Summary
Background and objectives Smoking is common in the hemodialysis population and is associated with increased all-cause mortality and development of cardiovascular disease. Cause-specific outcomes have not yet been examined in detail. This study investigated the association of baseline smoking status with all-cause, cardiovascular, and infection-related morbidity and mortality in patients undergoing long-term hemodialysis.
Design, setting, participants, & measurements Post hoc analysis of the HEMO Study in patients with available comorbidity, clinical, and nutritional data. Cox proportional hazards regression models were fit to estimate the association of smoking status with mortality. Poisson and negative binomial regression models were fit to estimate the association of smoking status with hospitalization rate.
Results Complete data were available for 1842 individuals (44% male, 63% black, 45% diabetic). Mean age was 58±14 years. At baseline, 17% were current smokers and 32% were former smokers. After case-mix adjustment, compared with never smoking, current smoking was associated with greater infection-related mortality (hazard ratio [HR], 2.04; 95% confidence interval [CI], 1.32–3.10) and all-cause mortality (HR, 1.44; 95% CI, 1.16–1.79) and greater cardiovascular (incidence rate ratio [IRR], 1.49; 95% CI, 1.22–1.82), infection-related (IRR, 1.35; 95% CI, 1.11–1.64) and all-cause (IRR, 1.43; 95% CI, 1.24–1.65) hospitalization rates. The population attributable fraction (i.e., fraction of observed deaths that may have been avoided) was 5.3% for current smokers versus never-smokers and 2.1% for current versus former smokers.
Conclusions Active smoking is prevalent in the chronic hemodialysis population and is associated with greater all-cause, cardiovascular, and infection-related morbidity and mortality.
Introduction
Smoking is common in the general population (23.1% of men and 20.6% of women) (1), despite an abundance of evidence associating it with increased risk for coronary artery disease (2), cancer (3), and greater mortality (4,5). Smoking has also been associated with an increased risk for both respiratory (6,7) and nonrespiratory (8) infections in the general population. Active smoking has been reported in up to 14.2% of individuals undergoing hemodialysis (9), a population that experiences an excessively high mortality rate.
In hemodialysis patients, most deaths are related to cardiovascular and infection-related conditions (10); cardiovascular mortality rates are 10- to 20-fold higher (11), and mortality rates from sepsis are up to 30-fold higher (12), compared with those in the general population. Advanced kidney disease has been associated with impaired metabolism of nicotine and cotinine (13); therefore, prolonged exposure to these and other byproducts of cigarette smoke may be particularly harmful in individuals undergoing maintenance hemodialysis.
To our knowledge, no study to date has examined cardiovascular and infection-related mortality and morbidity associated with smoking in the hemodialysis population. We therefore examined the association between smoking status, mortality, and morbidity in the Hemodialysis (HEMO) Study, a large randomized, controlled trial in patients undergoing maintenance hemodialysis that has collected detailed exposure and adjudicated outcomes data. We hypothesized that former and current smokers would have an increased risk for all-cause, cardiovascular, and infection-related mortality and hospitalization rates compared with never-smokers.
Methods
Study Design and Population
The study protocol was deemed exempt by the Partners Healthcare Institutional Review Board. All data were abstracted from the HEMO Study with the permission of the National Institute of Diabetes and Digestive and Kidney Diseases. The design of the HEMO Study has been previously reported (14,15). Briefly, HEMO was a prospective, multicenter, randomized clinical trial of standard versus high dialysis doses and low- versus high-flux membranes among prevalent adults undergoing thrice-weekly in-center hemodialysis. Exclusion criteria included a baseline serum albumin level <2.6 g/dl, residual urea clearance ≥1.5 ml/min per 35 L of urea distribution volume, inability to consistently achieve an equilibrated Kt/V >1.3, or the presence of end-stage comorbid conditions other than kidney failure. Of the 1846 HEMO Study participants, we excluded those who did not have an available record of smoking status at baseline (n=4); our final cohort consisted of 1842 individuals.
Exposures and Outcomes
The primary exposure of interest was baseline smoking status. In HEMO, smoking status was originally categorized as never, quit >20 years ago, quit ≤20 years ago, or current; for the purposes of this study, previous smoking history was collapsed into a single category of “former” smoking.
The primary outcomes were time from randomization to all-cause, cardiovascular, or infection-related death. Secondary outcomes included (1) time from randomization to all-cause, cardiovascular, or infection-related death or hospitalization and (2) rates of hospitalization. Deaths and hospitalization events during the study were reported to study investigators, who adjudicated the cause after review of hospital records and autopsy and narrative reports using standardized methods and criteria previously reported (14). In these analyses, participants were considered at risk for outcomes from the date of randomization until death or censoring (transplant, transfer, change in modality, or study completion on December 31, 2001).
Study Data
Per HEMO protocol, all study data were obtained via patient interview, chart review, and self-reported questionnaires. Demographic data, including sex, race, age, and hemodialysis vintage, were recorded at baseline. Other variables of interest included medication use; comorbid conditions (diabetes, ischemic heart disease, peripheral vascular disease, cerebrovascular disease, congestive heart failure, respiratory disease, cancer, and drug and alcohol use); dialysis treatment and hemodynamic variables; triceps skin-fold thickness and mid-arm muscle circumference (calculated as arm circumference − π × triceps skin-fold thickness; both in cm); and laboratory measures. Comorbid conditions were graded on the Index of Co-existing Disease (ICED) scale; analytically these were dichotomized (0 if ICED score=0; 1 if ICED score≥1) except for congestive heart failure and respiratory disease. For these conditions, more granular categorization was deemed important because smoking-related lung disease and heart failure may coexist and present as dyspnea necessitating hospitalization (0 if ICED score=0; 1 if ICED score=1; 2 if ICED score≥2) (16).
Statistical Analyses
Continuous variables were examined graphically and recorded as means (±SDs) or medians (with interquartile ranges) as appropriate. Comparisons were made using t tests, Wilcoxon rank-sum tests, ANOVA, or Kruskal-Wallis tests. Categorical variables were examined by frequency distribution, recorded as proportions and comparisons made using the chi-squared test.
For all analyses, exposure variables and covariates were considered as the most proximate value preceding at-risk time. The relationship between categories of smoking status (never, former, current) with time to death or hospitalization was examined by proportional hazards regression. Hospitalization rate was analyzed with Poisson or negative binomial regression, according to observed distribution.
Covariates for all models were selected on the basis of clinical and biologic plausibility. In the main analyses, the model included terms for demographic variables, dialysis access, serum sodium, and comorbid conditions that were not thought to serve as causal pathway intermediates (specifically excluding ischemic heart disease, congestive heart failure, peripheral vascular disease, cerebrovascular disease, malignancy, and respiratory disease). Because the prognostic significance of body size may differ according to sex, two-way sex-by-postdialysis weight cross-product terms were included. In secondary analyses, we explored potential mechanisms by examining for effect estimate attenuation upon adjustment for nutritional measures (serum creatinine, phosphorus, and total cholesterol levels; mid-arm muscle circumference and triceps skin-fold thickness; ultrafiltration requirement; reported caloric, protein, and sodium intake; and self-reported appetite) and comorbid conditions that were plausibly felt to lie on causal pathways between smoking and outcomes (ischemic heart disease, congestive heart failure, peripheral vascular disease, cerebrovascular disease, malignancy, and respiratory disease).
The linearity assumption for continuous variables was tested by graphical examination of Martingale residual plots and by comparative model fit diagnostics using the Akaike information criterion. The proportionality assumption was tested by scaled Schoenfeld residual testing. All survival models were stratified on clinical center. Overdispersion of incidence rate models was examined by goodness-of-fit testing. The population attributable fraction of mortality from smoking was estimated from adjusted hazard ratios (HRs) using the following formula: pd(HR −1/HR), where pd represents the proportion of cases exposed to the risk factor (17,18).
For all models, effect modification of smoking status on the basis of flux (high versus low) and Kt/V (high versus standard) assignment was examined via the inclusion of two-way cross-product terms, with significance adjudicated via likelihood ratio testing. Nominal two-sided P values <0.05 were considered to represent statistically significant differences. Analyses were performed using Stata 10.0MP software (College Station, TX).
Results
Smoking Status and Baseline Characteristics
The primary cohort consisted of 1842 individuals, of whom 17.3% were current smokers and 32.7% were former smokers at baseline. Mean age was 58±14 years; 44.5% of patients were diabetic, and 25.6% had peripheral vascular disease. Current smokers were more likely than never-smokers to be male, black, and younger; to have been undergoing hemodialysis for a longer period; to have more respiratory disease and greater ultrafiltration requirements; and to concurrently use alcohol and illicit drugs. Current smokers were less likely to have diabetes or peripheral vascular disease at baseline and had higher serum phosphorus and creatinine, but lower total cholesterol (Table 1). Former smokers were more likely to be older; to have more vascular disease, heart failure, and respiratory disease; and to have longer sessions and greater ultrafiltration requirements than never-smokers.
Characteristics of total study cohort according to baseline smoking status
Active smoking was associated with lower body mass index, lower triceps skin-fold thickness and mid-arm muscle circumference, and higher reported caloric and protein intake at baseline in unadjusted analyses (Table 2). Upon adjustment for age and sex, differences in body mass index, triceps skin-fold thickness, and macronutrient intake were attenuated, and current smokers had greater adjusted mid-arm muscle circumference (Supplemental Table A).
Distribution of anthropometric variables and nutritional indices according to baseline smoking status
Associations with All-Cause, Cardiovascular, and Infection-Related Mortality
Overall, participants contributed 4779 years of at-risk time, during which 780 deaths (312 cardiovascular and 197 infection-related) occurred; median follow-up time was 2.2 years. Compared with never smoking, former smoking was associated with a 26% greater adjusted risk for all-cause death (adjusted HR, 1.26; 95% confidence interval [CI], 1.07–1.50) and current smoking, with a 44% greater adjusted risk (adjusted HR, 1.44; 95% CI, 1.16–1.79). For cardiovascular-related mortality, there was a graded trend in risk according to smoking status (increasing from former to current versus never), but estimates failed to reach conventional levels of statistical significance. For infection-related mortality, current smoking was associated with an increased adjusted risk for death compared with never smoking (adjusted HR, 2.04; 95% CI, 1.35–3.10); former smoking took on an intermediate association with outcome but did not achieve statistical significance (Figure 1). We found no evidence for interaction according to vascular access type (P>0.50) and no evidence for effect modification of the association between smoking status and the outcomes of interest according to HEMO study Kt/V or flux assignment.
Adjusted hazard ratios (95% confidence intervals [CIs]) for all-cause, cardiovascular, and infection-related mortality according to baseline smoking status. The reference group is never-smokers. Multivariable estimates were adjusted for age, sex, race (black versus nonblack), HEMO Study Kt/V and flux group assignments, postdialysis weight, sex-by-postdialysis weight cross-product terms, diabetic status, access (fistula, graft, catheter), urine volume (≤200 ml/d, >200 ml/d), dialysis session length (≤180, 181–209, 210–239, ≥240 minutes), serum sodium level, and alcohol and drug abuse (never, former, current).
The population attributable fractions (i.e., the fractions of observed deaths that may have been avoided) were 5.3% for current smokers versus never-smokers and 2.1% for current versus former smokers.
Associations with Composite Outcome of Death or Hospitalization
For the composite cardiovascular outcome, participants contributed 3714 years of at-risk time, during which 831 individuals reached the end point. There was a graded association between smoking status (increasing from former to current versus never) and the composite cardiovascular outcome: The adjusted HRs were 1.18 (95% CI, 1.00–1.40) and 1.47 (95% CI, 1.20–1.81), respectively. For the composite infection-related outcome, participants contributed 3626 years of at-risk time, during which 793 individuals reached the end point. Again, a graded association between smoking status (increasing from former to current versus never) and outcome was noted: The adjusted HRs were 1.32 (95% CI, 1.11–1.56) and 1.38 (95% CI, 1.12–1.71), respectively (Figure 2).
Adjusted hazard ratios (95% confidence intervals [CIs]) for cardiovascular and infection-related composite end points of death or hospitalization according to baseline smoking status.The reference group is never-smokers. Multivariable estimates were adjusted for age, sex, race (black versus nonblack), HEMO Study Kt/V and flux group assignments, postdialysis weight, sex-by-postdialysis weight cross-product terms, diabetic status, access (fistula, graft, catheter), urine volume (≤200 ml/d, >200 ml/d), dialysis session length (≤180, 181–209, 210–239, ≥240 minutes), serum sodium level, and alcohol and drug abuse (never, former, current).
Associations with All-Cause, Cardiovascular, and Infection-Related Hospitalization Rates
During the study, a total of 1874 unique all-cause, 868 cardiovascular-related, and 918 infection-related hospitalizations were recorded. After adjustment for case mix, there was a graded association between smoking status (increasing from former to current versus never) and all-cause hospitalization rates: The adjusted incidence rate ratios were 1.25 (95% CI, 1.12–1.40) and 1.43 (95% CI, 1.24–1.65), respectively. A similar pattern was observed for cardiovascular and infection-related hospitalization rates (Figure 3).
Adjusted incidence rate ratios (95% confidence intervals [CIs]) of all-cause, cardiovascular, and infection-related hospitalizations according to baseline smoking status. The reference group is never-smokers. Multivariable estimates were adjusted for age, sex, race (black versus nonblack), HEMO Study Kt/V and flux group assignments, postdialysis weight, sex-by-postdialysis weight cross-product terms, diabetic status, access (fistula, graft, catheter), urine volume (≤200 ml/d, >200 ml/d), dialysis session length (≤180, 181–209, 210–239, ≥240 minutes), serum sodium level, and alcohol and drug abuse (never, former, current). Associations with all-cause mortality were estimated by negative binomial regression; associations with cardiovascular and infection-related hospitalization were estimated by Poisson regression.
Exploration of Potential Mechanisms
To examine for potential mechanisms underlying the observed associations, exploratory models were fit adjusting for nutrition-related and potential smoking-related comorbid conditions. The overall patterns of association were similar to those seen in the primary analyses; however, observed effect estimates generally attenuated to a greater degree in former smokers than in current smokers, raising the possibility of confounding or pathway intermediacy by the variables included in multivariable models (Supplemental Tables B and C).
Discussion
The main findings of this study were that current and former smoking (compared with never smoking) were associated with increased risks for (1) all-cause and infection-related mortality; (2) cardiovascular and infection-related death and hospitalization; and (3) all-cause, cardiovascular, and infection-related hospitalization rate.
In the general population, the association of smoking with cardiovascular disease and death has long been recognized (2,19); the estimated prevalence of smoking in adults in the United States is 19.3% (20). Considering that cardiovascular disease remains the leading cause of death in hemodialysis patients, it is concerning that the prevalence of smoking remains so high; patients with ESRD have presumably had more healthcare encounters and opportunity for education and smoking cessation interventions. According to our estimates based on population attributable fraction, 5.3% of the excess risk for death could be theoretically eliminated if hemodialysis patients never smoked; a smaller fraction (2.1%) could be eliminated if current smokers were to quit. These compare with estimates of 6.1%–13.1% in other studies (21–23). The less dramatic advantage of smoking cessation in hemodialysis patients may be due to competing risks from the higher burden of comorbid conditions, which may obscure potential health benefits.
Foley et al., using U.S. Renal Data System Wave 2 data (hemodialysis and peritoneal dialysis patients), reported that active smoking at dialysis initiation was associated with greater adjusted risk for new-onset heart failure, peripheral vascular disease, and mortality compared with nonsmoking. Interestingly, they found no significant excess risk for these outcomes in former smokers compared with nonsmokers (9). However, the study relied on administrative codes rather than outcomes committee–determined end points, raising the possibility of ascertainment bias limiting diagnostic accuracy (24,25). A study of peritoneal dialysis patients found a 22% higher adjusted risk for all-cause mortality among current or former smokers compared with never-smokers; this study did not use an adjudicated outcomes committee, which prevented a more granular analysis of cause-specific mortality (26). On the other hand, a recent systematic review and meta-analysis failed to find an association of smoking with increased cardiovascular events in hemodialysis patients (27).
In our analyses, we found a trend toward greater risk for cardiovascular death with current smokers compared with never-smokers. However, we did find significantly increased risk in both former and current smokers for the composite outcome of cardiovascular death and hospitalization and for cardiovascular hospitalization rates. The lack of association with cardiovascular mortality in our analyses may simply be an issue of power: We recorded only 312 events for cardiovascular death compared with 831 events for cardiovascular death or first hospitalization. In that regard, it should also be pointed out that the direction and magnitude of the effect estimates were remarkably similar across the outcomes considered, supporting the inference of an association between smoking and cardiovascular disease in this population.
The findings of increased risk for infection-related mortality, death and hospitalization, and hospitalization rates associated with current smoking status were particularly intriguing. Smoking has been associated with numerous abnormalities in immune function, including depression of immunoglobulin levels (except IgE) (28), decreased ability of alveolar macrophages to secrete IL-1 (29) and TNFα (30), and abnormalities of T cell functioning (31,32). In the general population, smoking has been associated with an increased risk for pneumococcal infection in immunocompetent individuals (odds ratio in current smokers compared with nonexposed nonsmokers, 4.1; 95% CI, 2.4–7.3) (6) and with an increased risk for influenza (odds ratio, 2.42; 95% CI, 1.53–3.83) in a study of male Israeli army recruits (7). Infection remains the second highest cause of death in long-term hemodialysis patients (10). Our analyses suggest that current smoking may be a major modifiable risk factor for infection-related hospitalization (adjusted incidence rate ratio, 1.35; 95% CI, 1.11–1.64) and mortality (adjusted HR, 2.04; 95% CI, 1.35–3.10) in this population. Indeed, they are the first to associate former and current smoking status with increased infection-related mortality in the chronic hemodialysis population. Of note, additional extraneous factors may predispose to a greater risk for infection-related complications among hemodialysis patients (e.g., socioeconomic background [33] and dialysis access [34]). We were careful to adjust for potential confounders where possible and found no evidence for interaction according to vascular access type.
The pharmacokinetics of tobacco-related metabolites are altered in advanced kidney disease. In one study, individuals with varying degrees of CKD were infused with nicotine, 0.028 mg/kg, for 10 minutes, followed by serial measurements of nicotine and cotinine over the subsequent 24-hour period. The authors found significant correlations between GFR and clearance of these compounds (13). Another study also found elevated nicotine levels 1 hour after smoking a cigarette in hemodialysis patients compared with individuals with normal kidney function; these levels remained higher than those in controls, even after 4–5 hours of hemodialysis in a smoke-free environment (35). If this is true for nicotine, it may also be true for toxic compounds derived from smoking. These studies suggest that those with advanced kidney disease may be subject to longer exposure to tobacco-related compounds and metabolites, which may contribute to an accentuated risk for adverse health outcomes. We did not find evidence for effect modification of the association of smoking with outcomes by dose or membrane flux, but it is unknown whether dose or flux influences removal of such compounds. Future studies should measure and assess the dialyzability of tobacco-related compounds and metabolites in long-term hemodialysis patients.
Our findings suggest a potential benefit for smoking cessation in long-term hemodialysis patients. In the general population, smoking cessation has been associated with significant reductions in morbidity and mortality; for example, a meta-analysis reported that smoking cessation after a myocardial infarction or cardiac surgery was associated with a 36% reduction in all-cause mortality and a 32% reduction in subsequent nonfatal myocardial infarctions (36). In our analyses, we found that both current and former smokers had higher risks for morbidity and mortality compared with never-smokers, with the risk being highest among current smokers. The lack of information on pack-years of smoking and actual quit date make it difficult to infer more about the magnitude of health benefits to be derived from quitting in the hemodialysis population. In the general population, an initial 25%–50% reduction in excess cardiovascular mortality is seen within 1–2 years of quitting smoking, but only after 10–15 years do the risks of former smokers approximate those of never-smokers (37,38). Given the reduced life expectancy of hemodialysis patients compared with the general population, the full extent of the beneficial effects of smoking cessation may be obscured by competing mortality risks.
The high proportion of active smokers in the hemodialysis population is concerning because of the risk-multiplying effects of smoking that we have identified. The hemodialysis population has more exposures to healthcare providers than do most other patient populations because of the thrice-weekly treatment schedule. That 17% of patients in this study were active smokers is sobering evidence of the behavioral patterns and addictive potential associated with smoking, and a reminder to address smoking cessation as part of routine dialysis rounds. Observational studies with the ability to compare outcomes between hemodialysis patients who quit smoking and those who continue to smoke may provide supportive evidence for this assertion.
Strengths of this study include detailed exposure and outcome assessment in the setting of a randomized, controlled trial. In addition, the use of an adjudicated outcomes committee provides additional confidence in the reported associations. However, several limitations merit consideration. The original classification used in the HEMO Study prevented us from examining more granular associations with pack-years, time since smoking cessation, and a more accurate estimate of “dose.” In this regard, the presented effect estimates for former smokers are likely to be conservative in relation to true associations. The presence of residual confounding based on covariates not considered, or incomplete adjustment for those considered, remains a possibility. Finally, differences in the study population examined (from a randomized, controlled trial completed in 2001) and the general hemodialysis population may limit the generalizability of our findings.
In conclusion, the prevalence of current smoking remains unacceptably high in chronic hemodialysis patients and is associated with greater morbidity and mortality. The frequency of interactions between hemodialysis patients and healthcare personnel affords unique opportunities for targeted interventions to help patients quit smoking.
Disclosures
S.M.B. has served as an advisor to Amgen, C.B. Fleet Company, and Proctor & Gamble. He has received speaking honoraria from Fresenius Medical Care North America. His spouse is employed by Astra Zeneca.
Acknowledgments
We thank the HEMO Study investigators and the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) data repository for the data used in this study. The HEMO Study was performed by the HEMO Study investigators and supported by the NIDDK. This paper was not prepared in collaboration with the investigators of the HEMO Study and does not necessarily reflect the opinions or views of the HEMO Study or the NIDDK.
F.R.M. was supported by a clinical fellowship grant from the National Kidney Foundation (2011-13) and has received support from the Scholars in Clinical Science Program of Harvard Catalyst, the Harvard Clinical and Translational Science Center (award no. UL1 RR025758), and financial contributions from Harvard University and its affiliated academic health care centers. S.M.B. is supported by DK079056. S.S.W. is supported by DK075941 and U01DK085660.
The content is solely the responsibility of the authors and does not necessarily represent the official views of Harvard Catalyst, Harvard University and its affiliated academic health care centers, the National Center for Research Resources, or the National Institutes of Health.
Footnotes
Published online ahead of print. Publication date available at www.cjasn.org.
This article contains supplemental material online at http://cjasn.asnjournals.org/lookup/suppl/doi:10.2215/CJN.03880412/-/DCSupplemental.
- Received April 20, 2012.
- Accepted July 16, 2012.
- Copyright © 2012 by the American Society of Nephrology
References
In this issue
