Bliwise DL, Kutner NG, Zhang R, Parker KP. Survival by Time of Day of Hemodialysis in an Elderly Cohort. JAMA. 2001;286(21):2690-2694. doi:10.1001/jama.286.21.2690
Author Affiliations: Departments of Neurology (Dr Bliwise) and Rehabilitation Medicine (Dr Kutner), School of Medicine, Department of Biostatistics, Rollins School of Public Health (Ms Zhang), and Department of Adult and Elder Health, Nell Hodgson School of Nursing (Dr Parker), Emory University, Atlanta, Ga.
Context Patients with end-stage renal disease (ESRD) typically undergo hemodialysis
(HD) during the morning or afternoon, with time of treatment generally based
on space availability or patient preference. No studies have investigated
variation in patient survival as a function of the time of day when they receive
Objective To investigate the association of elderly patients' HD treatment shift
with their continued survival, controlling for well-established HD-related
mortality risk factors.
Design, Setting, and Participants An 11-year follow-up cohort study conducted among 242 ESRD patients
aged 60 years or older who underwent HD at 58 dialysis facilities in Georgia
either during a morning shift (n = 167) or an afternoon shift (n = 75) and
who completed baseline (1998) and follow-up (1991) interviews.
Main Outcome Measure Mortality from all causes occurring through July 7, 1999, as verified
by death-certificate reviews, and compared by morning vs afternoon-shift HD.
Results Morning-shift HD patients survived significantly longer than afternoon-shift
patients (median survival, 941 days vs 470 days; P<.001).
A Cox proportional hazards model indicated that the morning shift was protective
(relative risk, 0.71; 95% confidence interval, 0.53-0.95) independent of age,
race, sex, body mass index, functional status, diabetic ESRD, cardiovascular
comorbidity, weekly hours of dialysis, and months of dialysis.
Conclusions Possible explanations for differential survival in association with
morning vs afternoon dialysis include salutary effects of sleep in the morning
or less efficient biochemical exchange during afternoon dialysis. Results
from this cohort study may warrant prospective observational studies and randomized
clinical trials that systematically alter the time of day at which HD is administered.
Every year in the United States, more than 300 000 patients receive
treatment for end-stage renal disease (ESRD) and incur an annual mortality
rate of approximately 20%.1 Most patients receive
in-center hemodialysis (HD) throughout the course of their illness, customarily
in the morning or afternoon. Few data suggest that the time of day when HD
is performed has any functional consequence for patients undergoing the procedure.
In this study, we present data from a well-characterized cohort of elderly
HD patients for whom differential survival was associated with HD temporal
variation and was independent of obvious medical or demographic factors that
might otherwise have explained this relationship.
This study comprised 242 HD patients who were members of a previously
described cohort of 349 elderly ESRD patients receiving chronic dialysis at
a baseline interview in 1988.2- 6
Complete derivation details about the parent cohort were reported earlier.6 Patients were randomly derived from a stratified sample
of all Georgia patients who were aged 60 years or older and were registered
in the ESRD Network as of November 1987.5 The
ESRD Network is an administrative unit of the Health Care Financing Administration,
now the Centers for Medicare and Medicaid Services, which manages ESRD Medicare
in specified geographical areas. Stratification was based on race (white or
black), sex, and residence (urban or rural). Selected patients' demographics
(age, race, sex, or primary ESRD diagnosis) did not differ from those of the
total population of dialysis patients aged 60 years or older in the Georgia
ESRD Network. Patients were affiliated with 58 dialysis treatment facilities
across the state. All patients gave informed consent before entering the study,
which the Emory University Medical School Institutional Review Board approved.
Patients who were interviewed in 1988 and still lived in Georgia were
contacted for follow-up interviews in 1991. The response rate was 95%. For
inclusion in this analysis, patients had to be treated by in-center HD at
baseline and at the 3-year follow-up if they survived to that date, with the
exception of 2 patients who were receiving in-center HD at baseline and received
successful transplants. Of the original cohort, 3 patients who underwent home
HD, 42 who underwent peritoneal dialysis, and 1 who recovered renal function
were ineligible for this analysis. Thus, we did not use an intent-to-treat
analysis. Of the remaining 303 patients treated by in-center HD, 24 could
not be included in this analysis because their dialysis shift changed after
the study began, 4 could not be included because insufficient information
was available about their dialysis shift times at baseline or follow-up, and
33 could not be included because of missing data on other variables included
in the multivariable analyses. Thus, 242 patients were available for inclusion
in this analysis. Two of these patients (morning shift) experienced failed
transplants and did not receive dialysis for only a few days.
Time of day for HD was recorded at baseline (1988) and follow-up (1991)
patient interviews.2 Patients whose in-center
HD was initiated between 6:00 and 11:00 AM were considered to have undergone
morning dialysis (n = 167). Patients whose in-center dialysis was initiated
between 11:30 AM and 4:00 PM were considered to have undergone afternoon dialysis
(n = 75). The HD patients included and excluded in the shift analyses reported
here did not have different demographic or health-status characteristics,
except that included patients were more likely to be male. Between included
and excluded patients, there was no significant difference in poststudy survival
or months of dialysis before study entry.
A comprehensive interview and a review of medical records were conducted
for all patients at the beginning of the study and for patients participating
in the 3-year follow-up.2 Characteristics of
HD (months of dialysis before study entry and number of hours of dialysis
per week) were confirmed by the dialysis facility, as was history of diabetes
mellitus as a primary cause of ESRD. Cardiovascular disease was defined as
a history of myocardial infarct, stroke, and treated or untreated hypertension.
Body mass index (BMI; computed as weight in kilograms divided by height in
meters squared) and serum albumin levels were also recorded. For further analyses,
BMI was dichotomized at more than 23 kg/m2 vs less than 23 kg/m2, according to the median of the study population. Months of dialysis
before study entry and hours of dialysis per week were dichotomized at the
median of the distribution for each (36 months and 10.5 hours, respectively).
Functional status was assessed on the basis of interview questions determining
to what extent patients were capable of simple activities (eg, climbing stairs
or walking around the block) and sedentary during the daytime. Functional
impairment was classified according to Guttman scaling as most severe, moderately
severe, least severe, or none.7 Patients' vital
status was monitored from the baseline interview in 1988 (beginning of the
study) to July 7, 1999. Death certificates were obtained to determine causes
The primary analysis used in this study was the Cox proportional hazards
model, in which continued survival in days after the beginning of the study
was modeled by using a variety of demographic and health predictors. Two patients
who received successful transplants were censored at the date of transplant,
and the only patient who survived was censored as of the study's ending date,
July 7, 1999. Binary variables were coded 1 for age at least 70 years, male
sex, black race, BMI more than 23 kg/m2, HD for at least 36 months,
HD at least 10.5 hours weekly, higher functional status (ie, least severe
or no functional impairment), diabetes as the primary cause of ESRD, and the
presence of cardiovascular comorbidity. Referent categories were coded 0 for
younger age, female sex, white race, lower BMI, fewer months of HD, fewer
hours of HD each week, lower functional status, absence of diabetes as the
cause of ESRD, and absence of cardiovascular comorbidity. Because our previous
work indicated that race, sex, and BMI had a significant interaction effect
on survival (ie, higher BMI was protective in black men and women and in white
men),8 2- and 3-way interaction terms for these
variables were included in the model. The effect of dialysis shift was examined
by using afternoon shift as the referent category. Examination of proportionality
assumptions in the Cox model indicated no significant time × risk factor
interactions for any variable in the model. Data were analyzed by using SAS
Version 8 (SAS Institute, Cary, NC).
Demographic and medical characteristics of the 242 patients as a function
of shift are shown in Table 1.
On average, HD patients who underwent morning dialysis survived more than
a year longer than patients who underwent afternoon dialysis (Figure 1). Patients who were dialyzed in the morning were more likely
to be black and have higher BMI, but there were no other statistically significant
differences between the 2 groups.
According to the Cox regression model, univariate tests of association
with patients' continued survival in days after the beginning of the study
are summarized in Table 2. Black
race, higher BMI, absence of cardiovascular comorbidity, higher functional
status, and morning dialysis shift were significantly associated with HD patients'
continued survival (all P<.05).
Table 2 presents the results
of the multivariable Cox model predicting survival as related to demographic
and health variables, as well as time of day of dialysis. As we have reported
elsewhere,8 the interaction of race, sex, and
BMI affects survival in this population. Higher functional status also predicted
patients' continued survival. Additionally, consistent with the data shown
in Table 1 and Table 2, the protective effect for patients' survival of the morning
shift was sustained in this multivariable model. Analysis of death certificates
indicated no differences between morning- and afternoon-shift patients for
any cause of death including cardiovascular (eg, cardiac arrest or cerebrovascular
accident; 55% vs 47%), infectious (eg, septicemia or pneumonia; 13% vs 17%),
or other causes (eg, malignancy, hyperkalemia, or elective withdrawal from
dialysis; 32% vs 36%; P = .47). Review of death certificates
indicated that only 3 patients voluntarily withdrew from HD treatment; all
3 were treated on a morning shift.
Usual weight gain between dialysis treatments was examined as an indicator
of compliance with dialysis regimen.9- 11
Between morning- and afternoon-shift patients, there was no significant difference
in interdialytic weight gain (mean [SD] gain, 2.06 [1.06] kg vs 1.97 [1.02]
kg, respectively; P = .81).
Analysis of changing comorbidities for individuals surviving to the
3-year follow-up and reinterviewed indicated that, although patients at follow-up
had lower functional status, lower BMI, and greater cardiovascular comorbidity,
these patterns occurred equally for morning- and afternoon-shift patients.
There was also no significant difference in hours of weekly dialysis throughout
the 3 years for individuals dialyzed in the morning vs the afternoon.
Both morning- and afternoon-shift patients showed a trend for decreased
albumin levels throughout the 3 years preceding follow-up (0.11 g/dL vs 0.70
g/dL, respectively), but the extent of this decrease did not significantly
differentiate patients in the 2 shifts (P = .19).
The risk factors we examined did not account for the higher survival
rate of elderly ESRD patients undergoing HD who were dialyzed during the morning.
There may be other selection factors that lead ESRD patients to be assigned
to or to select morning shift, but they remain unidentified.
If the time of day of dialysis significantly affects survival, some
speculation regarding putative mechanisms that underlie the effect may not
be premature. For example, biochemical clearance may be optimized during morning-shift
HD. In a group of 124 patients representing 4 shifts, Mattana et al12 reported that HD patients who were dialyzed late
in the day had relatively higher levels of potassium and phosphorus than those
undergoing HD earlier in the day, implying less effective dialysis, perhaps
owing to interaction with the evening meal. Although we did not have our population's
laboratory data on hyperkalemia or hyperphosphatemia and therefore were unable
to test this hypothesis, the possibility exists that some impairment of biochemical
clearance might have hastened mortality in patients being dialyzed in the
Optimization of medical interventions by introduction of treatments
at specific times of day is not without precedent. For example, nocturnal
administration of calcitriol reduced dialysis patients' risk of hypercalcemia.13 Perhaps the most dramatic examples of temporal optimization
of treatment are from chemotherapy.14,15
In metastatic colorectal carcinoma, for example, antitumor activity was maximized
and adverse effects minimized by the administration of 5-fluorouracil and
leucovorin during the early morning.16 Similar
time-of-day susceptibility to best chemotherapeutic response has been noted
for renal cell carcinoma.17 Although the mechanisms
underlying these chemotherapeutic observations are undoubtedly different from
those observed in our study, these results suggest that simple manipulations
of when treatments are instituted profoundly affect health, which assumes
increasing importance as more frequent and longer dialysis treatment is more
Sleep during HD, which is common, may also contribute to the continued
survival of patients treated in the morning. Individuals undergoing HD in
the morning are sleepier than those being dialyzed at other times of day,19 and the notion that sleep is vital for health and
that loss of sleep hastens morbidity and mortality has been borne out amply
by numerous epidemiologic human studies and experimental animal studies. For
example, studies from well-defined populations suggest that chronic short
sleep may be a harbinger for all causes of mortality.20,21
In experimentally sleep-deprived animals, sleep loss has been associated with
compromised thermoregulatory, immunologic, and, perhaps most relevant for
the current discussion, protein metabolic functions.22,23
Data such as these imply that sleep during morning dialysis may represent
a beneficial compensatory response to sleep loss imposed by early waking times.
This explanation assumes differences in the quantity or quality of intradialytic
sleep as a function of the time of day. This hypothesis remains to be verified
As a final caveat, our data indicating apparently protective effects
of morning dialysis were seen in an elderly cohort. Whether younger patients
undergoing HD would derive similar benefit remains unclear. For example, age-dependent
sleep pattern changes that occur in elderly individuals might predispose them
to the early-morning arising required for morning dialysis, and their increased
tendency to nap might also enhance the likelihood of their sleeping during
the procedure.24,25 To test this
hypothesis, studies examining the differential effect of HD shift in younger
populations would be required, and polysomnographic studies18
documenting the amount and quality of sleep occurring on various shifts would
also be necessary. Randomized clinical trials may be required to determine
definitively whether morning shift is protective for any age group. In the
interim, further observational studies using larger databases such as the
United States Renal Data System should investigate whether dialysis shift,
independent of cumulative hours of weekly dialysis or duration of dialysis
treatment,26 significantly affects dialysis
outcomes, including patients' survival.