Journal of Preventive Medicine. 2021;60:e179–e187.
    
    

13. 13.Babalola OO, Ojo LO, Aderemi MO. Lead levels in some biological samples of auto-mechanics in Abeokuta, Nigeria. Indian J Biochem Biophys. 2005;42:401–403.
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=16955744&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F07%2F24%2F2024.07.20.24310751.atom) 

14. 14.Hendrie HC, Osuntokun BO, Hall KS, et al. Prevalence of Alzheimer’s disease and dementia in two communities: Nigerian Africans and African Americans. Am J Psychiatry. 1995;152:1485–1492.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1176/ajp.152.10.1485&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=7573588&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F07%2F24%2F2024.07.20.24310751.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=A1995RX73000012&link_type=ISI) 

15. 15.Chandra V, Pandav R, Dodge HH, et al. Incidence of Alzheimer’s disease in a rural community in India: the Indo-US study. Neurology. 2001;57:985–989.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1212/WNL.57.6.985&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=11571321&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F07%2F24%2F2024.07.20.24310751.atom) 

16. 16.Centers for Disease Control and Prevention, National Center for Health Statistics. National Vital Statistics System, Mortality 1999-2020 on CDC WONDER Online Database, released in 2021. Data are from the Multiple Cause of Death Files, 1999-2020, as compiled from data provided by the 57 vital statistics jurisdictions through the Vital Statistics Cooperative Program. [online]. Accessed at: [http://wonder.cdc.gov/ucd-icd10.html](http://wonder.cdc.gov/ucd-icd10.html). Accessed January 16, 2024.
    
    

17. 17.Ukraintseva S, Sloan F, Arbeev K, Yashin A. Increasing rates of dementia at time of declining mortality from stroke. Stroke. 2006;37:1155–1159.
    
    [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6OToic3Ryb2tlYWhhIjtzOjU6InJlc2lkIjtzOjk6IjM3LzUvMTE1NSI7czo0OiJhdG9tIjtzOjUwOiIvbWVkcnhpdi9lYXJseS8yMDI0LzA3LzI0LzIwMjQuMDcuMjAuMjQzMTA3NTEuYXRvbSI7fXM6ODoiZnJhZ21lbnQiO3M6MDoiIjt9) 

18. 18.Baek MS, Kim H-K, Han K, et al. Annual Trends in the Incidence and Prevalence of Alzheimer’s Disease in South Korea: A Nationwide Cohort Study. Front Neurol. 2022;13:883549.
    
    

19. 19.Chan KY, Wang W, Wu JJ, et al. Epidemiology of Alzheimer’s disease and other forms of dementia in China, 1990-2010: a systematic review and analysis. Lancet. 2013;381:2016–2023.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/S0140-6736(13)60221-4&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=23746902&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F07%2F24%2F2024.07.20.24310751.atom) 

20. 20.Horton CJ, Weng H-Y, Wells EM. Association between blood lead level and subsequent Alzheimer’s disease mortality. Environmental Epidemiology. 2019;3:e045.
    
    

21. 21.Jain RB. Trends and variability in blood lead concentrations among US adults aged 20–64 years and senior citizens aged ≥65 years. Environ Sci Pollut Res. 2016;23:14056–14067.
    
    

22. 22.Mannino D, Homa D, Matte T, Hernandez-Avila M. Active and passive smoking and blood lead levels in U.S. adults: Data from the Third National Health and Nutrition Examination Survey. Nicotine & Tobacco Research. 2005;7:557–564.
    
    

23. 23.Dodge HH, Shen C, Pandav R, DeKosky ST, Ganguli M. Functional Transitions and Active Life Expectancy Associated With Alzheimer Disease. Arch Neurol. 2003;60:253.
    
    

24. 24.Griffin BA, Anderson GL, Shih RA, Whitsel EA. Use of alternative time scales in Cox proportional hazard models: implications for time-varying environmental exposures. Stat Med. 2012;31:3320–3327.
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=22531976&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F07%2F24%2F2024.07.20.24310751.atom) 

25. 25.National Health and Nutrition Examination Survey [online]. Accessed at: [https://www.cdc.gov/nchs/nhanes/index.htm](https://www.cdc.gov/nchs/nhanes/index.htm).
    
    

26. 26.Ezzati TM, Massey JT, Waksberg J, Chu A, Maurer KR. Sample design: Third National Health and Nutrition Examination Survey. National Center for Health Statistics. Vital Health Stat 2(113) [online]. Vital Health Stat 2(113).; 1992. Accessed at: [https://www.cdc.gov/nchs/data/series/sr\_02/sr02_113.pdf](https://www.cdc.gov/nchs/data/series/sr_02/sr02_113.pdf).
    
    

27. 27.Johnson CL, Paulose-Ram R, Ogden CL, Carroll MD, Kruszon-Moran D, Dohrmann SM, Curtin LR. National Health and Nutrition Examination Survey: Analytic guidelines, 1999– 2010 [online]. National Center for Health Statistics. Vital Health Stat 2(161).; 2013. Accessed at: [https://www.cdc.gov/nchs/data/series/sr\_02/sr02_161.pdf](https://www.cdc.gov/nchs/data/series/sr_02/sr02_161.pdf).
    
    

28. 28.Public-use Linked Mortality Files [online]. National Center for Health Statistics; 2022. Accessed at: [https://www.cdc.gov/nchs/data/datalinkage/public-use-linked-mortality-file-description.pdf](https://www.cdc.gov/nchs/data/datalinkage/public-use-linked-mortality-file-description.pdf). Accessed January 16, 2024.
    
    

29. 29.National Center for Health Statistics, Centers for Disease Control and Prevention. ANALYTIC AND REPORTING GUIDELINES: The Third National Health and Nutrition Examination Survey, NHANES III (1988-94) [online]. 1996. Accessed at: [https://wwwn.cdc.gov/nchs/data/nhanes/analyticguidelines/88-94-analytic-reporting-guidelines.pdf](https://wwwn.cdc.gov/nchs/data/nhanes/analyticguidelines/88-94-analytic-reporting-guidelines.pdf).
    
    

30. 30.Harrell FE. Regression modeling strategies: with applications to linear models, logistic and ordinal regression, and survival analysis. Second edition. Cham Heidelberg New York: Springer; 2015.
    
    

31. 31.Weyermann M, Brenner H. Factors affecting bone demineralization and blood lead levels of postmenopausal women--a population-based study from Germany. Environ Res. 1998;76:19–25.
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=9466893&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F07%2F24%2F2024.07.20.24310751.atom) 

32. 32.Landrigan PJ, Todd AC. Lead poisoning. West J Med. 1994;161:153–159.
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=7941534&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F07%2F24%2F2024.07.20.24310751.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=A1994PC71000005&link_type=ISI) 

33. 33.Fathabadi B, Dehghanifiroozabadi M, Aaseth J, et al. Comparison of Blood Lead Levels in Patients With Alzheimer’s Disease and Healthy People. Am J Alzheimers Dis Other Demen. 2018;33:541–547.
    
    

34. 34.Sinai A, Mokrysz C, Bernal J, et al. Predictors of Age of Diagnosis and Survival of Alzheimer’s Disease in Down Syndrome. J Alzheimers Dis. 2018;61:717–728.
    
    

35. 35.Fenga C, Gangemi S, Alibrandi A, Costa C, Micali E. Relationship between lead exposure and mild cognitive impairment. J Prev Med Hyg. 2016;57:E205–E210.
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=28167858&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F07%2F24%2F2024.07.20.24310751.atom) 

36. 36.Lindgren KN, Masten VL, Ford DP, Bleecker ML. Relation of cumulative exposure to inorganic lead and neuropsychological test performance. Occup Environ Med. 1996;53:472–477.
    
    [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6NToib2VtZWQiO3M6NToicmVzaWQiO3M6ODoiNTMvNy80NzIiO3M6NDoiYXRvbSI7czo1MDoiL21lZHJ4aXYvZWFybHkvMjAyNC8wNy8yNC8yMDI0LjA3LjIwLjI0MzEwNzUxLmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ==) 

37. 37.Weisskopf MG, Wright RO, Schwartz J, et al. Cumulative Lead Exposure and Prospective Change in Cognition among Elderly Men: The VA Normative Aging Study. American Journal of Epidemiology. 2004;160:1184–1193.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1093/aje/kwh333&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=15583371&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F07%2F24%2F2024.07.20.24310751.atom) 

38. 38.Nordberg M, Winblad B, Fratiglioni L, Basun H. Lead concentrations in elderly urban people related to blood pressure and mental performance: results from a population-based study. Am J Ind Med. 2000;38:290–294.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1002/1097-0274(200009)38:3<290::AID-AJIM7>3.0.CO;2-T&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=10940966&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F07%2F24%2F2024.07.20.24310751.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000088584300007&link_type=ISI) 

39. 39.Prada D, Colicino E, Power MC, et al. APOE ε4 allele modifies the association of lead exposure with age-related cognitive decline in older individuals. Environ Res. 2016;151:101–105.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/j.envres.2016.07.034&link_type=DOI) 

40. 40.Montine TJ, Bukhari SA, White LR. Cognitive Impairment in Older Adults and Therapeutic Strategies. Hook V, editor. Pharmacol Rev. 2021;73:152–162.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1124/pharmrev.120.000031&link_type=DOI) 

41. 41.Shawki A, Anthony SR, Nose Y, et al. Intestinal DMT1 is critical for iron absorption in the mouse but is not required for the absorption of copper or manganese. Am J Physiol Gastrointest Liver Physiol. 2015;309:G635–647.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1152/ajpgi.00160.2015&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=26294671&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F07%2F24%2F2024.07.20.24310751.atom) 

42. 42.Stokes AC, Weiss J, Lundberg DJ, et al. Estimates of the Association of Dementia With US Mortality Levels Using Linked Survey and Mortality Records. JAMA Neurol. 2020;77:1543–1550.
    
    

43. 43.Hoyert DL, Rosenberg HM. Alzheimer’s disease as a cause of death in the United States. Public Health Rep. 1997;112:497–505.