Stress Biomarker or Disease Score? Major Study Questions What Allostatic Load Really Measures

For more than two decades, scientists have used a concept called allostatic load to capture the cumulative ‘wear and tear’ that chronic and repeated stress inflicts on the body. The idea, first articulated by neuroendocrinologist Bruce McEwen in 1998, holds that sustained activation of the body’s stress systems leaves measurable fingerprints across the cardiovascular, metabolic, neuroendocrine and immune systems. By combining biomarkers such as blood pressure, cholesterol, HbA1c and C-reactive protein into a single index, researchers have treated allostatic load as the biological pathway through which social hardship, adversity and psychological strain ‘get under the skin’. A new study, however, raises an uncomfortable possibility: the way allostatic load is actually measured may be blurring the very distinction it was designed to capture, quietly merging subclinical stress biology with fully diagnosed disease.

The research, led by Cara L. Booker and colleagues at the University of Essex and published in SSM – Population Health, combines a systematic review of 428 population-based studies with an empirical analysis of more than 17,000 adults drawn from four major cohort studies: the English Longitudinal Study of Ageing, the Health and Retirement Study in the United States, the Midlife in the United States study, and Understanding Society, the UK Household Longitudinal Study. The team’s central question was deceptively simple: how much does allostatic load, as currently measured, actually overlap with multimorbidity, the co-occurrence of two or more long-term health conditions in the same individual? The answer matters because allostatic load is supposed to represent early, subclinical physiological dysregulation, while multimorbidity is a clinical endpoint. If the two constructs are largely the same thing in practice, thousands of studies interpreting allostatic load as a distinct marker of chronic stress may be measuring something rather different.

The systematic review, which screened 9,892 records and followed PRISMA 2020 guidelines, revealed striking inconsistency in how the construct is operationalised. Across the 428 included papers, the number of biomarkers used ranged from two to 28, spanning one to seven physiological systems. Cardiovascular and metabolic markers dominated: 410 of 428 studies included both metabolic and cardiovascular systems, while far fewer incorporated the primary neuroendocrine mediators that sit at the heart of allostatic theory. Systolic blood pressure appeared in 396 studies, diastolic blood pressure in 362, and C-reactive protein in 317. Despite this heavy reliance on a small core of markers, no consensus set of biomarkers has emerged, and the most common scoring approach remained a simple count of high-risk values.

The review’s most revealing findings concerned thresholds. When researchers use sample-based cut-points, typically the highest-risk quartile of their own study population, those thresholds can drift relative to established clinical criteria. The review found that quantile-based cut-offs for diastolic blood pressure averaged 83 mmHg, above the clinical threshold of 80 mmHg, and fasting glucose cut-offs averaged 7.4 mmol/L, exceeding the clinical 7 mmol/L. For body mass index, the typical quantile cut-off of 28.5 kg/m² sat well above the overweight threshold of 25. By contrast, sample-based thresholds for HbA1c, LDL cholesterol and total cholesterol fell below their clinical values. In other words, depending on the biomarker, a ‘high-risk’ allostatic load score may either exceed diagnostic criteria or remain comfortably within the normal range, positioning the index at very different points along the continuum from subclinical dysregulation to overt disease.

Medication use compounds the problem. Older adults frequently take antihypertensives, statins and glucose-lowering drugs that pharmacologically normalise the very biomarkers allostatic load indices count. A recent multi-cohort consensus statement warned that blood pressure and cholesterol, the ‘mainstays’ of allostatic load, may be inappropriate components in studies of older adults unless medication is explicitly considered. Yet the new review found that only about 36 percent of studies reported how medications were handled. Among those that did, most simply assigned participants taking relevant medication to the highest-risk category for that biomarker, a reasonable correction on its face, but one that injects diagnostic information directly into the exposure.

To test these concerns empirically, the team constructed eight different versions of allostatic load within each cohort: counts of high-risk biomarkers using sample quartiles or clinical cut-points, with and without medication adjustment, a pooled common-biomarker index, and a brief five-item score based on C-reactive protein, resting heart rate, HDL cholesterol, waist-to-height ratio and HbA1c, as recommended by the recent consensus statement. They then quantified overlap with three outcomes, general multimorbidity, cardiometabolic multimorbidity and immune multimorbidity, using C-statistics from logistic regression models adjusted for age, sex, ethnicity and survey year. C-statistics measure how well the allostatic load score discriminates between people with and without multimorbidity, with values above 0.8 indicating high overlap and values near 0.5 indicating none.

The results showed moderate-to-high overlap across the board. C-statistics ranged from roughly 0.68 to 0.82 for general multimorbidity, 0.69 to 0.86 for cardiometabolic multimorbidity, and 0.64 to 0.77 for immune multimorbidity. Strikingly, the different operationalisation strategies, sample quartiles, clinical thresholds, pooled biomarkers and the simplified five-item index, performed nearly identically, echoing earlier evidence that scoring algorithms differ little in predictive power. This convergence suggests that all current approaches draw on essentially the same metabolic and cardiovascular information, and that the brief five-item score captures little that longer biomarker batteries do not. The authors argue this reflects a broader feature of allostatic load research: indices are driven by downstream metabolic and inflammatory alterations rather than by the dynamic regulatory processes the theory describes.

Medication adjustment made things worse rather than better from a construct-validity standpoint. Incorporating medication information raised C-statistics in all four cohorts, with increases of up to +0.15 in the ageing studies ELSA and HRS, and the effect was largest for cardiometabolic multimorbidity. Because cardiometabolic conditions, including diabetes, hypertension, coronary heart disease and stroke, are defined by exactly the biomarker abnormalities that populate allostatic load indices, overlap with cardiometabolic multimorbidity exceeded that with general or immune multimorbidity in nearly every specification. Overlap with immune multimorbidity, covering conditions such as asthma, arthritis, chronic lung disease and thyroid disease, was consistently lower, likely reflecting the scarcity of inflammatory biomarkers and the absence of established cut-points for many of them. The pattern implies that what allostatic load most strongly detects is cardiometabolic disease itself, not a distinct stress-driven process preceding it.

The authors are careful to note that the overlap does not invalidate allostatic load as a concept. Multimorbidity may represent the clinical endpoint of allostatic processes, with chronic stress shifting physiological set points over time until dysregulation crosses into diagnosable disease. Cross-sectional measurement, however, cannot disentangle whether elevated allostatic load scores reflect the physiological embodiment of existing conditions or a shared pathophysiology, and the study’s reliance on data collected between 2004 and 2012 in predominantly White UK and US samples limits generalisability. The authors also acknowledge that C-statistics capture only one dimension of construct alignment and that complete-case analyses were used rather than multiple imputation.

The practical implications are nonetheless substantial. The authors recommend that researchers explicitly state where their chosen operationalisation sits on the subclinical-to-clinical continuum, consider sociodemographic influences on biomarker levels, including sex- and ethnicity-specific thresholds such as those proposed for BMI, and align medication-handling decisions with the analytical goal: identifying disease states or characterising latent physiological dysregulation. More fundamentally, they call for longitudinal designs, life-course biomarker collection and greater transparency in threshold definition. Until then, the study suggests, an index hailed as the biology of chronic stress may, in many published analyses, be functioning in practice as a partial census of diagnosed disease, a distinction that could reshape how hundreds of findings linking stress to health are interpreted.

Subject of Research: The conceptual and analytic overlap between allostatic load and multimorbidity

Article Title: A systematic review and empiric examination of the conceptual and analytic overlap between allostatic load and multimorbidity

Article References: A systematic review and empiric examination of the conceptual and analytic overlap between allostatic load and multimorbidity. (n.d.). https://doi.org/10.1016/j.ssmph.2026.101962

Image Credits: AI Generated

DOI: 10.1016/j.ssmph.2026.101962

Keywords: allostatic load, multimorbidity, chronic stress, biomarkers, cardiometabolic disease, physiological dysregulation, medication adjustment, clinical cut-points, population health, systematic review, ageing, cohort studies