No Survival Benefit in Octogenarians and Nonagenarians with Extended Hemodialysis Treatment Time

Background: The population of elderly end-stage renal disease patients initiating dialysis is rapidly growing. Although longer treatment is supposed to benefit for hemodialysis (HD) patients through more solute clearance and slower fluid removal, it is not yet clear how treatment session length affects mortality risk in octogenarians and nonagenarians. Methods: In a cohort of 112,026 incident HD patients between 2007 and 2011, we examined the association of treatment session length with all-cause mortality, adjusting for demographics and comorbid conditions. We also used restricted spline functions for age to evaluate continuous changes in the association of short (< 210 min) and extended (≥240 min) HD treatment (vs. 210 to < 240 min) with all-cause mortality over continuous age. Results: During the first 91 days of dialysis, patients aged ≥80 years tended to have the lowest treatment session length (median [interquartile range] 211 [193–230] min, r > 0.5). Longer treatment was associated with better survival in patients < 65 and 65 to < 80 years but not in octogenarians/nonagenarians. The association of extended treatment (≥240 min) with better survival was attenuated across age and not significant among patients aged ≥80 years with a hazard ratio of 1.10 (95% CI 0.99–1.20). Shorter treatment sessions (< 210 min) was associated with higher mortality across all age groups. Conclusion: Extended HD was not associated with lower mortality among octogenarians and nonagenarians, while it was associated with better survival among younger patients. Further studies are needed to determine the optimal treatment session length in elderly incident HD patients.

mortality over continuous age. Results: During the first 91 days of dialysis, patients aged ≥80 years tended to have the lowest treatment session length (median [interquartile range] 211  min, r > 0.5). Longer treatment was associated with better survival in patients < 65 and 65 to < 80 years but not in octogenarians/nonagenarians. The association of extended treatment (≥240 min) with better survival was attenuated across age and not significant among patients aged ≥80 years with a hazard ratio of 1.10 (95% CI 0.99-1.20). Shorter treatment sessions (< 210 min) was associated with higher mortality across all age groups. Conclusion: Extended HD was not associated with lower mortality among octogenarians and nonagenarians, while it was associated with better survival among younger patients. Further studies are needed to determine the optimal treatment session length in elderly incident HD patients.

Introduction
Despite medical advances and improvements in patient care, the mortality of end-stage renal disease (ESRD) patients on dialysis treatment remains unacceptably high [1]. Although there have been many studies for optimizing treatment session length in hemodialysis (HD) patients as an effort to improve survival by delivering the ideal dose of dialysis and reducing complications [2][3][4][5][6][7], it has not been clear yet which treatment session length is necessary for achieving better clinical outcomes.
Elderly patients are the fastest growing dialysis population according to the United States Renal Data System [1,8]. There are several factors that may warrant the need to modify the risk-benefit balance of extended dialysis particularly in this population [9][10][11]: (1) smaller increases in uremic burden or volumes due to lower dietary intake, compared with those in younger patients; (2) higher risk of HD-related complications; and (3) compromised quality of life by longer dialysis sessions.
Therefore, we examined whether there is a differential association between HD treatment session length and allcause mortality risk across age groups in a contemporary cohort of incident HD patients.

Materials and Methods
The study was approved by the Institutional Review Boards of the Los Angeles Biomedical Research Institute at Harbor-UCLA and the University of California Irvine Medical Center.

Study Population
We conducted a historical cohort study of incident HD patients receiving dialysis care from a large dialysis organization (LDO) in the United States, with comprehensive longitudinal data on patients' sociodemographics, comorbidities, laboratory tests, dialysis treatment characteristics, clinical events, and vital status. The original source population was a cohort of 208,820 incident dialysis patients receiving care in one of the LDO outpatient facilities over a 5-year period (January 2007 to December 2011). Among 133,162 incident ESRD patients who received dialysis treatment for at least 60 days and were treated only by in-center thrice-weekly HD during the follow-up, we excluded 21,036 patients if they were ≤18 and > 100 years old or had missing baseline HD treatment session length data. The follow-up time was divided into patient-quarters (i.e., 91-day periods from the date of first dialysis), where baseline was considered the first 91-day period.

Exposure Ascertainment
Quarterly averaged values of HD treatment session length during the first quarter of dialysis treatment were grouped into the following categories: < 180 min, 180 to < 210 min, 210 to < 240 min, and ≥240 min. Patients were also categorized according to their age at study entry: < 65, 65 to < 80, and ≥80 years.

Outcome Assessment
The primary outcome of interest was all-cause mortality. Patients were censored for death, kidney transplantation, transfer to a non-LDO dialysis unit, recovery of renal function or dialysis discontinuation, or at the end of the study (December 31, 2011). In the primary analyses, we sought to compare the association of treatment session length with all-cause mortality risk across age groups.
Sociodemographic, Dialysis Treatment, and Laboratory Characteristics Information regarding sociodemographics (including self-reported race/ethnicity and primary insurance type) as well as comorbid conditions (e.g., diabetes mellitus, hypertension, atherosclerotic heart disease, congestive heart failure, other cardiac diseases, cerebrovascular disease, chronic obstructive pulmonary disease, and history of cancer or liver disease) defined by the International Classification of Diseases-9 codes was also obtained from the LDO database. The modified Deyo-Charlson comorbidity index, where we excluded the presence or absence of kidney disease, was also obtained from the LDO database [12,13].
Blood samples for laboratory tests were collected before dialysis, except for post-dialysis serum urea nitrogen, using standardized techniques in all dialysis clinics. These samples were transported to a central laboratory in Deland, Florida, normally within 24 h, where they were measured using automated and standardized methods. Most laboratory tests were performed monthly, including serum creatinine, albumin, peripheral white blood cell count, total iron binding capacity, calcium, phosphorus, and bicarbonate. Serum intact parathyroid hormone was usually measured at least once per quarter. Hemoglobin was measured weekly to biweekly in most patients. The delivered dialysis dose was calculated by single-pool Kt/V using urea kinetic modeling. The normalized protein catabolic rate (nPCR) was calculated. To minimize measurement variability, all repeated measures for each 91day interval were averaged, and the quarterly mean values were used in analyses.

Statistical Analysis
Baseline characteristics across age groups and treatment time groups are summarized as proportions, mean ± SD, or medians (interquartile ranges [IQR]) depending on the data type. We used logistic regression to estimate the association between various clinical characteristics and the likelihood of extended treatment session length (treatment session length ≥240 vs. 180 to < 240 min) within age groups. We stratified each age group according to the treatment session length, and using Cox proportional hazard models, we compared the association of treatment session length with all-cause mortality risk across age groups using 3 hierarchical levels of adjustment: (1) Unadjusted model: Included the main predictor, treatment session length (2) Case-mix adjusted model: Included the unadjusted model covariates as well as age, sex, race/ethnicity, primary insurance type, 10 comorbid conditions (diabetes mellitus, hypertension, atherosclerotic heart disease, congestive heart failure, other cardiovascular disease, cerebrovascular disease, dyslipidemia, chronic obstructive pulmonary disease, liver disease, and history of malignancy), dialysis dose as measured by single-pool Kt/V, body mass index (BMI), and ultrafiltration rate (mL/h/kg body weight) Am  (3) Case-mix + malnutrition-inflammation-cachexia syndrome model: Included the case-mix adjusted model covariates as well as 10 clinically relevant laboratory variables (white blood cell count, hemoglobin, serum albumin, creatinine, bicarbonate, uncorrected calcium, phosphorus, intact parathyroid hormone, total iron binding capacity, and nPCR) We defined the case-mix model as our preferred model, which included core sociodemographic measures and other confounders of the association between treatment session length and outcomes. The malnutrition-inflammation-cachexia syndrome model was designated an exploratory model, which included confounders as well as potential intermediates on the causal pathway of treatment session length and mortality association. We also conducted subgroup analyses in which we compared the association of short (< 210 min) or extended treatment (≥240 min) vs. conventional treatment (210 to < 240 min) with mortality across age groups.
Effect modifications of the association between treatment session length (short: < 210 min or extended treatment: ≥240 min vs. the standard treatment (210-240 min) and all-cause death by age were examined with restricted cubic spline functions for age in the case-mix adjusted model. The frequency of missing data was low (1%) for most laboratory tests, except for nPCR (4%) and creatinine (4%), and the multiple imputation method with 5 datasets was used in all regression analyses. All analyses and figures were generated with STATA MP, version 13.1 (StataCorp, College Station, TX, USA) and SigmaPlot version 12.5 (Systat Software, San Jose, CA, USA).

Study Population Description
The study population included 112,026 incident HD patients (online suppl. Fig. 1; for all online suppl. material, see www.karger.com/doi/10.1159/000494336). The mean ± SD age of the overall cohort was 63 ± 15 years, and the median treatment session length (IQR) was 211 (193-230) min. The crude all-cause mortality rate of the overall cohort was 135 death events per 1,000 patientyears of follow-up (95% CI 133-136). Across all age groups, the baseline treatment session length distribution showed 3 distinct curves peaking at approximately 180, 210, and 240 min. Patients aged ≥80 years tended to have the lowest baseline treatment session length (median [IQR] 204 [184-217] min), with the highest peak at approximately 180 min, whereas patients aged < 65 years had the greatest treatment session length (median [IQR] 214 [199-234] min), with the highest peak at approximately 240 min (Fig. 1). Baseline characteristics of the cohort stratified by age and baseline treatment session length are presented in Table 1. Across all age groups, patients with the highest treatment session length tended to be male; were more likely to be African American and less likely to be Hispanic or Asian; were more likely to use a central venous catheter as their primary vascular access; were more likely to have diabetes, atherosclerotic heart disease, and other cardiovascular diseases; and had higher BMI compared to those in the lowest treatment session length category. Ultrafiltration rate tended to be lower in the extended treatment time group in all age groups. However, intradialytic hypotension was more frequent in the extended treatment time group only in patients aged ≥80 years. Table 2 shows the association between clinical characteristics and the likelihood of extended treatment time (treatment session length ≥240 min) across age groups in the case-mix adjusted logistic regression models (reference: treatment session length ≥180 to < 240 min). Across all age groups, being female, Hispanic, or Asian; using Medicaid insurance; and having an arteriovenous fistula (AVF)/ arteriovenous graft (AVG) access type were associated with a lower likelihood of extended treatment time. Being African American, having a diagnosis of diabetes or other cardiac disease, and having higher baseline BMI and creatinine values were associated with a higher likelihood of extended treatment time.

Treatment Session Length and All-Cause Mortality Across Age Groups
In the case-mix adjusted Cox regression analysis, there was a linear inverse association in mortality risk with greater treatment session length in patients aged  Hemoglobin, g/dL  Fig. 2, online suppl. Fig. 2).

Treatment Session Length and Mortality Across
Subgroups Figure 3a and b shows the case-mix adjusted associations between short (< 210 min) and extended (≥240 min) treatment session lengths and all-cause mortality (reference: ≥210 to < 240 min) across a priori selected sub-groups and stratified by age. Compared to the association of mortality with short treatment session length that did not show much difference between age groups (Fig. 3a), there was an incremental change in the association of extended treatment session length with mortality between age groups (Fig. 3b). Extended treatment (≥240 min) compared to standard treatment (≥210 to < 240 min) tended to be associated with a lower risk of mortality in those aged < 65 years especially in male, African American, non-diabetes, without congestive heart failure, low BMI (≤25), high interdialytic weight gain (IDWG, >3% of body weight), high spKt/V (> 1.5), and low albumin (≤3.5 mg/dL). Extended treatment time was not associated with a lower risk of mortality in any subgroups of patients aged 65 to < 80 and ≥80 years. In the subgroup of white patients aged ≥80 years, extended treatment (≥240 min) was associated with a higher risk of mortality.

Effect Modification by Age on the Association of Treatment Session Length with All-Cause Mortality
We then examined the effect modification by age on the association of treatment session length with allcause mortality in a continuous pattern using a restricted cubic spline function in the case-mix adjusted model. Using patients with standard treatment (≥210 to < 240 min) as reference, treatment session length < 210 min was associated with a higher risk of mortality across all ages, though the effect was attenuated along with increase of age (Fig. 4a). Conversely, the HR of all-cause mortality in association with extended treatment session length (≥240 min) was modified by age; lower risk of mortality with extended treatment session length in younger patients was incrementally attenuated across age (Fig. 4b). In a sensitivity analysis using different cut-off values for treatment time, age also did not modify the association of shorter treatment session length (< 195 min) with all-cause mortality. However, the association of longer treatment time (≥225 min) with all-cause mortality was modified by age in a similar linear pattern, where it was associated with a higher risk of mortality in those aged 80 years, though it did not reach statistical significance (online suppl. Fig. 3).

Discussion
In a nationally representative cohort of incident US HD patients, we observed an association of longer treatment session length with a lower risk of all-cause mortality in patients aged < 65 years, even beyond the usual treatment session length of ≥240 min per session. However, better survival of patients with extended treatment session lengths ≥240 min was not observed among patients aged 65 to < 80 and ≥80 years. Risk for all-cause mortality in patients with an extended treat-ment session length (≥240 min) was modified by age, whereby a lower risk of mortality was attenuated incrementally across age and tended to be associated with a higher risk of mortality in patients aged ≥80 years, although this did not reach statistical significance.  to < 240 min in each age group was used as a reference group. CHF, congestive heart failure; BMI, body mass index; IDWG, interdialytic weight gain; spKt/V, single pool Kt/V; Alb, albumin. Despite the existence of clinical practice guidelines and quality improvement programs to achieve optimal delivery of dialysis dose, treatment session lengths have traditionally been shorter in the United States compared with other international counterparts [14]. Although there was a lack of strong evidence based on randomized clinical trials [15,16] and earlier studies failed to show an association between session length and mortality [17,18], longer HD treatment sessions have demonstrated benefits in the care [2][3][4]. Recent publications reported that compared to conventional HD (210-240 min), an extended treatment session of more than 4 h was associated with better survival [6,7,19,20]. But, there have been variable results regarding the association of extended dialysis treatment with mortality [6,15,16,21].
In the present study, increased treatment session length showed a linear association with improved survival in patients aged < 65 years and 65 to < 80 years, and extended treatment session length (≥240 min) was significantly associated with better survival in patients aged < 65 years. However, longer treatment sessions (≥240 min) were not associated with better survival in patients aged 65 and older and showed a higher risk of mortality in the subgroup of white patients aged ≥80 years. Different associations of extended treatment session length ≥240 min with all-cause mortality between patients < 80 and ≥80 years were shown uniformly in the subgroup analysis. The association of better survival resulting from extended treatment session lengths ≥240 min was attenuated with increased age. The significance of the association was lost and extended treatment session length was not associated with better survival in patients aged ≥80 years. Longer HD treatment sessions are considered beneficial because they enable the delivery of larger doses of dialysis, which enhances solute clearance, such as phosphorus and unknown potential mediators that cause complications in ESRD [22,23]. These longer sessions also confer better tolerance in the management of interdialytic weight gain with preferable ultrafiltration rates, which has been known to reduce left ventricular hypertrophy and interdialytic hypotension [20,24,25]. However, our findings suggest that the association of longer HD treatment with better survival should be applied carefully, as longer treatments are not advisable for all HD patients. The benefits of extending dialysis treatment time in younger patients for the purpose of reducing the uremic burden and interdialytic weight gain may be diminished in patients aged ≥80 years. Moreover, elderly patients were likely more vulnerable to the higher risk of intradialytic hypotension probably due to higher comorbidities and advanced vascular stiffness when treatment sessions are extended ( Table 1). The impact of the excerbated loss of nutrient, such as from hypercatabolism in extended treatment sessions may also be more significant and lead to exceed the advantage of slower and increasing dose of dialysis in older patients [26].
However, shorter treatment sessions (< 210 min) were associated with worse survival regardless of age. Worse survival was shown to be associated with shorter treatment (< 210 min) in all subgroups, although some of the results were not significant. The adjusted HR ratio analysis also showed worse survival in all age groups. A minimum treatment session length should be recommended to ensure the delivery of a proper dose of dialysis and to avoid such poor outcomes. Such a recommendation would also be helpful to avoid the possibility of an increased ultrafiltration rate due to a shortened dialysis time [27,28].
The strengths of our study include its examination of a large, nationally representative cohort of dialysis patients, including more than 15,000 patients aged ≥80 years; the examination of incident HD patients whose characteristics are not confounded by survivor bias; and comprehensive availability of detailed, longitudinal patient-level comorbidity, laboratory, and dialysis treatment data. However, several limitations should be mentioned. First, treatment session length data in the first 91 days were used for the analysis to examine the long-term effect of HD treatment time on survival. Though the correlations of treatment session length at baseline with the overall study period were good (r > 0.5), baseline value may not reflect changes in session length over the course of follow-up or the impact of these changes on mortality. The overall distribution of treatment session length did not vary considerably across the follow-up period as noted in our previous study [29]. There was a good separation of treatment session length over time between groups divided by baseline treatment time, although the difference in treatment time became smaller, especially during the first 6 months (online suppl. Fig. 4).
Second, cause-specific mortality, which may be helpful to better elucidate the underlying mechanisms of the treatment session length-mortality association, was not examined in this study.
Third, our study results may be subject to residual confounding due to the observational study design and inability to adjust for unmeasured potential confounders. Moreover, a greater mortality risk associated with Am J Nephrol 2018;48:389-398 DOI: 10.1159/000494336 shorter treatment time is at least partly explained by indication bias because dying patients are likely to have unstable hemodynamics and low dietary intake and, hence, are likely to receive prescriptions for shorter treatment time in clinical practice. Finally, as with all observational studies, our study did not confirm a causal association between patterns of treatment session length and mortality.
In conclusion, extended treatment session length was not associated with better survival among octogenarians and nonagenarians compared with younger patients. However, a shorter treatment session (< 210 min) was associated with higher mortality risk regardless of age. Future studies are needed to determine the underlying mechanisms and to further define the optimal, individualized treatment session length management strategies for specific patient populations.