Abstract
Background:
Allostatic load has been linked to an increased risk of death in various populations. However, to date, there is no research specifically investigating the effect of allostatic load on mortality in older cancer survivors.
Aims:
To investigate the association between allostatic load (AL) and mortality in older cancer survivors.
Method:
A total of 1,291 adults aged 60 years or older who survived for ≥1 year since cancer diagnoses were identified from the 1999-2010 National Health and Nutrition Examination Survey. AL was the exposure of interest incorporating 9 clinical measures/biomarkers; one point was added to AL if any of the measures/biomarkers exceeded the normal level. The sum of points was categorized as an ordinal variable to reflect low, moderate, and high AL. Our outcomes of interest were all-cause, cancer-specific, and cardiovascular disease (CVD)-specific mortality. Death was identified by linkage to the National Death Index. Multivariable Cox proportional hazards models were used to estimate adjusted hazard ratio (aHR) and 95% confidence intervals (CI) of mortality by AL category.
Results:
Overall, 53.6% of participants were male and 78.4% were white. The mean age of study participants at interview was 72.8 years (SD=7.1). A total of 546 participants died during the follow-up (median follow-up time: 8.0 years). Among them, 158 died of cancer and 106 died of cardiovascular events. Results from multivariable Cox proportional hazards models showed that higher ALS was positively associated with higher all-cause mortality (ALS=4-9 vs. ALS =0-1: aHR=1.52, 95% CI =1.17-1.98, p-trend<0.01) and higher cancer-specific mortality (ALS=4-9 vs. ALS =0-1: aHR=1.80, 95% CI =1.12-2.90, p-trend=0.01). The association between ALS and cardiovascular mortality was positive but non-significant (ALS=4-9 vs. ALS =0-1: aHR=1.59, 95% CI =0.86-2.94, p-trend=0.11).
Conclusions:
Our study suggests that older cancer survivors can have a higher risk of death if they have a high burden of AL.
Keywords: allostatic load, cancer survivorship, gerontology, epidemiology
Introduction
Cancer is considered an age-related disease. Based on data from the Surveillance, Epidemiology, and End Results (SEER) Program, the average age at cancer diagnosis is 66 years in the US 1. Epidemiologic evidence shows that, compared to individuals aged 20 years, individuals aged 45-49 years have a 13-fold higher risk of developing cancer, and those aged 60 years have a 40-fold higher cancer risk 1. In 2019, nearly 65% of cancer survivors were 65 years or older, and this percentage is expected to continue rising 2. Older cancer survivors, compared to their younger counterparts, face the burden of cancer and aging simultaneously, indicating that age-related physiological changes should be monitored in the cancer care continuum for older cancer survivors to improve their prognosis 3.
Allostatic load (AL) was first introduced by Mc-Ewen and Stellar in 1993 4. AL is the cumulative wear and tear on the human body, and it represents the total effects of chronic physiologic stress, which may be induced by internal (e.g. cellular senescence) or external (e.g. social disparity) factors 5-7. For example, AL increases with chronological age as a result of long-term repeated stress induced by different types of exposures in life course 8-10. In addition, sociodemographic factors, such as race, have also been found to be associated with AL; particularly, African Americans have a higher burden of AL compared to white counterparts 11.
A systematic review synthesizing evidence from 267 original investigations reports that cancer survivors have a significantly higher risk of high AL compared to healthy counterparts 12. Given that high AL is an indicator of unfavorable physiologic changes, older cancer survivors may be at risk of unfavorable prognosis when living with high AL. Therefore, understanding the relationship between AL and mortality in these people is important because such knowledge may help clinicians more precisely predict prognosis and improve survival for this older vulnerable population.
In this study, we investigated the association between AL and the risk of all-cause, cancer-specific, and cardiovascular mortality in older cancer survivors using the National Health and Nutrition Examination Survey (NHANES) cohort.
Method
Data source and study population
The NHANES is a series of cross-sectional surveys led by the Centers for Disease Control and Prevention (CDC). It uses questionnaires, physical examinations, and laboratory tests to collect health-related information from US residents 13. In this analysis, we used NHANES data from 1999 to 2010. Participants with the following characteristics were included in this study: (1) age ≥ 60 years at baseline interview, (2) with a history of cancer, (3) survived for at least 1 year since their cancer diagnosis, (4) and with no missing data regarding vital status, AL score (see below), or other relevant covariates. A total of 1,291 participants met the criteria and were included in our analysis. Details about the process of participants’ inclusion are presented in Supplementary Figure 1.
Exposure and outcome of interest
The exposure of interest in this analysis was AL. We applied the algorithm established in prior research to define AL (Table 1), which incorporated measures of inflammation, metabolic homeostasis, and cardiovascular condition 14-16. Specifically, nine indicators, including systolic blood pressure (SBP), diastolic blood pressure (DBP), heart rate, total cholesterol, high-density lipoprotein (HDL), body mass index (BMI), glycohemoglobin, c-reactive protein (CRP), and albumin, were used to calculate the level of AL. Measurements of SBP, DBP, heart rate, and BMI were taken during baseline physical examinations 17, and the remaining indicators were examined by laboratory tests using blood samples collected at baseline 18. We added 1 point to the AL score if any of the aforementioned indicators exceeded the normal level. Details about cutoffs used to estimate the AL score are presented in Table 1. The overall AL score was then categorized as an ordinal variable to reflect low (0-1), moderate (2-3), and high (≥4) AL.
Table 1.
Factors considered in allostatic load and cutoff used in analysis.
| Factors | Criteria |
|---|---|
| Systolic blood pressure, mmHg | 0: < 140 1: ≥140 |
| Diastolic blood pressure, mmHg | 0: < 90 1: ≥ 90 |
| Resting heart rate, beats/min | 0: < 90 1: ≥ 90 |
| Total cholesterol, mg/dL | 0: < 240 1: ≥ 240 |
| HDL cholesterol, mg/dL | 0: ≥ 50 1: < 50 |
| Body mass index, kg/m2 | 0: < 30 1: ≥ 30 |
| Glycated hemoglobin, % | 0: < 6.4 1: ≥ 6.4 |
| CRP, mg/L | 0: < 3 1: ≥ 3 |
| Albumin, g/dL | 0: ≥ 4 1: < 4 |
Abbreviation: CRP: C-reactive protein; HDL: high-density lipoprotein
Our outcomes of interest were all-cause, cancer-specific, and cardiovascular mortality. We specifically included cardiovascular mortality as an outcome because epidemiologic evidence suggests that over 10% of cancer survivors die from cardiovascular diseases 19, and this proportion can increase in older cancer survivors because of the mixed effects of aging and cancer treatment toxicities 20. Death was identified by linkage to the National Death Index (NDI) through December 31, 2015. The underlying cause of death was ascertained via the International Classification of Diseases, Tenth Revision (ICD-10). In this study, cancer-specific deaths were defined as death resulting from cancer, which was identified using the ICD-10 codes C00-C97. Similarly, cardiovascular deaths were defined as death resulting from cardiovascular diseases, which was identified using the ICD-10 codes I00-I09, I11, I13, I20-I51, and I60-I69.
Covariates
Information on demographic factors including age, sex, race, education, and marital status was obtained from the baseline interview. Age was categorized as an ordinal variable (60 to 69 years, 70 to 79 years, and ≥80 years). Sex was dichotomized as male and female. Race was considered as white, black, and other. Education was categorized as high school or less, attended college, and college graduate. Marital status was treated as a binary variable (married and not married). We also included factors related to lifestyles and health conditions. Smoking status was based on self-reported data; participants were categorized as never smokers, current smokers (if ever smoked at least 100 cigarettes in life), and former smokers. Regular physical activity was defined as taking part in moderate or vigorous activities during the past 30 days, and it was treated as a binary variable (no vs. yes). Daily energy intake (kcal/day) was estimated based on food items recorded during a 24-hour food recall at baseline interview, and we categorized it as an ordinal variable to approximate quartiles (≤1366.1 kcal/day, 1366.2 to 1738.4 kcal/day, 1738.5 to 2151.9 kcal/day, and >2151.9 kcal/day). We incorporated 11 self-reported health conditions in our study, including heart attack, coronary heart disease, stroke, congestive heart failure, emphysema, chronic bronchitis, diabetes, osteoporosis, hypertension, chronic kidney disease, and arthritis. Multimorbidity was defined as having 2 or more of the aforementioned conditions. Cancer-related information was measured by self-report at baseline, and we included cancer type (breast, prostate, colorectal, melanoma, and other), time elapsed since cancer diagnosis (1-4, 5-9, and ≥10 years), and history of more than one cancer for the current analysis. We also included the survey year in our analysis to control for the potential influence introduced by the wide time span. The selection of study covariates was based on a priori knowledge regarding their relationship with AL and risk of death 16,21,22.
Statistical analysis
We first conducted a descriptive analysis by the summarizing distribution of study characteristics by AL score (0-1, 2-3, 4+). Chi-square tests were used to assess if distributions of study covariates differed by AL score. Fisher's exact test was used when cell counts were lower than 5. The number of deaths and follow-up time was summarized by AL score as well. Follow-up time was calculated for each participant from the date of baseline interview until death or censoring. Kaplan–Meier curves were used to visualize risk for all-cause, cancer-specific, and cardiovascular mortality by AL. Log-rank tests were performed to examine if risk of death varied by AL score.
Hazard ratios (HR) and 95% confidence intervals (CI) of AL score were estimated by using Cox proportional hazards regression models. The first model only included AL and age to estimate an age-adjusted HR for all-cause, cancer-specific, and cardiovascular mortality. For cancer-specific death analysis, deaths from causes other than cancer were treated as censored. Likewise, for cardiovascular death analysis, deaths from other causes were treated as censored. In the multivariable model, we further adjusted for other covariates including sex, race, education, marital status, smoking status, physical activity, energy intake, multimorbidity, time elapsed since cancer diagnosis, history of more than one cancer, and survey year. Low AL (score 0-1) was treated as the reference group in all models. Trend tests were conducted by treating AL as a continuous variable in the model. The proportionality assumption of the model was tested based on rescaled Schoenfeld residuals and no violation was found. To assess if sampling bias influenced the association between AL and death, we fitted a model that adjusted for sampling weight and included the same set of covariates. The sampling weights were obtained directly from NHANES data, accounting for the complex survey design, including oversampling, non-response, and post-stratification adjustments, to ensure the sample's representativeness with respect to the total population counts from the Census Bureau. Since cancer-specific mortality and cardiovascular mortality could be competing risks, the Fine-Gray model was applied to estimate HR for both outcomes. The proportion of participants excluded from the full analysis due to missing covariate data was greater than 10%, therefore, we conducted a sensitivity analysis using multiple imputation to assess the robustness of the effect measure of AL.
We also conducted subgroup analysis to examine if some factors could interact with AL in relation to risk of death in older cancer survivors. Prior evidence suggests that levels of inflammation, metabolic homeostasis, and cardiovascular risk vary by age, sex, comorbidity, and sociodemographic factors 23-26. Thus, subgroup analyses were conducted to explore if age (60-75 vs. ≥75 years), sex (male vs. female), race (white vs. black), education level (some college or higher vs. high school or less), and multimorbidity (no vs. yes) could interact with AL in relation to risk of death in older cancer survivors. We categorized AL as a binary variable, where a score of 0-3 was classified as low level and a score of 4-9 was classified as high level. In all subgroup analyses, the Cox proportional hazards regression models adjusted for the same set of covariates as the primary multivariable model except the factors used for stratification. An interaction term between AL and the aforementioned variables (age, sex, race, education level, and comorbidity burden) was generated and added to the model; a Wald test was used to examine if the interaction term was significant.
All statistical analyses were performed with R 4.1.2. We considered p<0.05 to be statistically significant.
Results
A total of 1,291 eligible participants were included in our analysis. Of them, 53.6% were males, 78.4% were white, the mean age at baseline was 72.8 years (SD=7.1), and the median follow-up time was 7.7 years. During the follow-up period, a total of 546 deaths were observed. Among the observed deaths, 158 deaths were due to cancer, and 106 deaths were due to cardiovascular events. Among them, 31.3% had a low AL (score 0-1), 50.6% had a moderate AL (score 2-3), and 18.1% had a high AL (score 4+). Detailed distributions of other study covariates can be found in Table 2. Overall, participants were more likely to have high AL if they were female, black, less educated, not married, or less physically active; participants with lower energy consumption or multimorbidity were also more likely to have higher AL. Colorectal cancer survivors were more likely to have high AL compared to survivors of melanoma, breast, prostate, and lung cancer.
Table 2.
Characteristics of study population (N=1,291)
| Overalla | Allostatic load scoreb | ||||
|---|---|---|---|---|---|
| Study characteristics | (N=1,291) n (%) |
0-1 (N=404) n (%) |
2-3 (N=653) n (%) |
4+ (N=234) n (%) |
p-value |
| Age at interview (year) | |||||
| 60 to 69 | 436 (33.8) | 130 (29.8) | 220 (50.5) | 86 (19.7) | 0.35 |
| 70 to 79 | 532 (41.2) | 171 (32.1) | 260 (48.9) | 101 (19.0) | |
| ≥80 | 323 (25.0) | 103 (31.9) | 173 (53.6) | 47 (14.6) | |
| Sex | |||||
| Male | 692 (53.6) | 224 (32.4) | 360 (52.0) | 108 (15.6) | 0.04 |
| Female | 599 (46.4) | 180 (30.1) | 293 (48.9) | 126 (21.0) | |
| Race | |||||
| White | 1012 (78.4) | 343 (33.9) | 508 (50.2) | 161 (15.9) | <0.01 |
| Black | 152 (11.8) | 26 (17.1) | 81 (53.3) | 45 (29.6) | |
| Other | 127 (9.8) | 35 (27.6) | 64 (50.4) | 28 (22.0) | |
| Education | |||||
| High school or less | 665 (51.5) | 176 (26.5) | 357 (54.7) | 132 (19.8) | <0.01 |
| Attended college | 327 (25.3) | 87 (26.6) | 172 (52.6) | 68 (20.8) | |
| College graduate | 299 (23.2) | 141 (47.2) | 124 (41.5) | 34 (11.4) | |
| Marital status | |||||
| Not married | 449 (34.8) | 124 (27.6) | 227 (50.6) | 98 (21.8) | 0.02 |
| Married | 842 (65.2) | 280 (33.3) | 426 (50.6) | 136 (16.2) | |
| Smoking status | |||||
| Never smoke | 553 (42.8) | 191 (34.5) | 268 (48.5) | 94 (17.0) | 0.22 |
| Current smoker | 116 (9.0) | 29 (25.0) | 64 (55.2) | 23 (19.8) | |
| former smoker | 622 (48.2) | 184 (29.6) | 321 (51.6) | 117 (18.8) | |
| Physical activity | |||||
| No | 669 (51.8) | 158 (23.6) | 357 (53.4) | 154 (23.0) | <0.01 |
| Yes | 622 (48.2) | 246 (39.5) | 296 (47.6) | 80 (12.9) | |
| Energy intake (kcal/day) | |||||
| ≤ 1366.1 | 323 (25.0) | 87 (26.9) | 161 (49.8) | 75 (23.2) | 0.03 |
| 1366.2 to 1738.4 | 323 (25.0) | 109 (33.7) | 162 (50.2) | 52 (16.1) | |
| 1738.5 to 2151.9 | 322 (24.9) | 97 (30.1) | 161 (50.0) | 64 (19.9) | |
| >2151.9 | 323 (25.0) | 111 (34.4) | 169 (52.3) | 43 (13.3) | |
| Multimorbidity c | |||||
| No | 478 (37.0) | 214 (44.8) | 217 (45.4) | 47 (9.8) | <0.01 |
| Yes | 813 (63.0) | 190 (23.4) | 436 (53.6) | 187 (23.0) | |
| Time elapsed since cancer diagnosis (year) | |||||
| 1-4 | 373 (28.9) | 104 (27.9) | 199 (53.4) | 70 (18.8) | 0.56 |
| 5-9 | 321 (24.9) | 103 (32.1) | 159 (49.5) | 59 (18.4) | |
| ≥10 | 597 (46.2) | 197 (33.0) | 295 (49.4) | 105 (17.6) | |
| History of more than one cancer | |||||
| No | 1151 (89.2) | 350 (30.4) | 592 (51.4) | 209 (18.2) | 0.12 |
| Yes | 140 (10.8) | 54 (38.6) | 61 (43.6) | 25 (17.9) | |
| Cancer type | |||||
| Breast cancer | 207 (16.0) | 62 (30.0) | 112 (54.1) | 33 (16.0) | <0.01 |
| Prostate cancer | 255 (19.8) | 83 (32.5) | 129 (50.6) | 43 (16.9) | |
| Colorectal cancer | 105 (8.1) | 24 (22.9) | 55 (52.4) | 26 (24.8) | |
| Melanoma | 74 (5.7) | 28 (37.8) | 37 (50.0) | 9 (12.2) | |
| Lung cancer | 29 (2.2) | 4 (13.8) | 20 (69.0) | 5 (17.2) | |
| Other | 621 (48.1) | 203 (32.7) | 300 (48.3) | 118 (19.0) | |
| Survey year | |||||
| 1999-2000 | 153 (11.9) | 42 (27.5) | 86 (56.2) | 25 (16.3) | 0.81 |
| 2001-2002 | 206 (16.0) | 63 (30.6) | 106 (51.5) | 37 (18.0) | |
| 2003-2004 | 201 (15.6) | 57 (28.4) | 107 (53.2) | 37 (18.4) | |
| 2005-2006 | 169 (13.1) | 57 (33.7) | 85 (50.3) | 27 (16.0) | |
| 2007-2008 | 276 (21.4) | 86 (31.2) | 135 (48.9) | 55 (19.9) | |
| 2009-2010 | 286 (22.2) | 99 (34.6) | 134 (46.9) | 53 (18.5) | |
Column percentages were reported for overall sample
Row percentages were reported for each group defined by allostatic load and p-values were estimated by chi-square tests.
Fisher's exact test was used when cell counts<5.
We incorporated chronic kidney disease, osteoporosis, diabetes, arthritis, heart attack, coronary heart disease, stroke, hypertension, emphysema, chronic bronchitis, and congestive heart failure when defining multimorbidity
Kaplan-Meier curves (Figure 1) suggested that the risk of all-cause and cancer-specific mortality increased with AL (log-rank p-values<0.01). The Kaplan-Meier curve showed that risk of cardiovascular death did not differ significantly by AL (log-rank p-value=0.10). Results in age-adjusted Cox proportional hazards regression models indicated that the risk of death increased with AL. Multivariable Cox proportional hazards models suggested positive and significant associations of all-cause (high vs. low AL: aHR=1.52, 95% CI=1.17-1.98) and cancer-specific (high vs. low AL: aHR=1.80, 95% CI=1.12-2.90) mortality with AL. Although the multivariable Cox proportional hazards model also showed a positive association between AL and cardiovascular mortality, the effect measures were not statistically significant (high vs. low AL: aHR=1.59, 95% CI=0.86-2.94). While point estimates of aHRs were attenuated towards the null to some extent in models corrected for sampling weight, the overall pattern still indicated a positive relationship between AL and all-cause mortality (Table 3). Death rates for different types of cancer are presented in Supplementary Table 1. Participants with colorectal cancer history had the highest all-cause death rate (death rate=58.2, 95% CI=43.1-76.7, per 1,000 person-years) and cardiovascular death rate (death rate=12.4, 95% CI=6.0-22.7, per 1,000 person-years), whereas participants with prostate cancer history had the highest cancer-specific death rate (death rate=21.2, 95% CI=15.4-28.5, per 1,000 person-years).
Figure 1.
Kaplan–Meier curves for (a) all-cause, (b) cancer-specific, and (c) CVD-specific mortality. The vertical axis indicates probability of being alive. The horizontal axis shows time of follow-up. Abbreviations: AL: allostatic load, CVD: cardiovascular disease.
Table 3.
Association between allostatic load and mortality
| Allostatic load score |
No. deaths/person- years |
Deaths per 1,000 person-years (95% CI) |
Age-adjusted HR and 95% CI (N=1,291) |
aHR and 95% CIa (N=1,291) |
aHR and 95% CIb (N=1,291) |
|---|---|---|---|---|---|
| All-cause death | |||||
| 0 - 1 | 135/3,396.3 | 39.7 (33.4, 46.9) | REF | REF | REF |
| 2 - 3 | 292/5,158.3 | 56.6 (50.5, 63.3) | 1.47 (1.20, 1.80) | 1.33 (1.08, 1.65) | 1.26 (0.99, 1.61) |
| 4 - 9 | 119/1,784.8 | 66.7 (55.5, 79.3) | 1.88 (1.47, 2.41) | 1.52 (1.17, 1.98) | 1.27 (0.92, 1.74) |
| overall: 546/10339.4 | overall: 52.8 (48.6, 57.3) | p-trend < 0.01 | p-trend < 0.01 | p-trend = 0.08 | |
| Cancer-specific death | |||||
| 0 - 1 | 36/3,396.3 | 10.6 (7.4, 14.6) | REF | REF | REF |
| 2 - 3 | 83/5,158.3 | 16.1 (12.8, 19.9) | 1.53 (1.04, 2.27) | 1.40 (0.93, 2.11) | 1.17 (0.72, 1.88) |
| 4 - 9 | 39/1,784.8 | 21.9 (15.6, 29.8) | 2.17 (1.38, 3.42) | 1.80 (1.12, 2.90) | 1.50 (0.85, 2.65) |
| overall: 158/10339.4 | overall: 15.3 (13.0, 17.8) | p-trend < 0.01 | p-trend = 0.01 | p-trend = 0.17 | |
| Cardiovascular death | |||||
| 0 - 1 | 25/3,396.3 | 7.4 (4.8, 10.8) | REF | REF | REF |
| 2 - 3 | 60/5,158.3 | 11.6 (8.9, 14.9) | 1.65 (1.04, 2.64) | 1.52 (0.94, 2.47) | 1.74 (1.00, 3.01) |
| 4 - 9 | 21/1,784.8 | 11.8 (7.3, 17.9) | 1.89 (1.05, 3.37) | 1.59 (0.86, 2.94) | 1.32 (0.67, 2.60) |
| overall: 106/10339.4 | overall: 10.3 (8.4, 12.4) | p-trend = 0.02 | p-trend = 0.11 | p-trend = 0.20 |
Abbreviations: aHR: adjusted hazard ratio, CI: confidence interval, CVD: cardiovascular disease
The model adjusted for age, sex, race, education, marital status, smoking status, regular physical activity, energy intake, multimorbidity, time elapsed since cancer diagnosis, history of more than one cancer, and survey year.
The model corrected for the sampling weight and adjusted for the same set of covariates as the main model.
Similar to the results of the primary multivariable models, the results obtained in the subgroup analyses suggested that older cancer survivors with a higher AL were more likely to die than those with a lower AL (Table 4). A significant interaction was observed between AL and comorbidity burden when analyzing all-cause mortality; specifically, high AL was associated with an increased risk of all-cause mortality in older cancer survivors living with a high burden of comorbidities (aHR=1.34, 95% CI=1.06-1.70, p-interaction=0.04), whereas the association among those with lower comorbidity burden was non-significant (aHR=0.72, 95% CI=0.42-1.26). No significant interaction was observed in the remaining subgroup analyses (Table 4).
Table 4.
Subgroup analysis for association between allostatic load (4-9 vs. 0-3) and risk of death
| All-cause death | Cancer-specific death | Cardiovascular death | ||||
|---|---|---|---|---|---|---|
| No. death |
aHR and 95% CI |
No. death |
aHR and 95% CI |
No. death |
aHR and 95% CI |
|
| Age | ||||||
| 60 to 75 (N=730) | 210 | 1.22 (0.87, 1.71) | 80 | 1.76 (1.03, 2.91) | 33 | 1.08 (0.41, 2.82) |
| ≥75 (N=561) | 336 | 1.21 (0.91, 1.60) | 78 | 1.64 (0.93, 2.98) | 73 | 1.18 (0.65, 2.15) |
| p-interaction = 0.65 | p-interaction = 0.30 | p-interaction = 0.84 | ||||
| Sex | ||||||
| Male (N=692) | 319 | 1.24 (0.93, 1.67) | 98 | 1.33 (0.80, 2.20) | 64 | 1.05 (0.53, 2.10) |
| Female (N=599) | 227 | 1.31 (0.96, 1.80) | 60 | 1.51 (0.83, 2.74) | 42 | 1.41 (0.66, 3.01) |
| p-interaction = 0.72 | p-interaction = 0.97 | p-interaction = 0.52 | ||||
| Multimorbidity | ||||||
| No (N=478) | 168 | 0.72 (0.42, 1.26) | 56 | 0.88 (0.36, 2.13) | 35 | 0.82 (0.24, 2.85) |
| Yes (N=813) | 378 | 1.34 (1.06, 1.70) | 102 | 1.54 (1.00, 2.38) | 71 | 1.35 (0.79, 2.33) |
| p-interaction = 0.04 | p-interaction = 0.19 | p-interaction = 0.17 | ||||
| Race | ||||||
| White (N=1,012) | 429 | 1.12 (0.87, 1.44) | 111 | 1.34 (0.84, 2.13) | 90 | 0.93 (0.52, 1.68) |
| Black (N=152) | 75 | 1.55 (0.87, 2.77) | 32 | 2.83 (1.17, 6.81) | 12 | 1.08 (0.23, 5.20) |
| p-interaction = 0.18 | p-interaction = 0.25 | p-interaction = 0.32 | ||||
| Low education | ||||||
| No (N=626) | 233 | 1.09 (0.79, 1.52) | 74 | 1.26 (0.72, 2.21) | 39 | 1.48 (0.68, 3.23) |
| Yes (N=655) | 313 | 1.48 (1.12, 1.95) | 84 | 1.74 (1.05, 2.89) | 67 | 1.22 (0.64, 2.32) |
| p-interaction = 0.34 | p-interaction = 0.85 | p-interaction = 0.63 | ||||
Abbreviations: aHR: adjusted hazard ratio, CI: confidence interval, CVD: cardiovascular disease
Allostatic load score was treated as a binary variable in the model (4-9 vs. 0-3).
The model adjusted for the same set of covariates as the primary multivariable model except the factor used for stratification.
Compared to the primary models, the HRs obtained from our competing risk analysis did not change substantially (Supplementary Table 2). In models fitted by multiple imputation, the HRs of AL were slightly attenuated towards the null value, but the overlapping 95% CI estimates suggested that the effect measures of AL were stable (Supplementary Table 3).
Discussion
The results from our study suggest that the risk of all-cause mortality among older cancer survivors increases with AL. In addition to all-cause mortality, risks of cancer-specific and cardiovascular death were also higher as AL increased, although statistical significance was not observed for cardiovascular mortality. Interestingly, our study suggests a significant interaction between AL and the burden of comorbidities in relation to all-cause mortality: the impact of AL on the risk of all-cause mortality appears to be stronger in older cancer survivors living with a higher burden of comorbidity. One possibility is that high AL and high comorbidity burden synergistically increased the risk of mortality in our study population. However, a significant interaction between AL and comorbidity burden was not observed for cancer-specific and cardiovascular death. Although the estimated aHRs of AL are largely different by comorbidity burden for both cancer-specific and cardiovascular death, the wide 95% CIs of these effect measures, which could be due to the small sample size for these outcomes, suggest that the differences are non-significant.
Prior research assessing the impact of AL on the risk of death yielded results similar to what was observed in the present analysis. For example, one cohort study analyzing data from an older population in the United States (N=29,701, mean age for white participants=64.1 years, mean age for black participants=65.4 years) used 10 biomarkers/clinical measures (serum albumin, C-reactive protein, high-density lipoprotein, total cholesterol, heart rate, SBP, DBP, serum creatinine, blood urea nitrogen, and waist circumferences) to define AL 27. The authors adjusted for age, gender, race, education, income, physical activity, smoking, alcohol, type 2 diabetes, BMI, and baseline comorbidity burden in the multivariable Cox proportional hazards model. Results from their models suggest that each unit increase in AL can increase all-cause mortality by 24% (HR=1.24, 95% CI=1.22, 1.27) and cancer-specific mortality by 7% (HR=1.07, 95% CI=1.03, 1.12) 27. Our study expands the conclusion of their research by further assessing cardiovascular death and focusing on cancer survivors, making our conclusion more clinically relevant to older people with a prior cancer diagnosis.
There is currently no standardized formula to measure the burden of AL. However, existing methods of estimating AL generally assess three key areas: inflammation, metabolism, and cardiovascular conditions 28-30. All three aspects were strongly associated with the prognosis of older adults with prior cancer diagnosis in the present study. Inflammation is an adaptive response that cells and tissues use to combat adverse events like infection and injury 31,32, and epidemiologic research indicates that inflammation may negatively impact long-term survival in cancer survivors. For example, the Women’s Healthy Eating and Living Study followed 2,919 female breast cancer patients (mean age of 53 years at baseline) and found that c-reactive protein, a biomarker of acute inflammation, is positively associated with higher risks of mortality (HR=1.91, 95% CI=1.13-3.23) and recurrence (HR=1.69, 95% CI=1.17- 2.43) 33. High levels of blood glucose and lipid are indicators of disturbed metabolic homeostasis, which can play an important role in tumor growth and contribute to adverse outcomes in cancer survivors 34-37. Affected by toxicities of cancer treatment, cancer survivors have a higher risk of adverse cardiovascular events and are more vulnerable to cardiovascular condition-related death 19.
To our knowledge, this is the first observational study that explores the impact of AL on mortality among older cancer survivors. There are several methodological strengths in the design and analysis of our study. First, we used nine blood biomarkers and clinical measures that are routinely collected in clinical practice to represent inflammation, metabolic homeostasis, and cardiovascular condition. Second, in addition to the primary analysis, we conducted several sets of sensitivity analyses (e.g., weighted analysis, multiple imputation, and competing risk analysis). Results of these sensitivity analyses, along with results of primary models, further ensured robustness of the association between AL and risk of death in older cancer survivors.
However, several limitations should be considered when interpreting our results. First, cancer history was self-reported, which is less reliable than medical record or cancer registry data. Second, we do not have cancer-specific information on treatment, stage, and histological subtype, leading to some residual confounding in statistical analysis. Third, factors incorporated for AL were all measured at a single time point at baseline, making it impossible to do time-varying analysis for the cohort. Finally, the majority of study participants are white, thus the results could be less generalizable to cancer survivors of minority racial/ethnic groups.
Our findings highlight that health practitioners should carefully monitor changes in AL of older adults with a history of cancer, adjust healthcare strategies based on AL, and offer personalized care to older cancer survivors to reduce the risk of death and prolong survival, especially among those with multimorbidity. Many indicators we used to calculate the burden of AL are closely related to comorbidities and can be applied to clinical practice to reflect the seriousness of comorbidities. For example, SBP≥140 and DBP≥90 has been used as the standard to diagnose hypertension, and blood glycated hemoglobin is an indicator of long-term glycemic control in patients with diabetes. These suggest that managing co-existing illnesses in older cancer survivors will have the potential to simultaneously improve cancer outcomes.
In conclusion, older cancer survivors living with high AL are at an increased risk of death. To improve prognosis of older cancer survivors at the population level, lifestyle factors or health interventions that have the potential to lower AL (e.g. exercise, mental health surveillance, and social support) should be considered during cancer care continuum for older adults with prior cancer diagnosis38,39. Future cohort studies should seek to enroll cancer survivors of diverse racial/ethnic groups and collect biomarkers or clinical measures to reflect AL in a longitudinal manner to account for the time-varying nature of AL in analysis. Cancer-specific information, such as cancer type, stage, and histological subtype, should also be incorporated in future studies to improve validity.
Supplementary Material
Data Availability Statement:
The data that support the findings of this study are openly available at https://www.cdc.gov/nchs/nhanes/about_nhanes.htm, 13.
Acknowledgments:
This analysis was internally funded by the University of Florida Health Cancer Center. We acknowledge and thank members of the National Center for Health Statistics of the Centers for Disease Control and Prevention, as well as the participants who contributed time and data to the National Health and Nutrition Examination Survey.
Abbreviations:
- aHR
adjusted hazard ratio
- AL
allostatic load
- BMI
body mass index
- CVD
cardiovascular disease
- CDC
Centers for Disease Control and Prevention
- cHR
crude hazard ratio
- CI
confidence interval
- DBP
diastolic blood pressure
- HDL
high-density lipoprotein
- HR
hazard ratio
- ICD-10
International Classification of Diseases Tenth Revision
- NDI
National Death Index
- NHANES
National Health and Nutrition Examination Survey
- SBP
systolic blood pressure (SBP)
- SD
standard deviation
- SEER
Surveillance, Epidemiology, and End Results
Footnotes
Ethical Statement: Informed consent was obtained from all individual participants included in the study.
Conflicts of Interest (CoI): While Dr. Dongyu Zhang is currently employed at Johnson & Johnson, during the time of the study, he was affiliated with the University of Florida and declares no conflict of interest. The other authors also report no conflicts of interest.
References
- 1.Risk Factors: Age - National Cancer Institute. Published April 29, 2015. Accessed March 10, 2022. https://www.cancer.gov/about-cancer/causes-prevention/risk/age
- 2.Shapiro CL. Cancer Survivorship. N Engl J Med. 2018;379(25):2438–2450. doi: 10.1056/NEJMra1712502 [DOI] [PubMed] [Google Scholar]
- 3.Kobayashi LC, Westrick AC, Doshi A, et al. New directions in cancer and aging: State of the science and recommendations to improve the quality of evidence on the intersection of aging with cancer control. Cancer. 2022;128(9):1730–1737. doi: 10.1002/cncr.34143 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.McEwen BS, Stellar E. Stress and the Individual: Mechanisms Leading to Disease. Arch Intern Med. 1993;153(18):2093–2101. doi: 10.1001/archinte.1993.00410180039004 [DOI] [PubMed] [Google Scholar]
- 5.McEwen B, Lasley EN. Allostatic load: when protection gives way to damage. Adv Mind Body Med. 2003;19(1):28–33. [PubMed] [Google Scholar]
- 6.McEwen BS. Allostasis and Allostatic Load: Implications for Neuropsychopharmacology. Neuropsychopharmacology. 2000;22(2):108–124. doi: 10.1016/S0893-133X(99)00129-3 [DOI] [PubMed] [Google Scholar]
- 7.McEwen BS, Gianaros PJ. Central role of the brain in stress and adaptation: Links to socioeconomic status, health, and disease. Ann N Y Acad Sci. 2010;1186(1):190–222. doi: 10.1111/j.1749-6632.2009.05331.x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Karlamangla AS, Singer BH, McEwen BS, Rowe JW, Seeman TE. Allostatic load as a predictor of functional decline: MacArthur studies of successful aging. J Clin Epidemiol. 2002;55(7):696–710. doi: 10.1016/S0895-4356(02)00399-2 [DOI] [PubMed] [Google Scholar]
- 9.McEwen BS. Interacting mediators of allostasis and allostatic load: towards an understanding of resilience in aging. Metabolism. 2003;52:10–16. doi: 10.1016/S0026-0495(03)00295-6 [DOI] [PubMed] [Google Scholar]
- 10.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(19):2259–2268. doi: 10.1001/archinte.1997.00440400111013 [DOI] [PubMed] [Google Scholar]
- 11.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(2):e0228336. doi: 10.1371/journal.pone.0228336 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Guidi J, Lucente M, Sonino N, Fava GA. Allostatic Load and Its Impact on Health: A Systematic Review. Psychother Psychosom. 2021;90(1):11–27. doi: 10.1159/000510696 [DOI] [PubMed] [Google Scholar]
- 13.NHANES - About the National Health and Nutrition Examination Survey. Published January 8, 2020. Accessed March 21, 2022. https://www.cdc.gov/nchs/nhanes/about_nhanes.htm
- 14.Beckie TM. A Systematic Review of Allostatic Load, Health, and Health Disparities. Biol Res Nurs. 2012;14(4):311–346. doi: 10.1177/1099800412455688 [DOI] [PubMed] [Google Scholar]
- 15.Chen X, Redline S, Shields AE, Williams DR, Williams MA. Associations of allostatic load with sleep apnea, insomnia, short sleep duration, and other sleep disturbances: findings from the National Health and Nutrition Examination Survey 2005 to 2008. Ann Epidemiol. 2014;24(8):612–619. doi: 10.1016/j.annepidem.2014.05.014 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Parente V, Hale L, Palermo T. Association between breast cancer and allostatic load by race: National Health and Nutrition Examination Survey 1999-2008. Psychooncology. 2013;22(3):621–628. doi: 10.1002/pon.3044 [DOI] [PubMed] [Google Scholar]
- 17.NHANES Questionnaires, Datasets, and Related Documentation. Accessed March 22, 2022. https://wwwn.cdc.gov/nchs/nhanes/Default.aspx
- 18.NHANES - NHANES Biospecimen Program. Published August 23, 2021. Accessed March 22, 2022. https://www.cdc.gov/nchs/nhanes/biospecimens/biospecimens.htm
- 19.Sturgeon KM, Deng L, Bluethmann SM, et al. A population-based study of cardiovascular disease mortality risk in US cancer patients. Eur Heart J. 2019;40(48):3889–3897. doi: 10.1093/eurheartj/ehz766 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Obeng-Gyasi S, Tarver W, Carlos RC, Andersen BL. Allostatic load: a framework to understand breast cancer outcomes in Black women. Npj Breast Cancer. 2021;7(1):1–4. doi: 10.1038/s41523-021-00309-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Acheampong T, Jiang L, Ziogas A, Odegaard AO. Multi-Systemic Biological Risk and Cancer Mortality: The NHANES III Study. Sci Rep. 2020;10(1):5047. doi: 10.1038/s41598-020-61945-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Xing CY, Doose M, Qin B, et al. Prediagnostic Allostatic Load as a Predictor of Poorly Differentiated and Larger Sized Breast Cancers among Black Women in the Women’s Circle of Health Follow-Up Study. Cancer Epidemiol Biomarkers Prev. 2020;29(1):216–224. doi: 10.1158/1055-9965.EPI-19-0712 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Crimmins EM, Johnston M, Hayward M, Seeman T. Age differences in allostatic load: an index of physiological dysregulation. Exp Gerontol. 2003;38(7):731–734. doi: 10.1016/S0531-5565(03)00099-8 [DOI] [PubMed] [Google Scholar]
- 24.Hollar DW. Allostatic Load, Mobility Disability, and Viral Effects in Cancer: A Structural Equation Model. Cancer Invest. 2022;40(4):366–377. doi: 10.1080/07357907.2021.1993880 [DOI] [PubMed] [Google Scholar]
- 25.McEwen BS. Sex, stress and the hippocampus: allostasis, allostatic load and the aging process. Neurobiol Aging. 2002;23(5):921–939. doi: 10.1016/S0197-4580(02)00027-1 [DOI] [PubMed] [Google Scholar]
- 26.Tang KL, Rashid R, Godley J, Ghali WA. Association between subjective social status and cardiovascular disease and cardiovascular risk factors: a systematic review and meta-analysis. BMJ Open. 2016;6(3):e010137. doi: 10.1136/bmjopen-2015-010137 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Akinyemiju T, Wilson LE, Deveaux A, et al. Association of Allostatic Load with All-Cause and Cancer Mortality by Race and Body Mass Index in the REGARDS Cohort. Cancers. 2020;12(6):1695. doi: 10.3390/cancers12061695 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.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 Health Disparities. 2017;4(3):455–461. doi: 10.1007/s40615-016-0246-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Juster RP, McEwen BS, Lupien SJ. Allostatic load biomarkers of chronic stress and impact on health and cognition. Neurosci Biobehav Rev. 2010;35(1):2–16. doi: 10.1016/j.neubiorev.2009.10.002 [DOI] [PubMed] [Google Scholar]
- 30.Suvarna B, Suvarna A, Phillips R, Juster RP, McDermott B, Sarnyai Z. Health risk behaviours and allostatic load: A systematic review. Neurosci Biobehav Rev. 2020;108:694–711. doi: 10.1016/j.neubiorev.2019.12.020 [DOI] [PubMed] [Google Scholar]
- 31.Coussens LM, Werb Z. Inflammation and cancer. Nature. 2002;420(6917):860–867. doi: 10.1038/nature01322 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Medzhitov R. Origin and physiological roles of inflammation. Nature. 2008;454(7203):428–435. doi: 10.1038/nature07201 [DOI] [PubMed] [Google Scholar]
- 33.Villaseñor A, Flatt SW, Marinac C, Natarajan L, Pierce JP, Patterson RE. Postdiagnosis C-reactive protein and breast cancer survivorship: findings from the WHEL study. Cancer Epidemiol Biomark Prev Publ Am Assoc Cancer Res Cosponsored Am Soc Prev Oncol. 2014;23(1):189–199. doi: 10.1158/1055-9965.EPI-13-0852 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Cust AE, Kaaks R, Friedenreich C, et al. Metabolic syndrome, plasma lipid, lipoprotein and glucose levels, and endometrial cancer risk in the European Prospective Investigation into Cancer and Nutrition (EPIC). Endocr Relat Cancer. 2007;14(3):755–767. doi: 10.1677/ERC-07-0132 [DOI] [PubMed] [Google Scholar]
- 35.Tonorezos ES, Jones LW. Energy Balance and Metabolism after Cancer Treatment. Semin Oncol. 2013;40(6): 10.1053/j.seminoncol.2013.09.011. doi: 10.1053/j.seminoncol.2013.09.011 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Boroughs LK, DeBerardinis RJ. Metabolic pathways promoting cancer cell survival and growth. Nat Cell Biol. 2015;17(4):351–359. doi: 10.1038/ncb3124 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Gaude E, Frezza C. Tissue-specific and convergent metabolic transformation of cancer correlates with metastatic potential and patient survival. Nat Commun. 2016;7(1):13041. doi: 10.1038/ncomms13041 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Campo RA, Light KC, O’Connor K, et al. Blood pressure, salivary cortisol, and inflammatory cytokine outcomes in senior female cancer survivors enrolled in a tai chi chih randomized controlled trial. J Cancer Surviv Res Pract. 2015;9(1):115–125. doi: 10.1007/s11764-014-0395-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Zhang D, Zhao Y, Kaushiva A, Zhu Z, Wang JHY, Braithwaite D. Self-rated health in relation to fruit and vegetable consumption and physical activity among older cancer survivors. Support Care Cancer Off J Multinatl Assoc Support Care Cancer. 2021;29(5):2713–2722. doi: 10.1007/s00520-020-05782-6 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data that support the findings of this study are openly available at https://www.cdc.gov/nchs/nhanes/about_nhanes.htm, 13.