Interactions between intradialytic central venous oxygen saturation, relative blood volume, and all‐cause mortality in maintenance hemodialysis patients

In maintenance hemodialysis (HD) patients, low central venous oxygen saturation (ScvO2) and small decline in relative blood volume (RBV) have been associated with adverse outcomes. Here we explore the joint association between ScvO2 and RBV change in relation to all‐cause mortality.

K E Y W O R D S central venous oxygen saturation, hemodialysis, patient outcomes, relative blood volume, ultrafiltration rate

| INTRODUCTION
In maintenance hemodialysis (HD) patients, cardiovascular disease (CVD) is the leading cause of death. [1][2][3][4][5] In addition to traditional CVD risk factors, nontraditional risk factors, such as high ultrafiltration volume (UFV) and ultrafiltration rate (UFR), have been identified. 2,[5][6][7] Ultrafiltration frequently results in a decline in blood volume and hemodynamic stress, factors that have been associated with intradialytic hypotension (IDH) and poor perfusion of vital organs, such as heart, brain, gut, liver, and kidneys. 8,9 While it would be ideal to continuously assess organ perfusion during HD, we currently lack tools that can be applied routinely.
Central venous oxygen saturation (ScvO 2 ), a proxy for upper body blood flow (UBBF), is frequently used as a hemodynamic marker in non-dialysis populations. 10,11 In HD patients with a central venous catheter (CVC) as vascular access, ScvO 2 can be measured easily and continuously using devices approved by the U.S. Federal Drug Agency (FDA). Previous studies have shown that low ScvO 2 levels and high ScvO 2 variability are associated with high UFV 12 and poor HD patient outcomes. 13,14 Intradialytic relative blood volume (RBV) monitoring builds on continuous measurements of hematocrit (HCT) or serum protein. [15][16][17] RBV is used by some dialysis clinicians to guide fluid management. RBV curves with a decline of less than 1.39% per hour or, in another study, of less than 8% 3 h into the treatment have been associated with increased mortality. 18,19 Reviewing these studies reveals an apparent paradox: on one hand, a low ScvO 2 , which is related to higher UFV, is associated with increased mortality, while on the other hand, a steep decline in RBV, which is also related to higher UFV, is associated with a lower mortality. The goal of our study was to explore the joint association between ScVO 2 and RBV changes in relation to all-cause mortality in chronic HD patients. Additionally, to further investigate the association between SvcO 2 and fluid removal, and as RBV is directly associated with UFR, we explored the joint association of ScvO 2 and UFR in relation to all-cause mortality.

| Population and study design
This retrospective, multicenter cohort study was conducted in maintenance HD patients dialyzed in 17 U.S. facilities of the Renal Research Institute (RRI) between January 2012 and August 2015. In these clinics, the Crit-Line monitor (CLM; Fresenius Medical Care, Waltham, MA) is used as standard of care.
Patient characteristics, UFR, RBV, ScvO 2 , and other clinical factors were monitored during a 6-month baseline period that started with the first HD treatment with eligible ScvO 2 readings. Only patients with at least 10 HD treatments with eligible ScvO 2 recordings during baseline were included in the final analysis.

| Data eligibility
We included only patients with CVC as vascular access. To ensure appropriate data quality, treatments with a mean ScvO 2 above 80% were excluded, as such values are not consistent with central venous blood. A ScvO 2 > 80% is very unlikely, given the high oxygen extraction by the brain and the muscle tissue in the upper body, and the lower-than-normal hemoglobin (Hgb) concentration in dialysis patients. The most plausible reason for such high oxygen saturation levels is the intermittent use of a single needle in an arterio-venous access that was erroneously documented as a CVC use. Likewise, mean ScvO 2 levels below 25% were excluded as they are deemed incompatible with life. 20 A ScvO 2 < 25% is most likely due to a loose connection of the Crit-Line sensor to the measurement chamber. Lastly, ScvO 2 measurements concurrent with a RBV > 102% were excluded.

| Measurement of ScvO 2 and RBV
Intradialytic ScvO 2 was measured using the CLM. The CLM has been approved by the FDA for reporting HCT, Hgb, RBV, and oxygen saturation in the extracorporeal dialysis circuit. The CLM measures HCT and oxygen saturation 9000 times per minute and reports their averages once per minute.
Treatment-level RBV change was calculated as Per manufacturer, the CLM accuracy for oxygen saturation measurement is within 2%. The CLM measurements were stored on a server located in the dialysis facility and later transferred to the RRI research database.

| ScvO 2
Patients' mean, median, minimum, maximum, standard deviation, start-HD, and end-HD ScvO 2 were calculated per treatment and then averaged across all treatments per patient and subsequently across patients, as described previously. 14

| Clinical and laboratory data
Clinical data were retrieved from electronic health records. Continuous clinical variables were averaged over the 6-month baseline period. Body mass index was calculated using post-HD weight. Laboratory measurements (Spectra Laboratories, Rockleigh, NJ, USA) were downloaded to the RRI data warehouse and extracted to the study database.

| Comorbidities
Diabetes mellitus (DM), chronic obstructive pulmonary disease (COPD), and congestive heart failure (CHF) were defined using ICD-9 codes documented in the patients' electronic health records.

| Statistical analysis
Descriptive statistics comprised mean (± standard deviation; SD) for continuous variables and percentages for categorical variables. We stratified patients into four groups based on their ScvO 2 and RBV changes being above or below the respective medians of the study population. Groups were defined as follows: group 1, ScvO 2 above median and RBV change below median (i.e., a more negative RBV decline compared to median RBV decline); group 2, ScvO 2 and RBV change above median; group 3, ScvO 2 and RBV change below median; and group 4, ScvO 2 below median and RBV change above median. Survival times were compared between those groups using Kaplan-Meier analysis and Cox proportional hazards models. Cox models were adjusted for age, DM, and dialysis vintage. To evaluate the joint association between ScvO 2 and RBV changes with all-cause mortality, we also fitted Cox proportional hazard models with bivariate spline terms for ScvO 2 and RBV change, 21 the results of this analysis are presented as a contour plot (for a detailed interpretation see Figures 1 and 2, bottom panels).
Additional secondary analyses using the same methodology were performed to evaluate the interaction between ScvO 2 and UFR.
The study was approved by the New England Institutional Review Board (14-446) and conducted in accordance with the Declaration of Helsinki. To comply with the U.S. Health Insurance Portability and Accountability Act (HIPPA) definition of a de-identified data set, patient age was capped at 90 years and dates were transformed to a relative timeline.

| Patient baseline characteristics
Between January 1, 2012, and August 31, 2015, 579 patients with CVC as vascular access received 7937 HD treatments at 17 US RRI clinics. Of these, 216 patients with 5231 dialysis sessions met the inclusion criteria of at least 10 baseline HD treatments with ScvO 2 readings. Their mean age was 62.2 ± 15.7 years, the dialysis vintage was 2.7 ± 4.5 years, 55% were whites, 47% were males, 61% had DM, 23% had CHF, and 10% had COPD. Patients' baseline characteristics and HD-related parameters are shown in Table 1.

| Patient characteristics relative to survival status
Forty-four (20.4%) patients died during follow-up. Baseline characteristics stratified by survival status are F I G U R E 1 Top panel: Kaplan-Meier analysis of survival probabilities in the four subgroups of patients based on the level of central venous oxygen saturation (ScvO 2 ) and relative blood volume (RBV) above and below the median of 58.8% and À5.5%, respectively. Bottom panel: Contour plot showing the bivariate joint association of average ScvO 2 levels and RBV change with all-cause mortality. The results of this analysis are shown as a contour plot that can be read like a topographic map. It presents three features ("dimensions") in a 2D plot, where the first dimension (in our case, RBV change) is represented on the x-axis, the second dimension (average ScvO 2 ) on the y-axis, and the third dimension [hazard ratios (HRs)] as "altitude." Levels of HRs are indicated in colors from yellow to red; the black lines indicate identical HRs (e.g., 0.8, 1.0, 1.5), akin to altitude contour lines in maps. The asterisk represents the median average ScvO 2 and RBV change. The four groups based on the population's median ScvO 2 and RBV change used for analyses are depicted in their respective quadrant within the contour plot. Group 1, ScvO 2 above median and RBV change below median; group 2, ScvO 2 and RBV change above median; group 3, ScvO 2 and RBV change below median; and group 4, ScvO 2 below median and RBV change above median. [Color figure can be viewed at wileyonlinelibrary.com] depicted in Table 1. Patients who died were older (68.3 vs. 60.7 years), had longer vintage (3.1 vs. 2.6 years) and more often DM (77.2% vs. 57.5%), higher interdialytic weight gain (2.1 vs. 1.9 kg), lower normalized UFR (7.5 vs. 7.0 mL/h/kg), lower serum albumin (3.8 vs. 3.6 g/dL), higher ferritin levels (907.4 vs. 740.9 ng/mL), higher F I G U R E 2 Top panel: Kaplan-Meier analysis of survival probabilities in the four subgroups of patients based on the level of central venous oxygen saturation (ScvO 2 ) and ultrafiltration rate (UFR) above and below the median of 58.8% and À6.7 mL/kg/h, respectively. Bottom panel: Contour plot showing the bivariate joint association of average ScvO 2 levels and UFR with all-cause mortality. The results of this analysis are shown as a contour plot that can be read like a topographic map. It presents three features ("dimensions") in a 2D plot, where the first dimension (in our case, UFR) is represented on the x-axis, the second dimension (average ScvO 2 ) on the y-axis, and the third dimension [hazard ratios (HRs)] as "altitude." Levels of HRs are indicated in assorted colors from yellow to red; the black lines indicate identical HRs (e.g. 0.8, 1.0, 1.5), akin to altitude contour lines in maps. The asterisk represents the median average ScvO 2 and UFR. The four groups based on the population's median ScvO 2 and UFR used for analyses are depicted in their respective quadrant within the contour plot. Group 1, ScvO 2 above median and UFR below median; group 2, ScvO 2 and UFR above median; group 3, ScvO 2 and UFR below median; and group 4, ScvO 2 below median and UFR above median. [Color figure can be viewed at wileyonlinelibrary.com] T A B L E 1 Baseline characteristics of all patients and stratified by survival status.

| Association of ScvO 2 and RBV with all-cause mortality
We stratified patients into four groups based on the median population ScvO 2 (58.8%) and RBV (À5.5%) (Figure 1, top panel). Kaplan-Meier analysis indicated a significantly better survival in patients with higher ScvO 2 combined with greater RBV decline (group 1) when compared to the other three groups. In the adjusted model, all-cause mortality was highest in patients with ScvO 2 below median and RBV change above median (group 4; HR 6.32; 95% confidence interval [CI] 1.37-29.06), followed by patients with ScvO 2 below median and RBV change below median (group 3; HR 5.04; 95% CI 1.14-22.35) and ScvO 2 above median and RBV change above median (group 2; HR 4.52; 95% CI 0.95-21.36) ( Table 2). The joint association of mean intradialytic ScvO 2 and RBV change with all-cause mortality is shown in a contour plot (Figure 1, bottom panel). Average ScvO 2 appears to have a strong association with mortality. Patients with higher ScvO 2 levels showed a lower allcause mortality even in the presence of only small intradialytic RBV changes. In contrast, lower ScvO 2 levels are associated with higher mortality, with the worst survival found in those patients with lower ScvO 2 and small RBV changes.

| Association of ScvO 2 and UFR with all-cause mortality
Secondary analyses with the same methodology were conducted to explore the joint association between SvcO 2 and UFR and survival. We stratified the population by median ScvO 2 (58.8%) and median UFR (6.5 mL/kg/h) ( Figure 2, top panel), while the joint association of mean intradialytic ScvO 2 level and UFR with all-cause mortality is shown in a contour plot (Figure 2, bottom panel). The analysis corroborated the notion that higher mean intradialytic ScvO 2 levels are associated with better survival.
Univariate Cox analysis showed the best survival in group 2, patients with ScvO 2 and UFR above the respective medians. In contrast, worse survival was observed in group 4, patients with lower ScvO 2 combined with higher UFR (above the median). In multivariate analysis with T A B L E 2 Crude and adjusted hazard ratios (HR) for all-cause mortality associated with ScvO 2 and RBV levels below or above the median population.

| DISCUSSION
Our study in maintenance HD patients elucidates the interactions between intradialytic ScvO 2 and RBV changes in relation to all-cause mortality. The results illustrate that ScvO 2 levels are a major driver of the association between RBV changes and patient outcome. A larger RBV decline in combination with higher average intradialytic ScvO 2 was associated with improved survival, whereas all-cause mortality was highest in patients with low ScvO 2 combined with small RBV changes. To our knowledge, this study is the first to explore the joint association of ScvO 2 and RBV changes with all-cause mortality in chronic HD patients. ScvO 2 monitoring is a cornerstone in the management of critically ill patients. Lower ScvO 2 levels are associated with increased morbidity and mortality. 11,22-24 ScvO 2 is determined by UBBF, Hgb concentration, arterial oxygen saturation (SaO 2 ), and oxygen extraction by the organs and tissues draining into the superior vena cava. UBBF is mainly determined by cardiac function and the conductive properties of the conduit and resistance vessels. 24,25 Regarding the impact of SaO 2 on ScvO 2 , it is important to note that-everything else being equal (e.g., tissue oxygen extraction)-SaO 2 and ScvO 2 are linearly related. However, to attain the low ScvO 2 levels seen in our study, the SaO 2 as an explanation would have been very low. In HD, ScvO 2 decline during HD treatment has been shown to have an inverse relationship with UFV. 12,26 Moreover, lower average ScvO 2 levels 14 and high ScvO 2 variability 13 have been associated with increased all-cause mortality. Although not conclusively defined, we conceptualize that low intradialytic ScvO 2 levels are an indicator of a low cardiovascular reserve. Imposing high UFR in these patients may lead to a critical decline in organ perfusion and subsequent organ damage. 8,9 In patients with higher intradialytic ScvO 2 , high UFR may be better tolerated, and fluid overload can be easier prevented.
Cordtz et al. 27 assessed whether ScvO 2 could be used as a surrogate for cardiac output (as measured by thoracic impedance). They report results in 20 HD patients who were classified as either hypotension-prone or hypotension-resistant and in who ScvO 2 was measured at the start and end of the HD session. The authors found a significant decrease in ScvO 2 in hypotension-prone patients, with no ScvO 2 decrease observed in the hypotension resistant group.
RBV monitoring is used by some nephrologists to guide fluid management and reduce the incidence of hypotensive events while improving cardiovascular stability, although mixed results have been shown. 28,29 The RBV curve reflects the difference between UFR and T A B L E 3 Crude and adjusted hazard ratios (HR) for all-cause mortality associated with ScvO 2 and UFR levels below or above the median population. vascular refilling rate from the interstitium. 30 Flat RBV curves may indicate fluid overload 31,32 and Agarwal found that an RBV decline of less than 1.39% per hour was associated with higher mortality. 18 We have corroborated that notion in a recent publication that has shown that a RBV decline of less than 8% 3 h into the dialysis treatment is associated with increased mortality. 19 The present study not only corroborates these earlier findings but also adds to their interpretation by showing that low ScvO 2 levels augments the mortality risk associated with small RBV changes. Next to fluid overload 18 a small decline in RBV might also point to an inability to tolerate fluid removal. The concurrent presence of low ScvO 2 and small RBV change (group 4) might thus indicate the coexistence of poor cardiovascular reserve and fluid overload. The combination of higher ScvO 2 and more pronounced RBV decline (group 1) may comprise patients who tolerate ultrafiltration well and in whom fluid overload can be well treated. Of note, also low intradialytic ScvO 2 and larger RBV decline (group 3) was associated with a significantly increased mortality rate. The subgroup with higher intradialytic ScvO 2 levels and smaller declines in RBV (group 2) also showed an increased mortality rate in unadjusted analysis, which might be explained by the relation between a small decline in RBV and fluid overload. 33 Based on the nature of our work, it was not our intention to predict mortality. The development of a prediction model will require a much larger data set. Our goal was to shed light on the pathophysiology of the association between RBV, ScvO 2 , and mortality. ScvO 2 was identified as a major potential contributor of outcomes when exploring ScvO 2 and UFR combined: survival was worse in patients with low ScvO 2 levels and UFR above the population median, although the results did not reach statistical significance in adjusted analysis. High UFR has been associated with IDH, cardiac stunning, and mortality. [34][35][36] Our results add to these observations by indicating that low intradialytic ScvO 2 may compound the adverse effects of high UFR.

Variable
Patients who died also showed several notable differences when compared to survivors, most prominently an inflammatory phenotype, as indicated by lower serum albumin levels, higher NLR, and higher ferritin levels. These findings corroborate previous studies. 37,38 Admittedly, our study has some limitations. First, its retrospective nature prevents determination of causality. Objective indicators of fluid status (e.g., bioimpedance) are missing; these would allow a more extensive probing into the relationship between RBV changes and fluid status. We also acknowledge that RBV measurement can be affected by a number of factors that may be technical (e.g., measuring path obstruction, gas bubbles in the measurement chamber, wrong placement of the measurement chamber) or treatment-related (e.g., infusion of fluid during dialysis). We aimed to address these points by excluding session with implausible measurements of ScvO 2 and HCT. While we have not measured of access recirculation, we believe that access recirculation would not impact ScvO 2 , because there is neither a gain nor a loss of O 2 in the extracorporeal circuit given the low physical solubility of oxygen in liquids, such as the dialysate. We also do not have indicators of cardiac function (e.g., echocardiogram). Hence, while physiologically plausible, the relationship between ScvO 2 and cardiac output remains conjectural. In addition, treatment-level SaO 2 was not available to us, so we cannot exclude that intradialytic changes in respiratory function may have affected SaO 2 and consequently ScvO 2 . Lastly because ScvO 2 measurements require CVC as vascular access, our results cannot be extrapolated to patients with arteriovenous vascular access. Study strengths are the diverse dialysis population, the substantial number of dialysis treatments, standardized care protocols, and automated and systematic recording of ScvO 2 and RBV.
In conclusion, intradialytic ScvO 2 is an important effect modifier that should be considered when interpreting the association between RBV change, UFR, and outcomes. Mortality risk is particularly increased in patients with concurrent low ScvO 2 and small RBV changes. The integration of multiple physiological parameters may yield important clinical insights that remain elusive when studying a single parameter. To further our pathophysiological understanding, specifically designed prospective studies with concurrent biochemical and physiological measurements are warranted.