Ambient Temperature and Biomarkers of Heart Failure
Ambient Temperature and Biomarkers of Heart Failure
In this analysis of the association between temperature measures and biomarkers related to inflammation and cardiovascular function in a population of heart failure patients, we observed higher levels of BNP beginning with 2-day moving averages. These results reached statistical significance for moving averages of 3 and 4 days. CRP results followed a similar pattern but were delayed by 1 day. These two biomarkers reflect interrelated mechanisms known to be associated with heart failure prognosis and symptom severity. CRP is a an indicator of systemic inflammation and is associated with risk of heart failure decompensation (Sato et al. 1999), whereas BNP is more specifically a marker of increasing hemodynamic load used in diagnosis and risk stratification among patients with congestive heart failure (Koglin et al. 2001). Predischarge BNP is a strong independent predictor of postdischarge outcome (Logeart et al. 2004) and long-term prognosis (Latini et al. 2002). To our knowledge, this is the first evidence that within-person BNP levels may be elevated during episodes of increased temperature. The results reported in this study are scaled to a 5°C change. Because we were examining associations across a season, the within-person differences in temperature could be quite large. We selected a 5°C change to represent the typical day-to-day variability in temperature within the Boston area. This is well within the range of data that we observe and represents a relatively small within-person change.
The levels of the inflammatory markers we observed in our study population of individuals with heart failure were elevated, as expected in a population of heart failure patients. For example, a BNP level of 100 pg/mL is typically used as a clinical cutoff for diagnosis of heart failure and decompensation, and the prognostic value of elevated plasma BNP in symptomatic systolic heart failure has been consistently reliable in various clinical settings (Cheng et al. 2001). For context, the magnitude of the association between temperature and BNP observed in this study indicates that 10°C higher temperature is associated with approximately one-third of the impact of a transient increase in sodium load from 1,610 mg/day to 5,750 mg/day for subjects with established heart failure, a stimulus that may trigger decompensation in some patients (Damgaard et al. 2007).
Most prior studies of environmental determinants in heart failure patients have focused on hospitalizations and mortality rather than symptoms and preclinical changes. Studies that have examined seasonal trends in hospital admissions and mortality for heart failure have typically observed a winter peak (Barnett et al. 2008; Boulay et al. 1999; Stewart et al. 2002). In a recent study, Gotsman et al. (2010) examined the association between temperature and prognosis in heart failure patients. They showed greater rates of hospitalization in the winter time but also found that warmer weather and admission during summer months were positively associated with symptom severity, prognosis, and reduced survival. To our knowledge, there is only one panel study that has directly examined health effects of temperature in heart failure patients. In this study, which was conducted in Montreal, Quebec, Goldberg et al. (2008) found an inverse association between oxygen saturation and maximum air temperature and relative humidity. They also found that maximum ambient air temperature, higher relative humidity, and ozone on the concurrent day predicted poorer self-reported general health (Goldberg et al. 2009). In our analyses of associations between temperature and biomarkers of heart failure, we observed positive associations with higher air temperature moving averages but not with relative humidity or ozone. This result may be due to the differences in the outcomes and also the differences in time windows assessed, Goldberg only tested lags of 0–2 days (3-day average). Our findings suggest an association with BNP after only two days, but it is possible that longer integrated averages of temperature over several days, which reflect a period of prolonged elevated temperature, are necessary to observe upregulation of CRP.
We are not aware of any previous studies investigating the effects of temperature with BNP or with CRP in heart failure patients. However, prior studies have reported the associations between temperature and CRP in other populations, and the results have been mixed (Halonen et al. 2010; Hampel et al. 2010; Schneider et al. 2008). Halonen et al. (2010) estimated temperature effects in a relatively healthy cohort of men and also observed inverse associations between temperature and CRP for lags of 0–1 day and for moving averages of up to 4 weeks. Schneider et al. (2008) measured CRP within a population of myocardial infarction survivors and observed inverse associations with temperature for averages and lags < 1 week. However, a more recent study by Hampel et al. (2010) contrasted patients with coronary heart disease and pulmonary disease in their responses to temperature and found no association with CRP among patients with pulmonary disease and lower CRP associated with lower temperature (corresponding to a positive association) for almost all lags among coronary heart disease patients. As has been suggested by these authors, differences in populations and their comorbidities and patterns of medication use may explain discrepant findings (Hampel et al. 2010).
We did not observe any evidence of associations with either TNF or endothelin-1 and temperature. Endothelin-1 is released from endothelial cells and has strong vasoconstrictive properties. Higher levels of endothelin-1 have been associated with heart failure severity and higher mortality rates (Bolger et al. 2002; Wei et al. 1994; Yang et al. 2005). In this analysis of only two measures per subject, we may not have had sufficient power to detect within-person changes. We note that 90% of endothelin-1 is synthesized in the endothelium and acts on the vascular smooth muscle locally (Levin 1995). TNF is a cytokine associated with inflammatory properties that also has been associated with severity and risk stratification of heart failure in a number of studies (Levine et al. 1990; Sharma et al. 2000). Our results are consistent with those of Halonen et al. (2010) who did not observe any significant association between temperature and TNF among older men also residing within the same region as our study participants.
Although we addressed the shape of the exposure–response relationship using penalized splines in additional analyses, we did not observe any statistically significant deviation from linearity. The only models with estimated degrees of freedom (df) > 1 were for endothelin-1, and these associations were nonsignificant (the estimated 1-day moving average df was 2.3; p = 0.4). In our analyses, neither PM2.5 nor ozone confounded the association between temperature and biomarkers and were not significantly associated with biomarker levels. We previously observed no consistent association between BNP and PM2.5 in a similar population of heart failure patients in the Boston area (Wellenius et al. 2007). Similar results were also observed in Aberdeen, Scotland, where there was no change in the number of hematological parameters measured in response to air pollution exposures in a panel of 132 heart failure patients (Barclay et al. 2009). However, our findings differ from other studies that have shown that PM2.5 is associated with changes in inflammatory biomarker levels in older aging populations (Dubowsky et al. 2006; Wilker et al. 2011), individuals with diabetes (O'Neill et al. 2007), and persons with coronary artery disease (Delfino et al. 2008). Heart failure patients may spend less time outside and have different time-activity and exposure patterns from healthy individuals and from those with other chronic conditions (Oka et al. 1993; van den Berg-Emons et al. 2001). These differences could also potentially explain our findings of no substantial deviations from linearity in the exposure–response relationship, if these patients with chronic heart failure are more vulnerable to extreme heat and spend little time outdoors during periods of extreme cold.
Potential limitations of this study include a limited number of repeated measures on study participants. This study specifically enrolled patients with heart failure and systolic dysfunction. Therefore, the results may not be generalizable to patients with heart failure and preserved systolic function. Additionally, the purpose of the original study was to examine effects of tai chi. This intervention was found to be associated with improved quality-of-life measures, but randomization to tai chi was not associated with statistically significant changes in the levels of these biomarkers (Yeh et al. 2011), and we did not observe any evidence of effect modification by tai chi randomization status in our analysis. We are also limited in our ability to assess spatial heterogeneity of the temperature metric, because we used only one stationary site for our analyses. Most (88%) participants in our study population lived within 40 km of Boston and our central site monitor; therefore, we expected to have limited misclassification for our within-person analyses. However, this study did not collect information on air-conditioner usage and availability, which also may be a source of variability in heat-related physiologic responses (Curriero et al. 2002; O'Neill et al. 2005; Ostro et al. 2010).
As mentioned above, time-activity patterns (Oka et al. 1993; van den Berg-Emons et al. 2001) could potentially confound or modify associations between temperature and these biomarkers, but we did not have sufficient data to address this topic. It is also possible that hemoconcentration could, in part, explain the associations between temperature and biomarkers, in situations where participants become more dehydrated in warmer weather. Data on blood urea nitrogen or hematocrit were not available in this study, but we did include body weight at each visit as a marker of hydration status in heart failure patients and found that the changes in the estimated association between temperature and the biomarkers were negligible (< 5%).
In this study of 100 patients with heart failure and systolic dysfunction, we observed significantly higher BNP and CRP, which are biomarkers related to heart failure symptoms and prognosis, with 3- to 4-day moving averages of apparent temperature. These findings suggest that changes in temperature and meteorology may alter underlying physiologic responses in this vulnerable population. Because of their compromised hemodynamic response, individuals with heart failure may have increased stress in the ventricles, as reflected by a rapid increase in circulating BNP levels, which may be accompanied or perhaps followed by elevation in systemic inflammatory responses, as indicated by CRP. Given the large public health burden of heart failure, further work is needed to examine the role of ambient temperature in influencing preclinical changes that occur before an episode of acute heart failure decompensation.
Discussion
In this analysis of the association between temperature measures and biomarkers related to inflammation and cardiovascular function in a population of heart failure patients, we observed higher levels of BNP beginning with 2-day moving averages. These results reached statistical significance for moving averages of 3 and 4 days. CRP results followed a similar pattern but were delayed by 1 day. These two biomarkers reflect interrelated mechanisms known to be associated with heart failure prognosis and symptom severity. CRP is a an indicator of systemic inflammation and is associated with risk of heart failure decompensation (Sato et al. 1999), whereas BNP is more specifically a marker of increasing hemodynamic load used in diagnosis and risk stratification among patients with congestive heart failure (Koglin et al. 2001). Predischarge BNP is a strong independent predictor of postdischarge outcome (Logeart et al. 2004) and long-term prognosis (Latini et al. 2002). To our knowledge, this is the first evidence that within-person BNP levels may be elevated during episodes of increased temperature. The results reported in this study are scaled to a 5°C change. Because we were examining associations across a season, the within-person differences in temperature could be quite large. We selected a 5°C change to represent the typical day-to-day variability in temperature within the Boston area. This is well within the range of data that we observe and represents a relatively small within-person change.
The levels of the inflammatory markers we observed in our study population of individuals with heart failure were elevated, as expected in a population of heart failure patients. For example, a BNP level of 100 pg/mL is typically used as a clinical cutoff for diagnosis of heart failure and decompensation, and the prognostic value of elevated plasma BNP in symptomatic systolic heart failure has been consistently reliable in various clinical settings (Cheng et al. 2001). For context, the magnitude of the association between temperature and BNP observed in this study indicates that 10°C higher temperature is associated with approximately one-third of the impact of a transient increase in sodium load from 1,610 mg/day to 5,750 mg/day for subjects with established heart failure, a stimulus that may trigger decompensation in some patients (Damgaard et al. 2007).
Most prior studies of environmental determinants in heart failure patients have focused on hospitalizations and mortality rather than symptoms and preclinical changes. Studies that have examined seasonal trends in hospital admissions and mortality for heart failure have typically observed a winter peak (Barnett et al. 2008; Boulay et al. 1999; Stewart et al. 2002). In a recent study, Gotsman et al. (2010) examined the association between temperature and prognosis in heart failure patients. They showed greater rates of hospitalization in the winter time but also found that warmer weather and admission during summer months were positively associated with symptom severity, prognosis, and reduced survival. To our knowledge, there is only one panel study that has directly examined health effects of temperature in heart failure patients. In this study, which was conducted in Montreal, Quebec, Goldberg et al. (2008) found an inverse association between oxygen saturation and maximum air temperature and relative humidity. They also found that maximum ambient air temperature, higher relative humidity, and ozone on the concurrent day predicted poorer self-reported general health (Goldberg et al. 2009). In our analyses of associations between temperature and biomarkers of heart failure, we observed positive associations with higher air temperature moving averages but not with relative humidity or ozone. This result may be due to the differences in the outcomes and also the differences in time windows assessed, Goldberg only tested lags of 0–2 days (3-day average). Our findings suggest an association with BNP after only two days, but it is possible that longer integrated averages of temperature over several days, which reflect a period of prolonged elevated temperature, are necessary to observe upregulation of CRP.
We are not aware of any previous studies investigating the effects of temperature with BNP or with CRP in heart failure patients. However, prior studies have reported the associations between temperature and CRP in other populations, and the results have been mixed (Halonen et al. 2010; Hampel et al. 2010; Schneider et al. 2008). Halonen et al. (2010) estimated temperature effects in a relatively healthy cohort of men and also observed inverse associations between temperature and CRP for lags of 0–1 day and for moving averages of up to 4 weeks. Schneider et al. (2008) measured CRP within a population of myocardial infarction survivors and observed inverse associations with temperature for averages and lags < 1 week. However, a more recent study by Hampel et al. (2010) contrasted patients with coronary heart disease and pulmonary disease in their responses to temperature and found no association with CRP among patients with pulmonary disease and lower CRP associated with lower temperature (corresponding to a positive association) for almost all lags among coronary heart disease patients. As has been suggested by these authors, differences in populations and their comorbidities and patterns of medication use may explain discrepant findings (Hampel et al. 2010).
We did not observe any evidence of associations with either TNF or endothelin-1 and temperature. Endothelin-1 is released from endothelial cells and has strong vasoconstrictive properties. Higher levels of endothelin-1 have been associated with heart failure severity and higher mortality rates (Bolger et al. 2002; Wei et al. 1994; Yang et al. 2005). In this analysis of only two measures per subject, we may not have had sufficient power to detect within-person changes. We note that 90% of endothelin-1 is synthesized in the endothelium and acts on the vascular smooth muscle locally (Levin 1995). TNF is a cytokine associated with inflammatory properties that also has been associated with severity and risk stratification of heart failure in a number of studies (Levine et al. 1990; Sharma et al. 2000). Our results are consistent with those of Halonen et al. (2010) who did not observe any significant association between temperature and TNF among older men also residing within the same region as our study participants.
Although we addressed the shape of the exposure–response relationship using penalized splines in additional analyses, we did not observe any statistically significant deviation from linearity. The only models with estimated degrees of freedom (df) > 1 were for endothelin-1, and these associations were nonsignificant (the estimated 1-day moving average df was 2.3; p = 0.4). In our analyses, neither PM2.5 nor ozone confounded the association between temperature and biomarkers and were not significantly associated with biomarker levels. We previously observed no consistent association between BNP and PM2.5 in a similar population of heart failure patients in the Boston area (Wellenius et al. 2007). Similar results were also observed in Aberdeen, Scotland, where there was no change in the number of hematological parameters measured in response to air pollution exposures in a panel of 132 heart failure patients (Barclay et al. 2009). However, our findings differ from other studies that have shown that PM2.5 is associated with changes in inflammatory biomarker levels in older aging populations (Dubowsky et al. 2006; Wilker et al. 2011), individuals with diabetes (O'Neill et al. 2007), and persons with coronary artery disease (Delfino et al. 2008). Heart failure patients may spend less time outside and have different time-activity and exposure patterns from healthy individuals and from those with other chronic conditions (Oka et al. 1993; van den Berg-Emons et al. 2001). These differences could also potentially explain our findings of no substantial deviations from linearity in the exposure–response relationship, if these patients with chronic heart failure are more vulnerable to extreme heat and spend little time outdoors during periods of extreme cold.
Potential limitations of this study include a limited number of repeated measures on study participants. This study specifically enrolled patients with heart failure and systolic dysfunction. Therefore, the results may not be generalizable to patients with heart failure and preserved systolic function. Additionally, the purpose of the original study was to examine effects of tai chi. This intervention was found to be associated with improved quality-of-life measures, but randomization to tai chi was not associated with statistically significant changes in the levels of these biomarkers (Yeh et al. 2011), and we did not observe any evidence of effect modification by tai chi randomization status in our analysis. We are also limited in our ability to assess spatial heterogeneity of the temperature metric, because we used only one stationary site for our analyses. Most (88%) participants in our study population lived within 40 km of Boston and our central site monitor; therefore, we expected to have limited misclassification for our within-person analyses. However, this study did not collect information on air-conditioner usage and availability, which also may be a source of variability in heat-related physiologic responses (Curriero et al. 2002; O'Neill et al. 2005; Ostro et al. 2010).
As mentioned above, time-activity patterns (Oka et al. 1993; van den Berg-Emons et al. 2001) could potentially confound or modify associations between temperature and these biomarkers, but we did not have sufficient data to address this topic. It is also possible that hemoconcentration could, in part, explain the associations between temperature and biomarkers, in situations where participants become more dehydrated in warmer weather. Data on blood urea nitrogen or hematocrit were not available in this study, but we did include body weight at each visit as a marker of hydration status in heart failure patients and found that the changes in the estimated association between temperature and the biomarkers were negligible (< 5%).
In this study of 100 patients with heart failure and systolic dysfunction, we observed significantly higher BNP and CRP, which are biomarkers related to heart failure symptoms and prognosis, with 3- to 4-day moving averages of apparent temperature. These findings suggest that changes in temperature and meteorology may alter underlying physiologic responses in this vulnerable population. Because of their compromised hemodynamic response, individuals with heart failure may have increased stress in the ventricles, as reflected by a rapid increase in circulating BNP levels, which may be accompanied or perhaps followed by elevation in systemic inflammatory responses, as indicated by CRP. Given the large public health burden of heart failure, further work is needed to examine the role of ambient temperature in influencing preclinical changes that occur before an episode of acute heart failure decompensation.
