Our results show that AL was associated with several chronic diseases in a cohort of Puerto Rican older adults. Increasing categories of AL were significantly associated with abdominal obesity, hypertension, diabetes and self-reported CVD and arthritis, but not with self-reported cancer. Moreover, AL showed stronger associations with CVD, arthritis and hypertension, and abdominal obesity, than parallel categories of MetS. Only diabetes showed stronger OR when MetS was the predictor; neither predictor was significantly associated with self-reported cancer.
The mean (SD) for AL score in our cohort was 3.8 (1.7) for all participants. It is difficult to compare overall measures of AL across studies due to variation in definitions. Yet, considering AL as a number of extreme parameters, our results are comparable to what has been reported in the MacArthur Studies of Successful Aging (3.9 (2.1)), a study on factors affecting physical and cognitive functioning (Seeman et al., 2004
). However, the MacArthur cohort consists of high-functioning, but older, participants than our group (70–79 versus 45–75 years), suggesting that this population of Puerto Ricans experience high levels of AL at younger ages. The men in our study experienced higher dysregulation than women. Cross-culturally, women have been found to have higher AL scores than men (Stewart, 2006
). It is possible that the specific set of parameters and clinical definitions that we selected drove this observation, as some of the cutoff points used here differ for men and women, in contrast to the combined population-specific cutoffs used by other studies. The basis for higher disrupted clinically-relevant parameters in Puerto Rican men should be further studied.
Increasing categories of AL (excluding respective definitive parameters) were significantly associated with higher odds of abdominal obesity, hypertension, diabetes, and self-reported CVD and arthritis, in agreement with previous reports (Crews, 2007
; Seeman et al., 2001
). These trends suggest that Puerto Ricans should aim to maintain the number of overloaded allostatic parameters to less than two, as the odds for disease were greatly reduced under that value. Notably, we found that AL was associated with both metabolic and non-metabolic conditions (e.g., arthritis), suggesting that disease status in this population may relate to overall dysregulation across various body systems and is not limited to metabolic responses. Interestingly, including CRP in the AL definition did not alter the observations obtained with the original 10-parameter score, although it has been previously associated with cancer, arthritis and obesity (Emery, Gabay, Kraan, & Gomez-Reino, 2007
; Florez et al., 2006
; Heikkila et al., 2008; Otterness, 1994
Comparison of parallel AL and MetS categories showed that associations of AL with abdominal obesity, CVD, hypertension and arthritis were stronger than for MetS. Only the OR for diabetes were stronger when using MetS. Reports from the MacArthur Studies show that AL predicted morbidity and mortality similarly to MetS, and they attributed the largest contribution to the secondary mediators (Karlamangla et al., 2002
; Seeman et al., 2001
). Still, it is possible that the non-MetS primary parameters play a role on the association of AL with disease. Our exploratory analysis with individual parameters of AL showed that several primary mediators contributed to the final models in association with several conditions, supporting our observations. For diabetes, the stronger association for MetS may be due to the inclusion of triglycerides as a parameter; higher levels of this lipid are associated with risk of diabetes. The AL definition does not include triglycerides, but rather cholesterol, which is not as strongly linked to diabetes. Nonetheless, our results support previously reported observations, further strengthening the premise that AL may be a better predictor of disease than MetS alone.
The prevalence of the chronic diseases studied in this Puerto Rican community was high, relative to most other populations. This is a major concern, as these chronic conditions impact the societal and economic burdens for this group, exacerbating existing disparities. Our results suggest a likely contribution of AL to such disparities; thus, studies and interventions aimed at understanding and reducing AL may be helpful in curbing prevalence of chronic disease in Puerto Ricans and other similar high risk populations. Self-reported cancer prevalence in our sample was unexpectedly low, as previous reports show high numbers of co-morbidities in cancer patients (Extermann, 2007
) and relatively high cancer rates in mainland Puerto Ricans (Ho et al., 2009
; Pinheiro et al., 2009
). Bias in study participation or in self report cannot be ruled out.
Our study had some limitations. First, the Census track method of identifying Puerto Ricans limited the selection to areas with high concentration of Hispanics. Ethnic density has been associated with common social and environmental stressors (Pickett & Wilkinson, 2008
), as well as a shared physical environment that may not promote healthy lifestyles. Although we cannot ascertain the effect of density in our cohort, it is possible that this might have influenced the observed high prevalence of some conditions. Nonetheless, we adjusted for some lifestyle behaviors that may be shared by those living in ethnically-dense areas. Selection bias could influence observed associations, as those declining participation in the study had lived in the US longer and were somewhat more acculturated than those participating. However, neither acculturation nor years of living in the US were found to be significant covariates in our models. Expanding the recruitment efforts with community approaches may have improved our ability to capture a more inclusive sample of this population. There were no significant differences between those with complete AL data and those without, except for lower prevalence of diabetes and MetS in the latter group. This may be due to lower acceptance of the blood draw by individuals with diabetes, as they might not want to fast. As the prevalence of these conditions was high in our sample, we do not expect that this would be a major source of bias. Based on our recruitment strategies, high participation rates, and few differences between participants and non-participants, we believe that our sample is reasonably representative of Puerto Rican adults living in urban communities in mainland US.
Another limitation is that we are examining cross-sectional data, based on the assumption that this population has been under stress for a long time in order to develop the cumulative dysregulation. This limits our ability to define causal direction. A loop mechanism could be operating, where disease status affects other factors that increase stress and propagates AL (Crews, 2007
; Szanton et al., 2005
). This cycle perpetuates and the disease state may be accelerated. Longitudinal data will be very valuable in validating our results and establishing the predictive value of AL on disease. Finally, recent evidence suggests that the neuroendocrine markers used here may reflect a transient state rather than long-term response to stress (Gersten, 2008
). Circadian changes in some neuroendocrine parameters (Hansen, Garde, Skovgaard, & Christensen, 2001
) and blood pressure (Cicconetti, Donadio, Pazzaglia, D'Ambrosio, & Marigliano, 2007
) have also been reported; this may bias the estimation of the associations, as those whose interviews were taken in the evening may have different characteristics that those interviewed during the day. Other primary mediators, such as salivary alpha-amylase, or both high and low cutoffs for cortisol, as well as repeated measures for fluctuating parameters, may better reflect long-term effects (Loucks, Justerb, & Pruessner, 2008
), and should be examined in future studies.
Our study had several strengths. First, we used a definition of AL that has not been explored previously. Other groups have used different cutoff points (sample-based quartiles or z-scores) and methods (weighted summation and partitioning) within their population to define AL (Logan & Barksdale, 2008
). Although a previous study used clinical cutoffs, it did not include the same set of parameters and guidelines used here (Seeman et al., 2008
). Our summation of clinical cutoffs included medication use in the classification, as is done to define MetS (Grundy et al., 2005
), considering parameters that had been stabilized by drug treatment and that cannot be captured with cross-sectional measures. This definition resulted in stronger associations than when excluding medication use or adjusting for them in the models, suggesting that medication use by these participants reflected their overloaded parameters. The use of clinical cutoffs in future studies could allow for more accurate inferences and comparisons to other populations.
We carefully adjusted our models for behavioral cofactors, including physical activity and dietary intake, which are not explored often, but play a role in the AL model. Increased total fat and energy intake have been associated with stressful eating (Oliver, Wardle, & Gibson, 2000
; Rutters, Nieuwenhuizen, Lemmens, Born, & Westerterp-Plantenga, 2008), and may also contribute to chronic health conditions. Nonetheless, it is possible that other factors not studied here (outlook in life, social support, self esteem, social position, resilience, genomic and non-genomic (not altering gene expression) effects, or other dietary factors) could also influence AL (McEwen, 2008
; Rosmond, 2005
). Further research on the contributing factors to AL in this population should be pursued.
Our study has several implications. First, our results, while obtained as cross-sectional data, add to the body of evidence that AL is associated to some chronic diseases beyond the use of traditional MetS markers. This could lead to the potential inclusion of additional parameters as risk assessment markers for such conditions in future studies and guidelines. Studies on minority groups are increasingly employing the AL model as a possible mechanism to explain health disparities. Thus, if confirmed in longitudinal studies and diverse populations, AL may be adopted as a measure of cumulative physiological dysregulation and a predictor of disease in the same way that MetS has been studied as a marker of metabolic dysregulation in the past. Studies focusing on health disparities, on the aging process, on psychosocial aspects of health, and on stress, could benefit from using this measure. Our results also support the importance of development of interventions for reducing AL and thus, prevalence of chronic conditions in Puerto Ricans. Targeted interventions at the individual and community level could include smoking cessation and stress-reduction programs, low-fat dietary advice, and activities that increase physical activity (McEwen, 2008
In conclusion, Puerto Rican older adults experience physiological dysregulation that is associated with increased odds of abdominal obesity, hypertension, diabetes and self-reported CVD and arthritis, but not with self-reported cancer. AL showed stronger associations with most conditions, than parallel categories of MetS, suggesting that AL may be a better marker of disease beyond traditional metabolic parameters. These results have prospective policy and research implications for Puerto Ricans and other ethnic groups.