clinical investigation www.kidney-international.

org

The time of onset of intradialytic hypotension

during a hemodialysis session associates see commentary on page 1269
with clinical parameters and mortality OPEN

David F. Keane1,2,3, Jochen G. Raimann1, Hanjie Zhang1, Joanna Willetts4, Stephan Thijssen1 and
Peter Kotanko1,5
1
Research Division, Renal Research Institute, New York, New York, USA; 2Medical Physics and Engineering, Leeds Teaching Hospitals NHS
Trust, Leeds, UK; 3Leeds Institute for Cardiovascular and Metabolic Medicine, University of Leeds, Leeds, UK; 4Fresenius Medical Care,
Global Medical Office, Waltham, Massachusetts, USA; and 5Division of Nephrology, Icahn School of Medicine at Mount Sinai, New York,
New York, USA

I
Intradialytic hypotension (IDH) is a common complication ntradialytic hypotension (IDH) associates with adverse
of hemodialysis, but there is no data about the time of outcomes in hemodialysis (HD) patients, including
onset during treatment. Here we describe the incidence of patient self-reported symptom burden, access failure,
IDH throughout hemodialysis and associations of time of cardiovascular events, and mortality.1–3 Despite being one
hypotension with clinical parameters and survival by of the most common complications associated with HD,
analyzing data from 21 dialysis clinics in the United States there is no consensus on the definition, means to preven-
to include 785682 treatments from 4348 patients. IDH was tion, and management of IDH.
defined as a systolic blood pressure of 90 mmHg or under The challenges in defining IDH have been reviewed else-
while IDH incidence was calculated in 30-minute intervals where,4 but generally definitions include some or all of the
throughout the hemodialysis session. Associations of time following: (i) a systolic blood pressure (SBP) nadir; (ii) an
of IDH with clinical and treatment parameters were absolute reduction in SBP; and (iii) required interventions.
explored using logistic regression and with survival using Basic physiology suggests that reductions in blood pressure
Cox-regression. Sensitivity analysis considered further IDH during HD can be related to a combination of volume
definitions. IDH occurred in 12% of sessions at a median reduction, impaired cardiac output response to fluid removal,
time interval of 120-149 minutes. There was no notable and changes in total peripheral resistance.5 Unfortunately, it is
change in IDH incidence across hemodialysis intervals not trivial to comprehensively monitor these parameters, so
(range: 2.6-3.2 episodes per 100 session-intervals). Relative studies are scarce and include small numbers of patients.6
blood volume and ultrafiltration volume did not notably Interventions for managing IDH are lacking. There is
associate with IDH in the first 90 minutes but did thereafter. increasing evidence that use of cooled dialysate can prevent
Associations between central venous but not arterial IDH7 but as yet is not universally adopted. Other in-
oxygen saturation and IDH were present throughout terventions, including use of relative blood volume (RBV),8
hemodialysis. Patients prone to IDH early as compared to biofeedback-controlled ultrafiltration rate (UFR), and/or
late in a session had worse survival. Sensitivity analyses dialysate sodium9 and pharmacological management,10 all
suggested IDH definition affects time of onset but other lack a sufficient evidence base for widespread acceptance.
analyses were comparable. Thus, our study highlights the Numerous factors have been shown to associate with IDH
incidence of IDH during the early part of hemodialysis occurrence. These include demographic factors, comorbid-
which, when compared to later episodes, associates with ities, treatment prescriptions, serum osmolality, antihyper-
clinical parameters and mortality. tensive medications, and anemia.11–13 RBV at the time of IDH
Kidney International (2021) 99, 1408–1417; https://doi.org/10.1016/ has been shown to have significant intersubject variability, but
j.kint.2021.01.018 for a given individual, there appears to be a relatively stable
KEYWORDS: blood volume; hemodialysis; intradialytic hypotension; oxime- critical RBV preceding IDH.14 There are some data to suggest
try; survival that central venous oxygen saturation (ScvO2) decreases more
Copyright ª 2021, International Society of Nephrology. Published by in sessions where IDH occurs.15
Elsevier Inc. This is an open access article under the CC BY license (http://
Studies investigating IDH frequently define sessions based
creativecommons.org/licenses/by/4.0/).
on whether an IDH event occurs, or define patients as being
prone to IDH or not, but little attention has been given to
how the incidence of IDH onset varies within the HD session.
We aimed to: (i) characterize the incidence of IDH in discrete
Correspondence: David F. Keane, Renal Medicine, Leeds Teaching Hospitals time intervals within an HD session; (ii) investigate whether
Trust, Leeds, UK. E-mail: [email protected] any routinely collected parameters were associated with the
Received 1 September 2020; revised 17 December 2020; accepted 13 time of IDH; and (iii) consider whether the time of IDH is
January 2021; published online 17 February 2021 associated with mortality.

1408 Kidney International (2021) 99, 1408–1417

DF Keane et al.: Timing of intradialytic hypotension clinical investigation

METHODS study period. Comorbidities are defined by clinicians responsible for

Study design the patient’s care, and those extracted included presence or absence
This was a retrospective, observational cohort study utilizing of chronic heart failure, diabetes mellitus, and peripheral arterial/
routinely collected treatment data from the network of Renal vascular disease. Antihypertensive medication use was defined as
Research Institute clinics across the United States. We extracted data those patients with at least one open antihypertensive prescription in
from all dialysis sessions in 21 clinics between January 1, 2017 and the first 30 days of the study period. Mortality and cause of death
October 31, 2019 from both incident and prevalent patients, with no data come from the end-stage renal disease death notification form
exclusion criteria. Parameters extracted included demographic and (CMS 2746), and accompanying categories for cause of death were
laboratory variables, comorbidities, HD prescriptions, and blood used to define those deaths as cardiovascular related or not.
pressure, as well as hematocrit, RBV, and oxygen saturation (SO2), Intradialytic measurements. The CLM measures hematocrit,
measured by the Crit-Line monitor (CLM; Fresenius Medical Care SO2, and RBV every minute. SO2 was included as arterial oxygen
North America, Waltham, MA). The number of patients and sessions saturation (SaO2) for all sessions where vascular access was via a
in these clinics during the study period determined the sample size. fistula or graft and as ScvO2 for all sessions utilizing a central venous
The study was deemed to be exempt from review by the Western catheter. Where measurement of ScvO2 was available, we also
Institutional Review Board. To comply with the US Health Insurance calculated the estimated upper-body blood flow (eUBBF) in L/min
Portability and Accountability Act definition of a deidentified data throughout the treatment. This was based on the following formula:
set, patient age was capped at 90 years and dates were transformed to
a relative timeline. The study was reported in accordance with ¼ SaO2 
eUBBF
Strengthening the Reporting of Observational Studies in Epidemi- ½ð100  oxygen consumptionÞ = ðK  hemoglobin  ScvO2 Þ
ology guidance.16
where SaO2 was assumed to be 92%, ScvO2 was measured by CLM,
hemoglobin in g/L was calculated as CLM-measured hematocrit/
Definitions of data analyzed 2.94, K, the amount of oxygen (in ml) bound per g of hemoglobin,
IDH. An intradialytic SBP of <90 mm Hg has been shown to was 1.34, and oxygen consumption was calculated using de-
have the greatest association with mortality,17 and we used this as our mographic information.18 Further details on the assumptions un-
primary definition of IDH. Acknowledging the lack of consensus on derlying this formula can be found in previous work from this
IDH definition, we undertook sensitivity analyses with 3 other def- group.19
initions of IDH. Assimon and Flythe4 describe 3 main components To account for the fact that UFR can vary during a session,
of most definitions: characteristics of blood pressure behavior particularly in a session where IDH occurs, we defined the UFR for a
(largely a decline and/or a nadir value), interventions administered, given session as the mean of all recorded UFRs preceding an IDH
and patient-reported symptoms. Furthermore, it is highlighted that episode, where an episode occurred, and the mean of all recorded
definitions based on a nadir value only are likely to include events UFRs in a session where it did not. UFR was normalized to post-
that relate to chronic hypotension, which will skew results to earlier dialysis body weight and reported as ml/hour per kg.
times. Our data set had no information on patient-reported symp- Intradialytic timing data. Given that blood pressure mea-
toms, so we aimed to cover the other components using 3 additional surements were made every 30 minutes and on indication, it was not
definitions: possible to define IDH any more accurately than to have occurred
(i) A decline in SBP >30 mm Hg from predialysis levels. within a 30-minute interval. Therefore, we divided each session into
(ii) A combination of a nadir SBP <90 mm Hg plus a decline in 9 intervals, corresponding to 8 30-minute intervals from 0 up to 240
SBP >30 mm Hg from predialysis levels. minutes and 1 interval for times exceeding 240 minutes (Figure 1).
(iii) Nursing intervention based on fluids administered during
dialysis.
For the definition based on fluids administered, we removed the Data analysis
final 5 minutes of treatment data to avoid inclusion of rinsing fluids, Time of IDH onset. The primary objective was to describe the
calculated the modal value for fluids administered per session at each incidence of IDH throughout HD sessions. To calculate the incidence
center (assumed to be the priming fluid), and defined IDH as any of IDH in any given interval, we divided the total number of IDH
administered fluid of at least 50 ml once the total fluid administered events in an interval by the total number of sessions where at least
surpassed the mode for the given center. one blood pressure measurement had been made during that in-
Individual subjects were defined as IDH-prone when an IDH terval. Only the first case of IDH in any HD session was included in
episode occurred in >30% of their sessions. For secondary analyses, the analysis. To avoid bias related to subject-level variation in fre-
we also stratified all patients and sessions based on the timing of quency of IDH, we calculated the incidence for each session-interval
IDH. All patients with at least one IDH event were classified as prone at the subject level and then averaged across subjects. IDH incidence
to early-onset IDH if more than half of recorded IDH episodes was reported in each of the 9 intervals as the number of IDH events
occurred before 120 minutes and late onset otherwise. All HD ses- per 100 sessions-intervals.
sions where an IDH session occurred were classified as early or late Intradialytic measurements. Second, we investigated whether
onset based on whether the first IDH episode in the session was hemodynamic parameters measured by CLM and ultrafiltration
before or after 120 minutes. volume (UFV) were associated with imminent IDH and whether this
Clinical and demographic data. Blood pressure measurements depended on time into an HD session. We extracted RBV, SO2,
were made pre-HD and post-HD, every 30 minutes during the eUBBF, and UFV at the start of each session-interval and compared
treatment, and when clinically indicated, using automated oscillo- parameters between those sessions-intervals where IDH occurred
metric devices connected to the dialysis machine. Age and dialysis and those where it did not occur in that session-interval. To account
vintage were taken from the date of the first treatment session in the for repeated measures on individual subjects, a separate mixed effects

Kidney International (2021) 99, 1408–1417 1409

clinical investigation DF Keane et al.: Timing of intradialytic hypotension

Treatment start

Treatment end 30 min

240 min 60 min

Automated BP measurement

BP measurement on indica on

90 min
210 min

180 min
120 min

150 min

Figure 1 | Hemodialysis sessions-intervals used to define incidence of intradialytic hypotension across a treatment session. All
sessions routinely had automated blood pressure (BP) measured every 30 minutes, whereas BP measured on indication could occur at any
point in each session interval.

regression model was used at each time point with RBV, SO2, or UFV ratios and 95% confidence intervals. Two-sided P values of <0.05
as the dependent variable, the presence or absence of IDH as a were considered statistically significant. All statistical analyses were
dummy predictor, and subject as a random effect. This allowed us to performed using R 3.5.0.20
compare RBV and SO2 for those patients about to experience an IDH
episode with those who were stable for at least the following 30
minutes. RESULTS
Associations between time of IDH and clinical variables. To Study cohort
investigate associations between patient, treatment, or HD pre- The extracted data set included 4352 patients with 790,317
scription variables and the time of IDH, we constructed mixed sessions from 21 dialysis units. The number of patients and
effects logistic regression models, to account for repeated mea- sessions included in each part of the analysis can be seen in
surements on patients and the possibility of practice pattern dif- Figure 2. Patients prone to IDH were more likely to be older,
ferences between centers. Early/late onset of IDH was used as the White, male, and comorbid, have longer dialysis vintage, and
dependent variable, participant and treatment center were random less likely to be prescribed an antihypertensive, whereas ses-
effects, and the fixed effects were chosen based on plausibility and sions where IDH occurred had longer treatment times with
previous literature. Univariate analysis was performed for all
lower UFR and were linked to higher BMI and lower systolic
predictor variables, followed by 2 multivariate models: (i) those
variables available at all sessions (age, sex, body mass index
and diastolic blood pressure and pulse pressure (Table 1).
[BMI], comorbidities, interdialytic weight gain (IDWG), UFR,
SBP, dialysis vintage, and dialysate calcium [DCa]); and (ii) those Time of IDH onset
variables from model (i) plus albumin and dialysate to serum Using our primary definition of IDH (nadir SBP <90
sodium gradient, which are only measured one session per mm Hg), the median session-interval for first IDH in a ses-
month.
sion was 120 to 149 minutes. There was no clear trend in IDH
Survival analysis. We separately considered associations be-
incidence across session intervals, with a range of 2.6 to 3.2
tween time of IDH and both all-cause and cardiovascular mor-
tality. We plotted Kaplan-Meier survival curves for patients prone
episodes per 100 sessions-intervals (Figure 3).
to early-onset IDH and late-onset IDH and generated Cox pro-
portional hazard models for early versus late IDH group, adjusted Intradialytic measurements
for age, sex, BMI, chronic heart failure, and diabetes. Patients were Intradialytic RBV, UFV, SO2, and eUBBF results are plotted in
censored at the end of the study period, on transfer to a different Figure 4, separated based on whether IDH occurs in the
treatment modality, or out of the group of Renal Research Institute
subsequent 30 minutes or not. There were no clinically sig-
clinics.
nificant differences in RBV and UFV between sessions-
Statistical analysis. Incidence of IDH was described and
summarized using the median session interval for first IDH episode.
intervals where IDH is imminent compared with those
All regression analyses were based on complete case analysis. Linear where no episode occurs until 90 minutes, after which RBV
regression models were presented graphically as means  95% and UFV are both notably lower preceding IDH. There was a
confidence intervals for the 2 dummy-coded conditions representing clinically significant reduction in ScvO2 and eUBBF preceding
presence or absence of IDH in the following 30 minutes. Logistic and IDH at all time points, whereas no demonstrable difference
Cox proportional hazard regression models were presented as odds was observed for SaO2.

1410 Kidney International (2021) 99, 1408–1417

DF Keane et al.: Timing of intradialytic hypotension clinical investigation

Full data set

4352 Paents
790,317 Sessions
No intradialyc
treatment informaon
• 4 Paents
• 4635 Sessions
Cohort 1
Time of IDH onset
• 4348 Paents
• 785,682 Sessions
Paents and sessions Paents and sessions
with no CLM data with no IDH
• 1557 Paents • 1348 Paents
• 588,156 Sessions • 716,710 Sessions

Cohort 2 Cohort 3
Intradialyc measurements (i) Associaons with clinical variables
• 2791 Paents • 3000 Paents
• 197,526 Sessions • 68,972 Sessions
(ii) Survival analysis
• 1890 Paents prone to late-onset IDH
• 1110 Paents prone to early-onset IDH

Figure 2 | Patient flow diagram. CLM, Crit-Line monitor; IDH, intradialytic hypotension.

Associations between time of IDH onset and clinical variables of IDH (9% of total sessions), a wider range of values (0.1–2.4
Among demographic factors, being prone to IDH, increasing episodes per 100 sessions-intervals), and a slightly earlier
age, female sex, and lower BMI were all associated with early- median session-interval of 90 to 119 minutes.
onset IDH in adjusted models, whereas among treatment We repeated regression and survival analyses using a
factors, higher DCa and lower IDWG, UFR, and SBP showed combined definition of nadir SBP <90 mm Hg and a decline
similar tendencies (Table 2). in SBP $30 mm Hg, which produced comparable results to
those obtained using the primary defintion (Supplementary
Survival analysis Figures S1 and S2 and Supplementary Tables S3 and S4).
During a median follow-up of 628 days, there were 743
deaths, of which 325 were cardiovascular related. Unadjusted DISCUSSION
Kaplan-Meier analysis showed significantly higher all-cause We have demonstrated that IDH, defined as SBP <90
and cardiovascular mortality for those patients who tended mm Hg, occurs throughout HD at a relatively stable rate. This
to have IDH early in a session (P < 0.001; Figure 5). Adjusted pattern is dependent on how IDH is defined, reinforcing the
multivariate Cox analysis showed poorer survival in patients need for careful consideration of IDH definition in research
prone to early IDH (all-cause mortality hazard ratio, 1.4; 95% and clinical management. Furthermore, our results suggest
confidence interval, 1.2–1.7; cardiovascular mortality hazard that the ability of RBV and UFV to discriminate imminent
ratio, 1.6; 95% confidence interval, 1.2–2.2; Table 3). IDH onset is dependent on the time into an HD session.
Finally, we were able to show differences in clinical and
Sensitivity analysis treatment variables as well as survival between patients prone
Using a decline-based definition ($30 mm Hg) of IDH, the to IDH early and late in a session.
incidence was significantly higher than all other definitions
(total incidence, 51% of sessions; range of 9.4–14.2 episodes Time of IDH onset
per 100 sessions-intervals), although the median session- Despite a general assumption that most IDH episodes occur
interval did not change (Figure 3). Adding a reduction of at toward the end of a treatment,21 there are little published data
least 30 mm Hg from predialysis levels to the definition of on IDH timing. Zucchelli and Santoro described hemody-
IDH significantly reduced the total incidence (9% of sessions) namic monitoring of IDH-prone patients, showing a mean
and the incidence of IDH in the early part of HD, with only time of IDH of 178 minutes into a session.5 This is notably
1.0 and 1.7 incidents per 100 session intervals in the first and later than we found, although the study does not have a clear
second intervals, compared with 3.2 and 3.1 using our pri- definition of IDH and does not state whether it only includes
mary definition of IDH (Figure 3). The median session- the first IDH in any given session.
interval for first IDH increased by 30 minutes, to 150 to Intradialytic blood pressure changes, which are associated
179 minutes. Using fluid administered as a marker of in- with IDH, have been better characterized. A biphasic reduction
terventions to address IDH produced slightly lower incidence in SBP during HD has been described, consisting of a rapid

Kidney International (2021) 99, 1408–1417 1411

clinical investigation DF Keane et al.: Timing of intradialytic hypotension

Table 1 | Patient and treatment characteristics for the whole cumulative UFV at 30, 60, and 90 minutes between sessions
cohort and stratified by whether the patient is prone to IDH where IDH is imminent and where it is not, indicating
(>30% sessions) for patient data or whether the session average UFR was not notably different at these time points.
included an IDH episode for session data This could potentially point to the other prime driver of fluid
Patient data Total Not IDH-prone IDH-prone out of the circulation: the rapid changes in extracellular
No. of patients 4348 3799 (87) 549 (13) osmolality at the start of dialysis, which are associated with
Male sex 2524 (58) 2232 (59) 292 (53) osmotic fluid shifts and blood pressure instability.12
Dialysis vintage, yr 3.1 (4.1) 2.9 (3.9) 4.7 (5.2) The use of RBV for prediction and prevention of IDH is
Age, yr 62.1 (15) 61.7 (15.6) 64.9 (15.0) well described, but uncertainties around application remain.24
Race
White 1936 (45) 1656 (44) 280 (51) The RBV profiles observed herein are consistent with previ-
Black 1475 (34) 1303 (34) 172 (31) ous data25,26 and mirror the patterns in SBP described above.
Other 156 (4) 136 (4) 20 (4) The novel aspect in these data is that, although differences
Unknown 781 (18) 704 (19) 77 (14)
were statistically significant throughout HD, differences in
Comorbidity
CHF 423 (10) 335 (9) 88 (16) RBV preceding IDH are so small to be clinically insignificant
Diabetes 978 (23) 826 (22) 152 (28) during the early period of a session. Later in the session, RBV
PAD-PVD 204 (5) 156 (4) 48 (9) is notably reduced preceding an IDH episode.
Antihypertensive use 1054 (24) 969 (26) 85 (15)
Early-onset IDH 1253 (29) 886 (23) 367 (67)
Literature around intradialytic SO2 measurement is more
limited. ScvO2 has been shown to reduce with increasing
Session data Total No IDH session IDH session
UFV27,28 and to associate with survival,29 whereas variability
No. of sessions 785,682 690,070 (88) 95,612 (12) in SaO2 and ScvO2 is predictive of IDH occurrence.30 Our
BMI, kg/m2 28.7 (7.8) 28.5 (7.7) 30.2 (8.9) results demonstrate that ScvO2 and eUBBF, but not SaO2, are
IDWG, kg 2.1 (3.0) 2.1 (2.9) 2.2 (3.6)
Treatment time, min 220 (32) 221 (32) 226 (35)
notably reduced preceding an IDH event. This could be
UFR, ml/h per kg 8.8 (9.6) 8.9 (9.1) 7.9 (14.5) explained by the fact that ScvO2 and eUBBF (strongly related
Vascular access to ScvO2) can indicate decreased cardiac output and subse-
Fistula/graft 638,337 (81) 563,647 (82) 74,690 (78) quent inadequate tissue oxygenation, unlike SaO2, which is
Catheter 147,342 (19) 126,420 (18) 20,922 (22)
Blood flow, ml/min 421 (54) 421 (54) 417 (54) more reflective of pulmonary function.28,31
Pre-HD SBP, mm Hg 149 (27) 152 (27) 128 (23) These data may help to refine how CLM-measured RBV
Pre-HD DBP, mm Hg 78 (16) 79 (16) 69 (16) and SO2 are utilized and to better define outcomes to help
Post-HD SBP, mm Hg 138 (25) 141 (25) 114 (20) build the evidence base necessary for widespread adoption.
Post-HD DBP, mm Hg 73 (15) 74 (15) 62 (13)
Pulse pressure, mm Hg 71 (21) 72 (21) 59 (21)
Dialysate temperature,  C 36.6 (0.4) 36.6 (0.4) 36.5 (0.5) Associations between time of IDH onset and clinical variables
Dialysate sodium, mEq/L Patients prone to IDH in general were more likely to be fe-
<136 11,735 (1) 10,291 (1) 1444 (2)
male and comorbid, and have lower blood pressure, higher
136 18,067 (2) 15,670 (2) 2397 (3)
137 732,534 (93) 643,320 (93) 89,214 (93) BMI, longer dialysis vintage, and higher IDWG (Table 1),
138 12,617 (2) 11,254 (2) 1363 (1) which is consistent with previous literature.32 Most factors
>138 10,729 (1) 9535 (1) 1194 (1) associated with IDH occurrence per se were associated with
Dialysate calcium, mEq/L
early-onset IDH, with the exceptions being higher BMI and
2 10,513 (1) 8586 (1) 1924 (2)
2.25 316,381 (40) 276,154 (40) 40,228 (42) IDWG and lower DCa, which were associated with IDH in
2.5 356,720 (45) 318,235 (46) 38,487 (40) general while being associated with late-onset IDH.
3 101,379 (13) 86,427 (13) 14,952 (16) BMI is known to have complex associations with outcomes
3.5 675 (0.1) 655 (0.1) 20 (0.02)
in kidney disease,33 and this is likely to extend to the rela-
BMI, body mass index; CHF, chronic heart failure; DBP, diastolic blood pressure; HD,
hemodialysis; IDH, intradialytic hypotension; IDWG, interdialytic weight gain; PAD,
tionship between BMI and IDH. BMI is linked to IDWG and
peripheral artery disease; PVD, peripheral vascular disease; SBP, systolic blood therefore UFR, but patients with higher BMI appear to be
pressure; UFR, ultrafiltration rate. able to better tolerate larger UFV and fluid depletion.34 Two
Data are number (SD) for continuous variables and number (%) for categorical
variables. studies showing that BMI had no association with IDH,
whereas increased fat and reduced lean tissue were associated
with IDH,35,36 may point to the problems in BMI as a sur-
early decline across the first quarter of the treatment and a rogate for body composition. IDWG is independently linked
later, gentler decline, although with significant interindividual to IDH37 and mortality38; however, IDWG is also strongly
variation.22,23 This pattern could be linked to the relatively high linked to nutrition, and this has been shown to impact on the
incidence of IDH early in a session observed herein. relationship between IDWG and mortality.39 Body habitus
and nutrition may be important factors in future work
Intradialytic measurements considering the timing of IDH.
Clearly, the rate of removal of fluid from the circulation by Serum calcium is involved in myocardial contraction and
ultrafiltration is central to the pathogenesis of IDH. However, therefore DCa may have a role in maintenance of blood
our results show no clinically significant difference in the pressure during HD.40 Clinical trials have demonstrated

1412 Kidney International (2021) 99, 1408–1417

DF Keane et al.: Timing of intradialytic hypotension clinical investigation

Decline 30 mm Hg and nadir 90 mm Hg Fluid bolus intervenon

IDH episodes per 100

3.0 3.0

sessions-intervals
2.5 2.5
2.0 2.0
1.5 1.5
1.0 1.0
0.5 0.5
0.0 0.0

40
9

90 9

90 9
12 19

12 19
15 49

18 9

21 09

21 09
40

29
29

15 49

0
18 9
–5

–8

–5

–8
17

24

24
17

>2
–1

–1
1

2
>2

0–
0–

1
30

60

30

60
0–

0–

0–

0–

0–
0–

0–
0–
Nadir <90 mm Hg Decline 30 mm Hg
IDH episodes per 100

3.5 16.0
sessions-intervals

3.0 14.0
2.5 12.0
10.0
2.0
8.0
1.5 6.0
1.0 4.0
0.5 2.0
0.0 0.0

40
9

90 9
12 19

21 09
29

15 49

0
18 9
9

90 9
12 19

15 49

18 79

21 09

0
29

–5

–8

24
17
–5

–8

24

>2
–1

2
0–

1
–1

2
0–

30

60

0–
0–

0–
0–
30

60

0–

0–

0–

0–

Time into HD session (min) Time into HD session (min)

Figure 3 | Intradialytic hypotension (IDH) episodes per 100 sessions-intervals at risk, with time into hemodialysis (HD) session using
IDH definitions of decline in systolic blood pressure (SBP) >30 mm Hg and nadir SBP <90 mm Hg; fluids administered; nadir SBP <90
mm Hg; and decline in SBP >30 mm Hg.

increasing DCa improves blood pressure but not symptoms DCa may be related to patients being prone to IDH. This may
associated with IDH41 and that profiling DCa leads reduces explain the association with early-onset IDH (Table 2) and
intradialytic blood pressure instability.42 The risk of indica- the increase in both higher (3 mEq/L) and lower (2.25 mEq/
tion bias with our data must be considered as prescription of L) prescriptions in sessions where IDH occurred (Table 1).

100 1.4
eUBBF (L/min)

1.3
98
1.2
96 1.1
RBV (%)

94 1
0.9
92 65
90 63
61
ScvO2 (%)

88
59
3000
57
2500
55
2000 96.5
UFV (ml)

1500
96.0
SaO2 (%)

1000
95.5
500

0 95.0
0 30 60 90 120 150 180 210 240 0 30 60 90 120 150 180 210 240
Time (min) Time (min)
Figure 4 | Intradialytic relative blood volume (RBV), ultrafiltration volume (UFV), arterial oxygen saturation (SaO2), central venous
oxygen saturation (ScvO2), and estimated upper-body blood flow (eUBBF). Each data point represents an average value at the start of
every session-interval, stratified by whether intradialytic hypotension (IDH) occurred in the subsequent 30 minutes or not, with error bars
representing 95% confidence intervals. Comparisons of each variable at all time points showed statistically significant differences (P < 0.05).

Kidney International (2021) 99, 1408–1417 1413

clinical investigation DF Keane et al.: Timing of intradialytic hypotension

Table 2 | Mixed-effects logistic regression for the odds of IDH occurring in the first half of a session compared with later in the
treatment
Univariate Multivariate (model 1) Multivariate (model 2)
Variable OR 95% CI OR 95% CI OR 95% CI
a a
IDH-prone 2.61 2.35–2.90 1.83 1.67–2.01 1.67 1.43–1.94a
Age, yr 1.02 1.01–1.02a 1.02 1.01–1.02a 1.02 1.01–1.02a
Male sex 0.88 0.80–0.96a 0.76 0.71–0.82a 0.81 0.71–0.94a
BMI, kg/m2 0.98 0.97–0.98a 0.99 0.98–0.99a 0.98 0.97–0.99a
CHF 1.20 1.05–1.38 1.11 0.99–1.25 1.16 0.95–1.42
Diabetes 0.90 0.81–1.00a 0.99 0.91–1.08 1.09 0.93–1.26
PAD-PVD 1.14 0.95–1.38 0.90 0.77–1.05 0.77 0.59–1.00a
IDWG, kg 0.90 0.89–0.91a 0.94 0.93–0.95a 0.90 0.86–0.95a
Ultrafiltration rate, ml/h per kg 0.97 0.97–0.98a 0.99 0.99–1.00a 0.99 0.98–1.01
Pre-SBP, mmHg 0.97 0.97–0.97a 0.98 0.97–0.98a 0.97 0.97–0.98a
Dialysis vintage, yr 1.00 1.00–1.02 1.00 0.99–1.01 1.01 0.99–1.02
Dialysate calcium, mEq/L 1.13 1.02–1.25a 1.19 1.08–1.32a 1.38 1.07–1.77a
Albumin 0.55 0.49–0.61a — — 0.90 0.76–1.08
Dialysate to serum Na gradient, mEq/L 1.01 0.99–1.03 — — 1.01 0.99–1.03
BMI, body mass index; CHF, chronic heart failure; CI, confidence interval; IDH, intradialytic hypotension; IDWG, interdialytic weight gain; OR, odds ratio; PAD, peripheral artery
disease; PVD, peripheral vascular disease; SBP, systolic blood pressure.
a
P < 0.05.
Adjusted analysis for model 1 includes all covariates available for all treatments (785,682 sessions), whereas for model 2, it includes all variables from model 1 plus albumin
and sodium gradient in a limited number of treatments due to availability of these data (15,994 sessions).

The association between UFR and mortality is well or potentially to the reduction of UFR prescriptions in un-
described,43 but there are surprisingly little data describing stable patients.
associations between UFR and IDH, although Dialysis Out-
comes and Practice Patterns Study (DOPPS) data showed Survival analysis
UFR >10 ml/h per kg is associated with greater odds of Kaplan-Meier analysis (Figure 5) and Cox analysis suggest
IDH.44 We found UFR to be lower in patients prone to IDH that early-onset IDH is associated with both all-cause and
and increasing UFR to be associated with less chance of early- cardiovascular mortality. Comorbidities, including chronic
onset IDH. This could be related to normalization of UFR to heart failure and autonomic dysfunction, which compromise
body weight (given BMI was higher in IDH-prone patients) the ability to compensate for fluid removal on dialysis, may be

Figure 5 | Kaplan-Meier curves for patients who tended to have intradialytic hypotension (IDH) in the first half of a session compared
with those who tended to have IDH in the second half of a session. Data presented for all-cause and cardiovascular mortality, with
accompanying 95% confidence intervals.

1414 Kidney International (2021) 99, 1408–1417

DF Keane et al.: Timing of intradialytic hypotension clinical investigation

Table 3 | Cox proportional hazards analysis of the association of early-onset IDH with both all-cause and cardiovascular
mortality
All-cause mortality Cardiovascular mortality
Univariate Multivariate Univariate Multivariate
Variable OR 95% CI OR 95% CI OR 95% CI OR 95% CI
a a a
Early-onset IDH 2.26 1.95–2.61 1.40 1.16–1.69 2.22 1.78–2.78 1.64 1.23–2.18a
IDH-prone 2.36 2.00–2.79a 1.52 1.22–1.88a 1.80 1.37–2.37a 1.25 0.88–1.76
Age, yr 1.04 1.04–1.05a 1.03 1.03–1.04a 1.03 1.02–1.04a 1.02 1.01–1.03a
Dialysis vintage, yr 1.00 0.99–1.01 1.00 1.00–1.00 1.00 1.00–1.00 1.00 1.00–1.00
Male sex 0.99 0.86–1.15 1.24 1.03–1.48a 1.05 0.84–1.32 1.36 1.03–1.80a
Race (White vs. non-White) 1.78 1.52–2.09a 1.58 1.32–1.90a 1.33 1.04–1.69a 1.23 0.93–1.61
BMI, kg/m2 0.97 0.95–0.98a 0.97 0.96–0.99a 0.97 0.95–0.98a 0.97 0.95–0.99a
CHF 1.26 1.03–1.54a 1.16 0.93–1.45 1.36 1.01–1.83a 1.32 0.95–1.84
Diabetes 1.03 0.88–1.21 0.88 0.73–1.06 1.02 0.80–1.30 0.86 0.65–1.14
Albumin, g/dl 0.25 0.22–0.29a 0.27 0.22–0.34 0.31 0.25–0.40a 0.35 0.25–0.49a
IDWG, kg 0.94 0.91–0.96a 1.00 0.07–0.94 0.95 0.91–1.00 1.01 0.96–1.07
BMI, body mass index; CHF, chronic heart failure; CI, confidence interval; IDH, intradialytic hypotension; IDWG, interdialytic weight gain; OR, odds ratio.
a
P < 0.05.

linked to early-onset IDH and the association with survival. Research Institute centers across the United States. We included
Nevertheless, the associations were maintained after adjust- all patients who dialyzed in these clinics during the study
ment for a limited number of potential confounders period, removing selection bias. Sex, dialysis vintage, age, and
(Table 3). race data are all comparable to US Renal Data System data sets,
although comorbidity and antihypertensive medication data
Sensitivity analysis appear to be low, likely due to missing data being given the
Sensitivity analyses demonstrate the importance of how IDH same status as a null response in the source database. IDH rates
is defined for both research and management of IDH. The per session for the SBP-based definitions are also extremely
addition of a minimal reduction in SBP to a nadir value, similar to previously reported data,17 providing further support
reducing the dependence on the initial SBP, predictably for the generalizability of the results. CLM data were not
reduced the number of IDH incidents early in a session, available for all patients, but the use of CLM is a unit level
although there was notable incidence in the first half of decision rather than based on patient need.
dialysis in all definitions used. We repeated all secondary Consistent with observational data, residual confounding
analyses, including survival analysis, using the combined must be considered. Associations between modifiable vari-
definition of a decline- and a nadir-based SBP in an attempt ables, such as HD prescriptions, and IDH are likely to include
to reduce the imapct of IDH episodes relating to chronic confounding by indication.
hypotension rather than dialysis-induced hypotension. All
results were robust to this alternative definition. Conclusion
We have demonstrated a broad distribution in the time of
Clinical implications onset of IDH throughout a treatment session, which is
IDH is likely the result of several mechanistic pathways, associated with clinical, treatment, and outcome variables,
although there is often little differentiation in treatment. including survival. Consideration of time of IDH alongside
Given the associations with mortality, intervening to prevent well-described underlying mechanisms could potentially help
early-onset IDH may have significant impact on outcomes. to differentiate IDH episodes and facilitate effective use of
Further research is needed to explore the mechanisms and interventions, something that remains a challenge.
potential interventions around IDH early in treatment, but
promising options could include minimizing the effect of
DISCLOSURE
large osmotic changes in the early stages of dialysis (reduced PK and ST hold stock in Fresenius Medical Care. JW is employed by Fresenius
blood flows and sodium modeling), preserving central blood Medical Care North America. JGR, HZ, ST, and PK are employees of the Renal
Research Institute. The Renal Research Institute is a wholly owned subsidiary of
volume (modified dialysate temperature, positional changes,
Fresenius Medical Care. All the other authors declared no competing interests.
and ultrafiltration profiles), and further investigation of the
early dialysis-induced hypoxemia demonstrated elsewhere45
and supported by this study. ACKNOWLEDGMENTS
DFK was funded by a National Institute for Health Research (NIHR)
Clinical Lectureship (ICA-CL-2017-03-017). This article presents
Strengths and limitations independent research funded by the NIHR. The views expressed are
This is the first study aiming to characterize the time of onset those of the authors and not necessarily those of the National Health
of IDH. It is based on a large cohort from the network of Renal Service (NHS), the NIHR, or the Department of Health and Social Care.

Kidney International (2021) 99, 1408–1417 1415

clinical investigation DF Keane et al.: Timing of intradialytic hypotension

SUPPLEMENTARY MATERIAL 16. von Elm E, Altman DG, Egger M, et al. The Strengthening the Reporting of
Supplementary File (PDF) Observational Studies in Epidemiology (STROBE) statement: guidelines for
Figure S1. Sensitivity analysis: definition of IDH of SBP <90 mm Hg reporting observational studies. Lancet. 2007;370:1453–1457.
17. Flythe JE, Xue H, Lynch KE, et al. Association of mortality risk with various
and a decline of SBP >30 mm Hg from predialysis levels. Kaplan-
definitions of intradialytic hypotension. J Am Soc Nephrol. 2015;26:724–734.
Meier curves for patients who tended to have IDH in the first half of a 18. Mehrpour O, Sheikhazadi A, Ghadyani MH, et al. Brain weight of Iranian
session compared with those who tended to have IDH in the second population; the first report. J Forensic Leg Med. 2010;17:426–431.
half of a session for all-cause (left) and cardiovascular (right) mortality. 19. Rosales LM, Zhang H, Mateo M, et al. Tracking arteriovenous fistula
Figure S2. Sensitivity analysis: definition of IDH of SBP <90 mm Hg maturation: a novel approach. Blood Purif. 2019;47:240–245.
and a decline in SBP >30 mm Hg from predialysis levels. Intradialytic 20. R Core Team. R: A Language and Environment for Statistical Computing.
Vienna, Austria: R Foundation for Statistical Computing; 2013. Available
relative blood volume (RBV), ultrafiltration volume (UFV), arterial
at: http://www.R-project.org/. Accessed April 1, 2018.
oxygen saturation (SaO2), central venous oxygen saturation (ScvO2), 21. Daugirdas JT. Measuring intradialytic hypotension to improve quality of
and estimated upper-body blood flow (eUBBF). Each data point care. J Am Soc Nephrol. 2015;26:512–514.
represents an average value at the start of every session-interval, 22. Dinesh K, Kunaparaju S, Cape K, et al. A model of systolic blood pressure
stratified by whether IDH occurred in the subsequent 30 minutes or during the course of dialysis and clinical factors associated with various
not, with error bars representing 95% confidence intervals. blood pressure behaviors. Am J Kidney Dis. 2011;58:794–803.
Table S1. Mixed-effects logistic regression for the odds of IDH occurring 23. Goldstein BA, Chang TI, Winkelmayer WC. Classifying individuals based
in the first half of a session compared with later in the treatment. IDH on a densely captured sequence of vital signs: an example using
defined as SBP <90 mm Hg and a decline in SBP >30 mm Hg. Adjusted repeated blood pressure measurements during hemodialysis treatment.
J Biomed Inform. 2015;57:219–224.
analysis for model 1 includes all covariates available for all treatments
(785,682 sessions), whereas for model 2, it includes all variables from 24. Schneditz D, Kron J, Hecking M. Anything goes? high time for smart
blood volume monitors. ASAIO J. 2018;64:697–700.
model 1 plus albumin and sodium gradient in a limited number of
25. Maduell F, Arias M, Massó E, et al. Sensitivity of blood volume monitoring for
treatments due to availability of these data (15,994 sessions). *P < 0.05.
fluid status assessment in hemodialysis patients. Blood Purif. 2013;35:202–208.
Table S2. Cox proportional hazards analysis of the association of early- 26. Yu SJ, Kim DH, Oh DJ, et al. Assessment of fluid shifts of body
onset IDH with both all-cause and cardiovascular mortality. IDH defined compartments using both bioimpedance analysis and blood volume
as SBP <90 mm Hg and a decline in SBP >30 mm Hg. *P < 0.05. monitoring. J Korean Med Sci. 2006;21:75–80.
27. Harrison LE, Selby NM, McIntyre CW. Central venous oxygen saturation: a
potential new marker for circulatory stress in haemodialysis patients?
REFERENCES Nephron Clin Pract. 2014;128:57–60.
1. McIntyre CW. Haemodialysis-induced myocardial stunning in chronic
28. Zhang H, Chan L, Meyring-Wösten A, et al. Association between intradialytic
kidney disease—a new aspect of cardiovascular disease. Blood Purif.
central venous oxygen saturation and ultrafiltration volume in chronic
2010;29:105–110.
hemodialysis patients. Nephrol Dial Transplant. 2018;33:1636–1642.
2. Chang T, Paik J, Greene T, et al. Intradialytic hypotension and vascular
29. Chan L, Zhang H, Meyring-Wösten A, et al. Intradialytic central venous
access thrombosis. J Am Soc Nephrol. 2011;22:1526–1533.
oxygen saturation is associated with clinical outcomes in hemodialysis
3. Shoji T, Tsubakihara Y, Fujii M, et al. Hemodialysis associated patients. Sci Rep. 2017;7:8581.
hypotension as an independent risk factor for two-year mortality in 30. Mancini E, Perazzini C, Gesualdo L, et al. Intra-dialytic blood oxygen
hemodialysis patients. Kidney Int. 2004;66:1212–1220. saturation (SO2): association with dialysis hypotension (the SOGLIA
4. Assimon MM, Flythe JE. Definitions of intradialytic hypotension. Semin study). J Nephrol. 2016;30:811–819.
Dial. 2017;30:464–472. 31. Meyring-Wösten A, Luo Y, Zhang H, et al. Intradialytic hypertension is
5. Zucchelli P, Santoro A. Dialysis-induced hypotension: a fresh look at associated with low intradialytic arterial oxygen saturation. Nephrol Dial
pathophysiology. Blood Purif. 1993;11:85–98. Transplant. 2018;33:1040–1045.
6. Feng Y, Zou Y, Zheng Y, et al. The value of non-invasive measurement of 32. Sands JJ, Usvyat LA, Sullivan T, et al. Intradialytic hypotension: frequency,
cardiac output and total peripheral resistance to categorize significant sources of variation and correlation with clinical outcome. Hemodial Int.
changes of intradialytic blood pressure: a prospective study. BMC 2014;18:415–422.
Nephrol. 2018;19:310. 33. Kalantar-Zadeh K, Rhee C, Chou J, et al. The obesity paradox in kidney
7. Mustafa RA, Bdair F, Akl EA, et al. Effect of lowering the dialysate disease: how to reconcile it with obesity management. Kidney Int Rep.
temperature in chronic hemodialysis: a systematic review and meta- 2017;2:271–281.
analysis. Clin J Am Soc Nephrol. 2016;11:442–457. 34. Keane D, Chamney P, Heinke S, et al. Use of the body composition
8. Reddan D, Szczech LA, Hasselbad V, et al. Intradialytic blood volume monitor for fluid status measurements in subjects with high body mass
monitoring in ambulatory hemodialysis patients: a randomized trial. index. Nephron. 2016;133:163–168.
J Am Soc Nephrol. 2005;16:2162–2169. 35. Tian M, Zha Y, Qie S, et al. Association of body composition and intradialytic
9. Leung KCW, Quinn RR, Ravani P, et al. Randomized crossover trial of hypotension in hemodialysis patients. Blood Purif. 2020;49:334–340.
blood volume monitoring-guided ultrafiltration biofeedback to reduce 36. Zhou Q, Wang J, Xie S, et al. Correlation between body composition
intradialytic hypotensive episodes with hemodialysis. Clin J Am Soc measurement by bioelectrical impedance analysis and intradialytic
Nephrol. 2017;12:1831–1840. hypotension. Int Urol Nephrol. 2020;52:953–958.
10. Chang TI. Impact of drugs on intradialytic hypotension: 37. Stefánsson BV, Brunelli SM, Cabrera C, et al. Intradialytic hypotension and
antihypertensives and vasoconstrictors. Semin Dial. 2017;30:532–536. risk of cardiovascular disease. Clin J Am Soc Nephrol. 2014;9:2124–2132.
11. Chou J, Kalantar-Zadeh K, Mathew A. A brief review of intradialytic 38. Flythe JE, Curhan GC, Brunelli SM. Disentangling the ultrafiltration rate–
hypotension with a focus on survival. Semin Dial. 2017;30:473–480. mortality association: the respective roles of session length and weight
12. Singh AT, Mc Causland FR. Osmolality and blood pressure stability during gain. Clin J Am Soc Nephrol. 2013;8:1151–1161.
hemodialysis. Semin Dial. 2017;30:509–517. 39. Kurita N, Hayashino Y, Yamazaki S, et al. Revisiting interdialytic weight
13. Larkin JW, Reviriego-Mendoza MM, Usvyat LA, et al. To cool, or too cool: gain and mortality association with serum albumin interactions: the
is reducing dialysate temperature the optimal approach to preventing Japanese Dialysis Outcomes and Practice Pattern Study. J Ren Nutr.
intradialytic hypotension? Semin Dial. 2017;30:501–508. 2017;27:421–429.
14. Barth C, Boer W, Garzoni D, et al. Characteristics of hypotension-prone 40. van der Sande FM, Cheriex EC, van Kuijk WHM, et al. Effect of dialysate
haemodialysis patients: is there a critical relative blood volume? Nephrol calcium concentrations on intradialytic blood pressure course in cardiac
Dial Transplant. 2003;18:1353–1360. compromised patients. Am J Kidney Dis. 1998;32:125–131.
15. Perazzini C, Bolasco PG, Corazza L, et al. Prediction of intradialytic 41. Alappan R, Cruz D, Abu-Alfa AK, et al. Treatment of severe intradialytic
hypotension based on oxygen saturation variations. Comput Cardiol. hypotension with the addition of high dialysate calcium concentration
2013:40:1223–1226. to midodrine and/or cool dialysate. Am J Kidney Dis. 2001;37:294–299.

1416 Kidney International (2021) 99, 1408–1417

DF Keane et al.: Timing of intradialytic hypotension clinical investigation

42. Kyriazis J, Glotsos J, Bilirakis L, et al. Dialysate calcium profiling during 44. Saran R, Bragg-Gresham JL, Levin NW, et al. Longer treatment time and
hemodialysis: use and clinical implications. Kidney Int. 2002;61:276–287. slower ultrafiltration in hemodialysis: associations with reduced mortality
43. Assimon MM, Flythe JE. Rapid ultrafiltration rates and outcomes among in the DOPPS. Kidney Int. 2006;69:1222–1228.
hemodialysis patients: re-examining the evidence base. Curr Opin 45. Campos I, Chan L, Zhang H, et al. Intradialytic hypoxemia in chronic
Nephrol Hypertens. 2015;24:525–530. hemodialysis patients. Blood Purif. 2016;41:177–187.

Kidney International (2021) 99, 1408–1417 1417