Derivation of an allostatic load index in the Chilean GERO study cohort reveals the need for specific clinical thresholds in older populations

doi:10.21203/rs.3.rs-4719120/v1

Download PDF

Research Article

Derivation of an allostatic load index in the Chilean GERO study cohort reveals the need for specific clinical thresholds in older populations

https://doi.org/10.21203/rs.3.rs-4719120/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background

Only few studies of Allostatic Load (AL) have been conducted in Latin/Hispanic populations, and even fewer performed in South American cohorts. Here, we replicated a recently proposed algorithm to derive an AL index in a Chilean cohort of cognitive healthy adults.

Methods

Using data from the GERO cohort (n = 166, age 76.92 ± 5.07 years, 80.12% females) we generated an AL risk index through a comprehensive algorithm (ALCS). Due to the age and the large disparity between males and females of the sample, modifications to the original algorithm were introduced. An empirical algorithm (ALES), without including clinical thresholds was also derived for model comparison.

Results

Neither ALCS nor ALES achieved adequate goodness of fit. After age and sex corrections, both algorithms largely improved, but only ALES showed satisfactory model fit. ALCS largely improved when correcting for available age-specific clinical thresholds, although failed to achieve sufficient goodness of fit. A final inclusion of an additional inflammatory biomarker resulted in further improvement of ALES, but not on ALCS.

Conclusions

Even when satisfactory model fit was not achieved by the comprehensive scoring, it was largely improved when some clinical values were corrected for ages ≥65. The ALES algorithm, which do not include clinical thresholds, achieved satisfactory fit after correcting for sample disparities between sexes. This suggests the need to consider the suitability of clinical thresholds developed in younger populations and the need to develop specific thresholds for assessing older adults.

Trial registration

Geroscience Center for Brain Health and Metabolism Research Project Nº 1140423, approved in May 2015 by the Scientific Ethics Committee (SEC) of the Western Metropolitan Health Service (WMHS), Santiago (Chile). Trial Number NCT04265482.

Allostatic load

Scoring algorithms

Latin/Hispanic cohort

Age-specific clinical thresholds

Older populations

Allostatic Load (AL) describes the wear and tear resulting from physiological responses caused by stress exposure [1]. It reflects a process of maintaining homeostasis upon environmental demands, through transient changes in the basal activity of several physiological systems [1, 2]. Nonetheless, if the response is sustained through extensive time, the organism’s resources will deplete and damage the multiple systems involved. To date, evidence shows positive relationship between AL and risk of developing several pathological conditions, such as major depression, brain structure deterioration, cardiovascular disease, diabetes type 2, cancer, functional decline, and dementia [3]. Although evidence reports important ethnic modulatory effect on AL [4], most studies have been performed in white-caucasian cohorts from high-income countries, and only few reports can be found in Latin/Hispanic populations, usually among Mexican or Central-American immigrants within the US [5–8], with very little evidence from South American cohorts [9–11].

Initially, AL indexes included physiological biomarkers of neuroendocrine, metabolic, immune, and cardiovascular activity, classifying them as primary mediators, released as the first response of the hypothalamic-pituitary axis (HPA) upon stress exposure; and the secondary outcomes, activated by the primary mediators, such as glycemia, blood pressure or cholesterol levels [12]. Novel approaches have excluded the primary neuroendocrine mediators, and there is a broad consensus that at least one biomarker for the cardiovascular, metabolic, and inflammatory components must be included for AL estimation [13]. However, there is still no clear agreement regarding the algorithm to calculate the AL index, with up to 18 different equations described so far [13], although several studies favour more comprehensive approaches that include prescribed medications, clinical thresholds and sex differences among biomarkers [3, 14–17].

In a recent study we proposed a comprehensive composite to calculate an AL risk index a British cohort [18]. The algorithm included the high-quartile approach derived from sex-specific sample distribution, clinical reference values specific for age and sex when available, and current medications treatments that could potentially mask the values of some biomarkers. The algorithm showed better model fit criteria when compared to the purely empirical method and revealed strong associations with structural brain deterioration in a mid-life healthy population. Here, we aimed to derive an AL scoring using the comprehensive approach in non-demented, middle and older old participants from the Chilean GERO cohort [19]. This cohort is composed by an heterogenous group of Chilean participants, with a social context different from most published cohorts. This makes it attractive to study stressor factors, stress response and the mechanisms underlying dementia development in a different environmental and social context from those mostly analysed to date, especially as evidence shows that ethnicity and even geographical diversity can affect certain biomarkers used to estimate AL [4, 15, 20, 21]. Accordingly, given the GERO population is older than the usually reported in previous AL studies and high cardiovascular and obesity risk is reported for Chile by recent OECD publications [22], modifications had to be performed to improve the original algorithm.

Participants and Ethics

Data from the first assessment (T1) from the GERO cohort was used [19]. Participants were recruited between October 2017 and July 2021 from the general population through a door-to-door strategy across three districts in Santiago – Chile (Macul, Peñalolen and La Reina), assigned to three primary healthcare centres and selected according to their socioeconomic heterogeneity. Eligible participants were ≥70 years old, without dementia diagnosis and cognitive complaint either self-reported or by a close informant. Eligible participants underwent a first evaluation to confirm inclusion criteria after signing a written informed consent. The Geroscience Center for Brain Health and Metabolism Research Project Nº 1140423 and informed consents, was approved in May 2015 by the Scientific Ethics Committee (SEC) of the Western Metropolitan Health Service (WMHS), Santiago (Chile). A written informed consent to participate in the study is obtained for all participants of the GERO cohort. The study is registered under the clinical trial Number: NCT04265482 (https://clinicaltrials.gov/study/NCT04265482).

Biomarkers collection

Fourteen biomarkers were used to derive an AL index. The inflammatory component included Creatinine, albumin, C-reactive protein (CRP) and Erythrocyte Sedimentation Rate (ESR). The latter was included as a reliable proxy measure to replace serum fibrinogen [23] used in the original composite. Cardiovascular included systolic blood pressure (SBP), diastolic blood pressure (DBP), resting heart rate (RHR), and waist-to-hip ratio (WHR), and metabolic system included total cholesterol, high-density lipoprotein (HDL) cholesterol, low-density lipoprotein (LDL) cholesterol, glycemia, triglycerides, and body-mass index (BMI). Interlukin-6 (IL6), also commonly used for AL score creation [11, 24], was selected as additional inflammatory biomarker for potential model corrections. Samples were taken between 9 A.M. and 11 A.M. and peripheral blood was processed within 2h. Whole blood, buffy coat, plasma, and serum were collected in a fasted state at baseline evaluation and processed according to the guidelines published previously [25]. Subsequently, blood samples were stored in the GERO biobank for long-term storage at − 80°C at the Faculty of Medicine of the University of Chile. IL-6 and CRP inflammatory biomarkers were analysed using Luminex platform at the University of Chile (see full details in Slachevsky, 2020 [19]).

Original and corrected comprehensive AL algorithm (ALCS)

166 participants from a total of 291 selected for inclusion to the GERO cohort, with participants excluded if they had missing data in the variables used to create the AL index. (n = 125). All biomarkers were scored to create a comprehensive AL index (ALCS), as previously described[18]. Initial categories for “no-risk” (zero points), “at-risk” (one point), and “high-risk” (two points) were defined for each marker, based on both clinical reference values provided by Hospital Salvador (Santiago, Chile) and quartile calculations based on sex-specific sample distribution. When clinical upper limit (clinical-up) was higher than the 75th percentile, at-risk category was defined between ≥ p75 - ≤ clinical-up (no-risk: <p75 and high-risk: >clinical-up). When clinical upper limit was lower than p75, at-risk was defined between ≥ clinical-up - ≤p75 (no-risk: <clinical-up, high-risk: >p75). For reverse biomarkers (albumin, HDL cholesterol), if clinical lower limit (clinical-low) was below the 25th percentile (p25), at-risk was defined as ≤ clinical-low - ≥p25 (no-risk: >p25 and high-risk: <clinical-low).

Medication treatments coded through the Anatomic Therapeutic Chemical (ATC) classification system [26] were scored as high-risk (two points) as could potentially mask some biomarkers values, as follows: total cholesterol, triglycerides and LDL for lipid modifying agents (C10); systolic and diastolic blood pressure for anti-hypertensive medication (C02, C03, C09); resting heart rate for beta-blockers (C07) or calcium blockers (C08); and glycemia for insulin or analogues (A10).

After summing the scores and given only three female participants scored 0 points, percentiles of total scores – including zeros - were calculated from sex-specific distributions, to generate final AL risk categories: Very-low risk: 0 - <p5; Low risk: ≥p5 - ≤p25; Medium risk: >p25 - <p75; High risk: ≥p75. The decision algorithm is detailed in Supplementary Fig. S1a.

Given the age of the sample, age-corrected clinical thresholds were included for ESR [27], CRP [28] and age and gender categories for RHR provided by the British Cardiovascular Society. High and low quartiles were calculated from the whole sample distribution of total scores to generate final AL risk categories. Clinical thresholds and quartiles values are detailed in Supplementary Table S1.

Original and corrected empirical AL algorithm (ALES)

Following the empirical approach [29], a purely quartile-based AL index was derived from sex-specific distributions. Biomarkers were awarded 1 point if their value was ≥ p75, or ≤ p25 for albumin and HDL cholesterol. After scoring for medications, AL risk categories were generated from total scores by the same method used in ALCS. Corrected ALES final risk categories were generated by calculating percentiles from the whole sample distribution of total scores. The decision algorithm is detailed in Supplementary Fig. S1b.

Covariates

Demographic covariates included age, sex, and years of education. Household income was assessed as self-reported estimated total income in Chilean pesos from all members of the household, categorized from 65,000 to over 1,650,000 in variable ranges of 50,000, 100,000 or 200,000, according to criteria from the National Health Survey. For APOEε4 genotype status, DNA was extracted using the protocol Q suggested by the international human microbiome standards (IHMS SOP 06 V1), and binary categories were generated where 0 = no APOEε4 allele and 1 = one or both APOEε4 alleles. As additional modulator of AL, self-reported frequency of engagement in stimulating activities was assessed through a 15-item reverse-coded Likert scale (see details in Supplementary Table S2), with points ranging from 1 to 5 for “several times a week” to “never” (Thumala, et al. Article in preparation). Activity types were sub-divided as cognitive, social, recreational, physical, and productive.

Statistical analysis

All analysis were conducted using IBM SPSS Statistics v.27.0 [30]. Statistical differences between sex for age, years of education and original AL raw scores were evaluated by two-tailed unpaired t-test when normality and homoscedasticity were above the rejection value of < 0.05 (estimated by Shapiro–Wilk and Levene’s tests respectively). Otherwise, Mann Whitney-U rank sum test was used for non-parametric unpaired two-sample comparison.

For validation, original, corrected and IL6-included ALCS and ALES algorithms were analysed separately, including age, sex, and years of education as covariates in Multinomial Logistic Regression (MLR) models, with the very-low risk category used as reference. Algorithms showing satisfactory model fit (Likelihood ratio χ² tests with α = 0.05) were further assessed for final selection by highest Mc Fadden’s pseudo-R² [29], the Bayesian Information Criterion difference (ΔBIC)>-10 [31], and highest log-likelihood parameters. Classification tables were used to compare the accuracy of correct classification into each AL category, and differences were assessed through pooled probabilities and z-transformations. Inter-rater reliability was evaluated through contingency tables and Cohen’s kappa index. After model validation, a MLR was fitted to assess relationships between AL risk categories generated from the selected algorithm and age, sex, years of education, household income, APOEε4 genotype status, and engagement in cognitive, social, recreational, physical, and productive stimulating activities.

Participants

The analytical sample included 166 participants from the GERO cohort (80.12% females), with an average age of 76.92 (SD = 5.07) years old (females: 77.04, SD = 5.27), and a mean of 9.02 (SD = 4.79) years of education (females: 8.79, SD = 4.66). No significant differences were found by sex in age (Mann-Whitney U test: U = 2265.5) or education (Mann-Whitney U test: U = 1957, p = 0.33, full statistical results detailed in Supplementary Table S3). Demographic characteristics of the GERO participants included the analyses are detailed in Table 1.

Table 1

Descriptive Statistics of the GERO cohort (n = 166)a
Variable [missing]		Total	Females
Sex, n (%)		33 (19.88%)	133 (80.12%)
Age, mean (SD)		76.92 (5.07)	77.04 (5.27)
Years of education, mean (SD) [1]		9.02 (4.79)	8.79 (4.66)
Household income (chilean pesos), n (%) [20]	Less than 65,000	5 (3.42%)	4 (3.51%)
	65,000–136,999	0 (0%)	0 (0%)
	137,000–180,999	9 (6.16%)	7 (6.14%)
	181,000–250,999	14 (9.59%)	12 (10.53%)
	251,000–350,999	21 (14.38%)	18 (15.79%)
	351,000–450,999	33 (22.6%)	25 (21.93%)
	451,000–650,999	21 (14.38%)	17 (14.91%)
	651,000–850,999	16 (10.96%)	13 (11.4%)
	851,000–1,050,999	7 (4.79%)	6 (5.26%)
	1,051,000–1,250,999	11 (7.53%)	6 (5.26%)
	1,251,000–1,450,999	5 (3.42%)	5 (4.39%)
	1,451,000–1,650,000	2 (1.37%)	1 (0.88%)
	Over 1,650,000	2 (1.37%)	0 (0%)
APOEe4 carriers, n (%)		27 (16.27%)	21 (15.79%)
Engagement in stimulating activities questionnaire	Total score	42.69 (8.77)	41.89 (8.49)
mean (SD) [2]	Cognitive	9.96 (3.16)	10.02 (3.19)
	Social	8.38 (2.44)	8.26 (2.4)
	Recreational	6.51 (2.58)	6.37 (2.41)
	Physical	9.31 (3.25)	9.07 (3.37)
	Productive	8.53 (2.6)	8.17 (2.45)
AL risk categories, n (%)^b	Very-low risk	6 (3.61%)	4 (3.01%)
	Low risk	36 (21.69%)	25 (18.8%)
	Medium risk	79 (47.59%)	70 (52.63%)
	High risk	45 (27.11%)	34 (25.56%)
^a Participants from cohort with no missing data in variables used to create the AL score. ^b Categories generated from the algorithm selected as best fit.

AL algorithm selection and validation

Following the original proposal, comprehensive AL score (ALCS) and the empirical AL score (ALES) were calculated following the algorithm flowchart described in Supplementary Fig. S1, generating final AL categories of very-low, low, medium, and high risk from percentiles obtained from sex-specific distributions of total scores. Both categorical scores were separately regressed on the covariates sex, age, and years of education through multinomial logistic regressions (MLR), for algorithm selection based on best model fit criteria. Surprisingly, likelihood ratio tests revealed poor model fit for both original ALCS (χ²(9) = 10.371, p = 0.321), and ALES (χ²(9) = 16.309, p = 0.061; Table 2a, top row).

Table 2

Model fit evaluation of comprehensive (ALCS) and empirical (ALES) algorithmsa
a. Likelihood ratio tests for original, corrected and IL6 biomarker inclusion algorithms
ALCS-original	c² (9) = 10.371, p = 0.321		ALES-original		c² (9) = 16.309, p = 0.061
ALCS-corrected	c² (9) = 17.098, p = 0.047		ALES-corrected		c² (9) = 19.012, p = 0.025*
ALCS-corrected + IL-6	c² (9) = 16.952, p = 0.049		ALES-corrected + IL-6		c² (9) = 20.844, p = 0.013*
b. Comparison between algorithms showing best model fit
	Model Fitting	Pseudo R²	D R²	BIC	D BIC	Log-likelihood (LL)	D LL
ALES-corrected	c² (9) = 19.012, p = 0.025	0.048	REF	408.54	REF	-173.599	REF
ALES-corrected + IL-6	c² (9) = 20.844, p = 0.013	0.054	0.006	396.19	-12.356^b	-167.421	6.178
Rate of correct classification			Very-low risk	Low risk	Medium risk	High risk	Overall
ALES-corrected		observed	6	39	53	68	166
		predicted	0	1	20	51	72
		% correct	0	2.6	37.7	75	43.4
ALES-corrected + IL-6		observed	6	36	79	45	166
		predicted	0	5	70	5	80
		% correct	0	13.9	88.6	11.1	48.2
p̂ Comparisons		pooled p̂	0	0.08	0.682	0.496	0.458
		SEDp	0	0.297	0.118	0.234	0.081
		z	0	0.381	4.308	2.727	0.595
		p (a = 0.05)	0.5	0.352	< 0.001	0.003	0.276
Inter-rater reliability
p Agreement	p Error	Cohen's κ	SEκ	95% CI		Z transformed	P
0.843	0.023	0.798	0.046	0.707	0.889	17.199	< 0.001
^a Multinomial logistic regression (MLR): AL risk category = Sex + Age + Years of education + e * Satisfactory model fit (a = 0.05). BIC = Bayesian Information Criterion. BIC D = Difference in BIC compared with the classic empirical AL scoring algorithm. ^b Substantial decrease in BIC.

To improve model fitting, algorithms were corrected for two main features of the sample: the large disparity between the number of male and female participants, and their age distribution (min = 69, max = 92 years). For the former, we generated the final AL risk categories based on percentiles from the whole sample distribution instead of a sex-specific approach. This modification was supported as no significant differences were found between sex in the ALCS (two-tailed unpaired t-test; t(164) = 0.223) or ALES raw scores (Mann-Whitney U test: U = 2223.5, p = 0.91, full statistical results detailed in Supplementary Table S3). For the latter, available age-specific clinical thresholds were included for the ERS, CRP and RHR biomarkers (see details in Supplementary Table S1).

Corrected ALCS and ALES risk categories were regressed on the covariates, with only corrected ALES showing significant model fit (χ²(9) = 19.012, p = 0.025), whereas corrected ALCS evidenced a large but not significant improvement (χ²(9) = 17.098, p = 0.047; (Table 2a, middle row).

On a final approach, IL-6 was included as an additional inflammatory biomarker to prevent a possible overload of the cardiovascular component, as suggested by previous health reports for the Chilean population. Preserving previous sex and age corrections, MLR showed a significantly improved model fit for the ALES-corrected + IL-6 algorithm (likelihood ratio test: χ²(9) = 20.844, p = 0.013), but no changes were evidenced in the ALCS-corrected + IL-6 model (likelihood ratio test: χ²(9) = 16.952, p = 0.049; Table 2a, bottom row).

Showing the best model fit, ALES-corrected, and ALES-corrected + IL-6 algorithms were further compared for final selection (Table 2b). Although only slight differences favoured the ALES-corrected + IL-6 in Mc-Fadden’s pseudo-R² (DR²=0.006) and log-likelihoods (DLL = 6.178), it showed a substantial decrease of the Bayesian Information Criterion (DBIC=-12.356), compared to ALES-corrected. Further analysis between the ratios of correct classification into AL risk categories showed no overall significant differences among the algorithms, although middle risk category was significantly better ranked by (pooled p̂= 0.682, p < 0.001). Inter-rater reliability test showed a high level of agreement between algorithms (Cohen’s kappa test for 4 measures and 2 scorers: κ = 0.798, z-transformed = 17.199, p < 0.001). Altogether, the empirical AL score corrected for sex disparities and inclusion IL-6 resulted in the best algorithm for further analysis in the GERO cohort. The final algorithm flowchart is described in Fig. 1.

Associations between AL risk index and demographic variables

Associations between AL risk categories from the selected algorithm, demographic covariates, and engagement in stimulating activities were estimated through MLR and very-low risk as reference category (Table 3). The overall model showed satisfactory fitting (χ²(30) = 49.531, p = 0.014), with significant model effects only for sex (χ²(3) = 8.076, p = 0.044), and years of education (χ²(3) = 9.78, p = 0.021). Education was inversely associated with AL load, with significantly increased odds of being in low (aOR = 0.54, p = 0.038), medium (aOR = 0.53, p = 0.033) and high-risk AL (aOR = 0.53, p = 0.035). No significant associations were found between AL risk categories and sex, age, household income, APOEε4 genotype status, engagement in cognitive, social, recreational, physical, or productive stimulating activities.

Table 3

MLR for AL risk, demographic covariates and engagement in stimulating activities
(Very-low risk set as reference category)
	Low risk		Medium risk		High risk		Likelihood ratio tests
	β (aOR)	95% CI	β (aOR)	95% CI	β (aOR)	95% CI	c² (3)	p
Sex	1.46 (4.32)	[0.08, 229.04]	2.81 (16.6)	[0.32, 867.25]	1.67 (5.31)	[0.1, 276.53]	8.08	0.044 *
Age	0.55 (1.74)	[0.86, 3.48]	0.46 (1.59)	[0.79, 3.17]	0.43 (1.54)	[0.77, 3.09]	7.18	0.066
Years of education	-0.62 (0.54) *	[0.3, 0.97]	-0.64 (0.53) *	[0.3, 0.95]	-0.63 (0.53) *	[0.3, 0.96]	9.78	0.021 *
Household income	0.73 (2.08)	[0.82, 5.26]	0.74 (2.1)	[0.83, 5.29]	0.66 (1.94)	[0.77, 4.88]	4.19	0.242
APOEe4 carriers	-0.36 (0.7)	[0.05, 9.48]	0.02 (1.02)	[0.08, 12.61]	-1.39 (0.25)	[0.02, 3.66]	5.1	0.165
Cognitive activities	0.04 (1.04)	[0.65, 1.67]	0.21 (1.24)	[0.78, 1.96]	0.14 (1.15)	[0.72, 1.84]	4.33	0.228
Social activities	-0.27 (0.76)	[0.36, 1.62]	-0.13 (0.87)	[0.42, 1.82]	-0.17 (0.84)	[0.4, 1.77]	1.51	0.68
Recreational activities	0.47 (1.61)	[0.72, 3.57]	0.38 (1.47)	[0.66, 3.24]	0.39 (1.47)	[0.66, 3.27]	2.09	0.554
Physical activities	0.63 (1.87)	[0.85, 4.12]	0.59 (1.8)	[0.82, 3.94]	0.69 (2)	[0.91, 4.4]	6.45	0.092
Productive activities	-0.42 (0.66)	[0.33, 1.32]	-0.6 (0.55)	[0.28, 1.1]	-0.62 (0.54)	[0.27, 1.08]	6.98	0.072
β = Log odds. aOR: adjusted odds ratio. * p < 0.05

The present study this study is the first evaluation of AL in a Chilean cohort, based on a previously published comprehensive approach for AL scoring [18]. In the attempt to replicate this proposal, important considerations emerged regarding the demographic characteristics of the studied population, which led to need to correct the original algorithms to achieve an adequate model fit.

First, to account for the large disparity between female and male participants in the GERO cohort, percentiles for AL risk categorization were generated from the whole sample scores distribution instead of the original gender-specific proposal. Second, given the studied population was significantly older (76.92, SD=5.07, min=70) than the one used to derive the original ALCS (51.3, SD=5.48) [18], available clinical thresholds for individuals ³65 and ³70 years old were included in the comprehensive algorithm. Both corrections largely improved both the ALCS and the ALES, but only the latter showed satisfactory model fit when regressed against similar demographic covariates used to validate the original proposal. Finally, considering recent reports of high cardiovascular and obesity risk in Chilean population [22], IL-6 was included as additional inflammatory biomarker to prevent a possible bias of the scores by the cardiovascular component. The result was a further increase in model fit criteria when compared to the corrected ALES with no significant improvement in the corrected ALCS algorithm. Although these findings seem to contradict the previously reported validity of the comprehensive approach [18], they point to the biggest difference in how both ALCS and ALES algorithms are constructed: the inclusion of clinical reference values and the suitability for the studied population. The large improvement in model fit from the original to the corrected ALCS (likelihood ratio tests p values 0.321 and 0.047, respectively), reveals the importance of setting specific laboratory ranges for older cohorts when determining risk values. While global population significantly ages and new classifications for young-old (65-74), middle-old (75-84), and oldest-old (³85) has been proposed [32, 33], clinical thresholds are usually established in young adults aged 20-40 [34, 35]. Yet, increased amount of evidence reports that some changes from established reference values are expected in elderly healthy people, such as increased fasting serum glucose and erythrocyte sedimentation rates, or decreased serum creatinine levels [34, 36, 37], suggesting that some abnormalities may be relevant pathological signs whereas other could just reflect a normal aging process [38]. Thus, it should not surprise that our AL risk categories were not associated with age, as it happens in younger cohorts [9, 18]. Although some longitudinal studies performed in British and US cohorts [39, 40] report a linear increase of AL through age, van Deurzen et al., showed that such trajectory in later life was highly associated with date of birth cohort and the cumulative socioeconomical disadvantages [40]. Conversely, an earlier cross-sectional study preformed in 18,000 participants between 20-90 years old showed that AL was strongly associated with age until 60 years old, after which scores stabilized with very slight variations [41], suggesting that there might be a point when AL cannot be considered as a predictor of age-related physiological decline. However, in this study some biomarkers were scored using clinical thresholds not adjusted for older ages, leaving open the question of how valid these ranges are for establishing a reliable AL score in these populations.

Finally, while higher AL risk was strongly predicted by lower levels of education, no associations were found with sex, household income, APOEe4 genotype status, or engagement in stimulating activities. The lack of association to household income somehow disagree with previous studies performed in Latin-American cohorts, where AL was highly related to a range of socio-economic disadvantages, such as housing conditions, social or religious support and neighbourhood characteristics [5, 42]. Nonetheless, it should be noted that according to recent demographic analyses, educational level is not correlated with per-capita income in the Chilean population [43], thus it could be expected a significant relation to education in the absence of association to household income in the present cohort.

Understanding AL in older adults is crucial to predict vulnerability to age-related diseases and to develop interventions and care plans to improve their quality of life. As the first evaluation of AL in a Chilean population, we expect this study will serve as first step for future assessments of age-related outcomes in the GERO cohort. Although our sample size is small and skewed to majority female participants, current Chilean demographics reports higher proportion of females aged 60 or over with respect of males (55.3% vs 44.7%), with broader differences with increased age and predicting greater differences in the coming years [44]. An additional limitation of the present study relates to the possibility that our results might be underpowered given the limited sample size. However, previous studies related to AL in small South American samples and skewed to majority females [9, 10] have shown adequate model fitness and satisfactory power (n=54, 68% females; n=256, 73.4% females, respectively). However, we expect to confirm our results after recruiting new participants in the near future.

Altogether, given the small number of similar reports in Latin-American populations, our results will help to expand the existing evidence linking AL with cognitive decline and dementia risk established in white-caucasian cohorts.

Ethics: A written informed consent to participate in the study is obtained for all participants of the GERO cohort. The Geroscience Center for Brain Health and Metabolism Research Project Nº 1140423 and informed consents, was approved in May 2015 by the Scientific Ethics Committee (SEC) of the Western Metropolitan Health Service (WMHS), Santiago (Chile). The study is registered under the clinical trial Number: NCT04265482 (https://clinicaltrials.gov/study/NCT04265482).

Acknowledgements: Servicio de Salud Metropolitano Oriente, Hospital del Salvador and CESFAM Santa Julia (Macul), CESFAM Ossandón (La Reina), CESFAM Juan Pablo II (La Reina), and CESFAM Cardenal Silvia Henríquez (Peñalolén) for their contribution to the realization of the study. The authors would like to express their gratitude to the participants and family members from the GERO cohort study.

Funding: ANID/FONDAP/15150012; AS is partially supported by ANID/FONDECYT Regular Fondecyt 1231839. The contents of this publication are solely the responsibility of the authors and does not represent the official views of these institutions. The funding body has not involvement in the design of the study and collection, analysis, and interpretation of data and in manuscript writing.

Data availability: The data supporting the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.

Conflicts of Interest: The authors declare that they have no competing interests. The “Geroscience Center for Brain Health and Metabolism” is funded by the National Agency of Research and Technology (“Agencia Nacional de Investigación y Tecnología, ANIT”), entity under the Chilean Ministry of Science and Technology, through the Priority Areas Research Center Founding (“Fondo de Financiamiento de Centros de Investigación en Áreas Prioritarias, FONDAP”) N° 1510012. The project was founded in 2015 with US$ 5,284,791 (equivalent to national currency), for a period of 5 years (2015–2020). (https://www.gerochile.cl/web/).

McEWEN BS. Stress, Adaptation, and Disease: Allostasis and Allostatic Load. Ann N Y Acad Sci. 1998;840:33–44.
Schulkin J. Rethinking Homeostasis. The MIT Press; 2002.
Guidi J, Lucente M, Sonino N, Fava GA. Allostatic Load and Its Impact on Health: A Systematic Review. Psychother Psychosom. 2021;90:11–27.
Borrell LN, Rodríguez-Álvarez E, Dallo FJ. Racial/ethnic inequities in the associations of allostatic load with all-cause and cardiovascular-specific mortality risk in U.S. adults. PLoS One. 2020;15:e0228336.
D’Alonzo KT, Munet-Vilaro F, Carmody DP, Guarnaccia PJ, Linn AM, Garsman L. Estresse de aculturação e carga alostática entre mulheres imigrantes mexicanas. Rev Lat Am Enfermagem. 2019;27.
Niño MD, Chavez YZ. Stress and Allostatic Load Among Latina/os: Evidence from the Hispanic Community Health Study/Study of Latina/os. J Immigr Minor Heal. 2021;23:895–903.
Rodriquez EJ, Coreas SI, Gallo LC, Isasi CR, Salazar CR, Bandiera FC, et al. Allostatic load, unhealthy behaviors, and depressive symptoms in the Hispanic Community Health Study/Study of Latinos. SSM - Popul Heal. 2021;16:100917.
McClain AC, Xiao RS, Tucker KL, Falcón LM, Mattei J. Depressive symptoms and allostatic load have a bidirectional association among Puerto Rican older adults. Psychol Med. 2022;52:3073–85.
Soria CA, Remedi C, D’Alessio L, Roldán EJA. Sex and Age-Related Differences in Neuroticism and Allostatic Load Index in Urban Patients with General Anxiety Disorder Treated with Alprazolam. Open J Psychiatry. 2018;08:212–32.
Freire A do NF, Barbosa JF de S, Pereira DS, Gomes C dos S, Guerra RO. Allostatic load and stress biomarkers in a sample of community-dwelling older adults. Arch Gerontol Geriatr. 2020;87:104006.
Germano ML, dos Santos Gomes C, de Souza Barbosa JF, Neto NJ, Pereira DS, Ahmed T, et al. Allostatic load and physical performance in older adults: Findings from the International Mobility in Aging Study (IMIAS). Arch Gerontol Geriatr. 2023;109:104961.
Seeman TE, Singer BH, Rowe JW, Horwitz RI, McEwen BS. Price of adaptation--allostatic load and its health consequences. MacArthur studies of successful aging. Arch Intern Med. 1997;157:2259–68.
Duong MT, Bingham BA, Aldana PC, Chung ST, Sumner AE. Variation in the Calculation of Allostatic Load Score: 21 Examples from NHANES. J Racial Ethn Heal Disparities. 2017;4:455–61.
Rodriquez EJ, Livaudais-Toman J, Gregorich SE, Jackson JS, Nápoles AM, Pérez-Stable EJ. Relationships between allostatic load, unhealthy behaviors, and depressive disorder in U.S. adults, 2005–2012 NHANES. Prev Med (Baltim). 2018;110 December 2017:9–15.
Rodriquez EJ, Kim EN, Sumner AE, Nápoles AM, Pérez-Stable EJ. Allostatic Load: Importance, Markers, and Score Determination in Minority and Disparity Populations. J Urban Heal. 2019;96:3–11.
Juster R-P, de Torre MB, Kerr P, Kheloui S, Rossi M, Bourdon O. Sex Differences and Gender Diversity in Stress Responses and Allostatic Load Among Workers and LGBT People. Curr Psychiatry Rep. 2019;21:110.
Kerr P, Kheloui S, Rossi M, Désilets M, Juster R-P. Allostatic load and women’s brain health: A systematic review. Front Neuroendocrinol. 2020;59:100858.
Buller-Peralta I, Gregory S, Low A, Dounavi M-E, Bridgeman K, Ntailianis G, et al. Comprehensive allostatic load risk index is associated with increased frontal and left parietal white matter hyperintensities in mid-life cognitively healthy adults. Sci Rep. 2024;14:573.
Slachevsky A, Zitko P, Martínez-Pernía D, Forno G, Court FA, Lillo P, et al. GERO Cohort Protocol, Chile, 2017–2022: Community-based Cohort of Functional Decline in Subjective Cognitive Complaint elderly. BMC Geriatr. 2020;20:505.
Beese S, Postma J, Graves JM. Allostatic Load Measurement: A Systematic Review of Reviews, Database Inventory, and Considerations for Neighborhood Research. Int J Environ Res Public Health. 2022;19:17006.
Howard JT, Sparks PJ. Does allostatic load calculation method matter? Evaluation of different methods and individual biomarkers functioning by race/ethnicity and educational level. Am J Hum Biol. 2016;28:627–35.
Measuring distance to the SDG targets – Chile. In: The Short and Winding Road to 2030. OECD; 2022.
Beydoun MA, Dore GA, Canas JA, Liang H, Beydoun HA, Evans MK, et al. Systemic inflammation is associated with longitudinal changes in cognitive performance among urban adults. Front Aging Neurosci. 2018;10 OCT:1–12.
Booth T, Starr JM, Deary I. Modeling multisystem biological risk in later life: Allostatic load in the lothian birth cohort study 1936. Am J Hum Biol. 2013;25:538–43.
O’Bryant SE, Gupta V, Henriksen K, Edwards M, Jeromin A, Lista S, et al. Guidelines for the standardization of preanalytic variables for blood‐based biomarker studies in Alzheimer’s disease research. Alzheimer’s Dement. 2015;11:549–60.
Nahler G. anatomical therapeutic chemical classification system (ATC). In: Dictionary of Pharmaceutical Medicine. Vienna: Springer Vienna; 2009. p. 8–8.
ISP. Recomendaciones para la medición de la velocidad hemática de sedimentación. Rol ISP. 2017;:1–9.
Wener MH, Daum PR, McQuillan GM. The influence of age, sex, and race on the upper reference limit of serum C-reactive protein concentration. J Rheumatol. 2000;27:2351–9.
McLoughlin S, Kenny RA, McCrory C. Does the choice of Allostatic Load scoring algorithm matter for predicting age-related health outcomes? Psychoneuroendocrinology. 2020;120 July:104789.
IBM Corp. IBM SPSS Statistics for Windows (Version 27.0). 2020.
Raftery AE. Bayesian Model Selection in Social Research. Sociol Methodol. 1995;25:111.
Alterovitz SSR, Mendelsohn GA. Relationship goals of middle-aged, young-old, and old-old internet daters: An analysis of online personal ads. J Aging Stud. 2013;27:159–65.
Lee SB, Oh JH, Park JH, Choi SP, Wee JH. Differences in youngest-old, middle-old, and oldest-old patients who visit the emergency department. Clin Exp Emerg Med. 2018;5:249–55.
Vásárhelyi B, Debreczeni LA. Lab Test Findings in the Elderly. EJIFCC. 2017;28:328–32.
Schuff-Werner P. Challenges of laboratory diagnostics in the elderly. LaboratoriumsMedizin. 2018;42:105–7.
Kelso T. Laboratory Values in the Elderly. Emerg Med Clin North Am. 1990;8:241–54.
Risch M, Sakem B, Risch L, Nydegger UE. The SENIORLAB study in the quest for healthy elderly patients. LaboratoriumsMedizin. 2018;42:109–20.
Lapin A, Böhmer F. Laboratory diagnosis and geriatrics: More than just reference intervals for the elderly ... Wiener Medizinische Wochenschrift. 2005;155:30–5.
Tampubolon G, Maharani A. Trajectories of allostatic load among older Americans and Britons: longitudinal cohort studies. BMC Geriatr. 2018;18:255.
van Deurzen I, Vanhoutte B. A Longitudinal Study of Allostatic Load in Later Life: The Role of Sex, Birth Cohorts, and Risk Accumulation. Res Aging. 2019;41:419–42.
Crimmins EM, Johnston M, Hayward M, Seeman T. Age differences in allostatic load: an index of physiological dysregulation. Exp Gerontol. 2003;38:731–4.
Morales-Jinez A, Gallegos Cabriales E, D’Alonzo K, Ugarte-Esquivel A, López-Rincón F, Salazar-González BC. Social Factors Contributing to the Development of Allostatic Load in Older Adults: A Correlational- Predictive Study. Aquichan. 2018;18:298–310.
Prieto J. Poverty traps and affluence shields: Modeling the persistence of income position in chile. Res Econ Inequal. 2021;29 May:169–207.
CIPEM - Conocimiento e investigacion de personas mayores. Características sociodemográficas de las personas mayores. Santiago, Chile; 2023.

No competing interests reported.

BullerPeraltaetal2024SupplementaryMaterial.docx

Download PDF

Version 1

posted

You are reading this latest preprint version

Derivation of an allostatic load index in the Chilean GERO study cohort reveals the need for specific clinical thresholds in older populations

Status:

Version 1

Abstract

Figures

Background

Methods

Participants and Ethics

Biomarkers collection

Original and corrected comprehensive AL algorithm (ALCS)

Original and corrected empirical AL algorithm (ALES)

Covariates

Statistical analysis

Results

Participants

AL algorithm selection and validation

Associations between AL risk index and demographic variables

Discussion

Conclusions and limitations

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1