impact of fluid status and inflammation and their interaction on survival a study in an international hemodialysis patient cohort

clinical investigation www.kidney-international.org Impact of fluid status and inflammation and their interaction on survival: a study in an international hemodialysis patient cohort OPEN Marijke J.E Dekker1,2, Daniele Marcelli3, Bernard J Canaud3, Paola Carioni3, Yuedong Wang4, Aileen Grassmann3, Constantijn J.A.M Konings2, Peter Kotanko5,6, Karel M Leunissen1, Nathan W Levin5,6, Frank M van der Sande1, Xiaoling Ye5, Vaibhav Maheshwari5, Len A Usvyat5,7 and Jeroen P Kooman1; for the MONDO Initiative Department of Internal Medicine, Division of Nephrology, Maastricht University Medical Center, Maastricht, Netherlands; 2Department of Internal Medicine, Division of Nephrology, Catharina Hospital Eindhoven, Netherlands; 3Fresenius Medical Care, Bad Homburg, Germany; University of California, Santa Barbara, Santa Barbara, California, USA; 5Renal Research Institute, New York, New York, USA; 6Icahn School of Medicine at Mount Sinai Health System, New York, New York, USA; and 7Fresenius Medical Care North America, Waltham, Massachusetts, USA In hemodialysis patients extracellular fluid overload is a predictor of all-cause and cardiovascular mortality, and a relation with inflammation has been reported in previous studies The magnitude and nature of this interaction and the effects of moderate fluid overload and extracellular fluid depletion on survival are still unclear We present the results of an international cohort study in 8883 hemodialysis patients from the European MONDO initiative database where, during a three-month baseline period, fluid status was assessed using bioimpedance and inflammation by C-reactive protein All-cause mortality was recorded during 12 months of follow up In a second analysis a three-month baseline period was added to the first baseline period, and changes in fluid and inflammation status were related to all-cause mortality during six-month follow up Both pre-dialysis estimated fluid overload and fluid depletion were associated with an increased mortality, already apparent at moderate levels of estimated predialysis fluid overload (1.1–2.5L); hazard ratio 1.64 (95% confidence interval 1.35–1.98) In contrast, post-dialysis estimated fluid depletion was associated with a survival benefit (0.74 [0.62–0.90]) The concurrent presence of fluid overload and inflammation was associated with the highest risk of death Thus, while pre-dialysis fluid overload was associated with inflammation, even in the absence of inflammation, fluid overload remained a significant risk factor for short-term mortality, even following improvement of fluid status Kidney International (2017) j.kint.2016.12.008 -, -–-; http://dx.doi.org/10.1016/ KEYWORDS: body composition; fluid status; hemodialysis; inflammation Correspondence: Marijke Dekker, Department of Internal Medicine, Division of Nephrology, Maastricht University Medical Center, Post Box 5800, 6202 AZ Maastricht, Netherlands E-mail: mje.dekker@gmail.com Received 28 July 2016; revised 14 November 2016; accepted December 2016 Kidney International (2017) -, -–- Copyright ª 2016, International Society of Nephrology Published by Elsevier Inc This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) I n dialysis patients, the presence of extracellular fluid overload (FO) is associated with hypertension and left ventricular hypertrophy, while reversal of FO improves blood pressure (BP) and cardiac remodeling.1,2 Recent studies have shown that severe FO, defined as an expansion of the extracellular fluid by more than 15% (around 2.5 L in a person weighing 70 kg), is associated with lower survival in both hemodialysis (HD) and peritoneal dialysis patients.3–5 However, whether lower degrees of FO are also associated with mortality or the risk associated with FO increases incrementally with higher levels of FO has yet to be studied Moreover, circumstantial evidence suggests that fluid depletion (FD) is also associated with adverse outcomes.6 This is a highly relevant clinical question, because FD may contribute to intradialytic hypotension, impaired organ perfusion, and cardiac stunning.7 Also, high ultrafiltration volumes required to normalize extracellular fluid volumes are associated with increased mortality8 and can lead to postdialysis FD The association between depleted fluid status and mortality has not been investigated previously Importantly, the hazards of abnormal fluid status may be modified by inflammation Smaller studies have shown that inflammation, as reflected by increased C-reactive protein (CRP) levels, is associated with indicators of FO.9,10 However, it is not known to what extent the concurrent presence of FO and inflammation is associated with an increased mortality risk, or whether the mortality risk of FO persists after adjustment for inflammation Moreover, the nature and time line of this interaction has yet to be studied Whereas FO might lead to inflammation (e.g., because of translocation of endotoxins through an edematous bowel7), inflammation might conversely precede FO (e.g., due to a reduction in lean body mass which, if undetected, may result in inadequate dry weight prescription) clinical investigation MJE Dekker et al.: Fluid status and inflammation in dialysis patients This study in a multinational HD population had goals Firstly, to investigate the association between both moderate and severe FO and that of FD with all-cause mortality Secondly, to assess the joint and separate effects of inflammation and fluid status on survival Lastly, to identify the dynamic interaction of these parameters RESULTS We included 8883 patients with a median age of 63 years, 57.2% male, and a median dialysis vintage of 3.6 years (Table and Figure 1) Patient characteristics stratified by predialysis fluid status are presented in Table When the calculated postdialysis fluid status was analyzed, the same trend of an increasing risk of death by increasing levels of postdialysis FO was observed (Figure 2b) With regard to BP, the same trend of increasing BP with higher level of fluid overload was observed both before and after dialysis (Table 2) However, postdialysis FD was associated with significantly better survival compared with postdialysis normovolemia (HR: 0.74, 95% CI: 0.62–0.90) (Figure 2b) The majority of patients with postdialysis FD were normovolemic before dialysis, but a small percentage of patients also originated from the predialysis moderate and extreme FO groups (Figure 3) Concurrent presence of FO and inflammation Categorical analysis of mortality risk associated with different levels of fluid state In the adjusted analysis, levels of predialysis FO were associated with risk of death Compared to normovolemia, hazard ratios (HR) for mortality increased from 1.64 (95% confidence interval [CI]: 1.35–1.98) in the group with moderate FO to 4.23 (95% CI: 3.16–5.65) in patients with extreme FO (Figure 2a) In addition, predialysis FD was also associated with increased mortality (HR: 2.03, 95% CI: 1.32–3.12) The association between fluid status and mortality was confirmed in a second analysis in which predialysis fluid status was divided by normohydration weight on a patient base level The association with mortality was assessed via quantiles, and the same trend of incrementally increased risk of death with higher levels of fluid overload and with fluid depletion was observed (data not shown) Predialysis systolic BP increased incrementally with fluid status It was lowest in the predialysis FD group, and highest in patients with the most extreme FO (Table 2) CRP was lowest in the normovolemia group, and increased incrementally with both FO and FD Serum albumin levels decreased with increasing FO Inter-dialytic weight gain (IDWG) and ultrafiltration rate (UFR) were the lowest in predialysis normovolemic patients and increased with both predialysis FO and FD, with the longest treatment time in the group with extreme FO predialysis (Table 2) The number of baseline multifrequency bioimpedance spectroscopy (MF-BIS) measurements did not materially affect the results (data not shown) Table | Demographic characteristics at baseline for the total cohort (N [ 8883) and subgroup (N [ 5450) Variable Age (yr) Males Dialysis vintage (yr) Diabetes Mellitus Catheter access Northern Europea Eastern Europea Western Europea Southern Europea a Total cohort Dynamic cohort Mean (SD) Mean (SD) 63 (14.8) 57.2% 3.6 (1.6–6.9) 17.1% 18.6% 21 (0.2%) 4145 (46.7%) 333 (3.7%) 4384 (49.4%) 61.7 (15.6) 56.7% 3.7 (1.8–7.2) 17.7% 15.5% (0.1%) 3116 (57.2%) 133 (2.4%) 2198 (40.3%) Stratification for the regions within Europe was based on the United Nations geographic scheme Concurrent presence of inflammation and FO increased the risk of death in a dose-dependent fashion Predialysis FD concomitant with inflammation was also associated with a higher mortality risk In patients with predialysis severe FO in whom no inflammation was present, the HR was 3.09 (95% CI: 2.20–4.36); the HR increased to 6.02 (95% CI: 4.41–8.23) when inflammation was present (Table 3) Also in the normovolemia group, an increased risk was observed when inflammation was present (HR: 2.53, 95% CI: 1.81–3.53) Dynamic interactions between fluid and inflammatory status In this part of the study we included 5450 patients, 56.7% female, 17.7% with diabetes, and 84.5% with an arterio-venous fistula (Table 1) The highest percentage of patients (40.9%, n ¼ 2228) was categorized into the FO–/inflammation not present (I–) group and the lowest percentage (13.4%, n ẳ 732) into the FOỵ/I group (Figure 4) Patients in the FO–/I– group were younger, more often female, and had higher hemoglobin, albumin, creatinine, and normalized protein catabolic rate (nPCR) They were also less likely to have diabetes, and their predialysis BP was lower than patients categorized into any of the FOỵ groups A subset analysis in patients in the first year of dialysis showed materially identical results, with equal distribution over all groups (Supplementary Material B) Changes in fluid and inflammation status between the first and second baseline periods Figure depicts patients switching between the different groups In general, the majority of patients remained in the same group throughout the study period Of the 773 patients in the FOỵ/Iỵ group, 10.7% changed to the FO/I group in the second period Of the 1717 patients in the FO/Iỵ group in baseline period 1, 12.5% experienced FO in the subsequent period (switched to the FOỵ/Iỵ group and FOỵ/I groups), whereas of the 732 FOỵ/I patients, 26.8% experienced inammation during follow-up In a subset analysis with patients stratified by access type, starting in the FO–/I– group was significantly less frequent in patients with catheter access (32.1%) compared to those with an arteriovenous fistula (42.5%) (P < 0.001) They also more frequently remained in of the inflammation groups (Supplementary Material C) Change in fluid and inflammation status and its association with mortality Patients who continued to experience inflammation and FO were at highest risk of death (HR: 9.44, Kidney International (2017) -, -–- MJE Dekker et al.: Fluid status and inflammation in dialysis patients clinical investigation Figure | Study design For this study we included patients of the European subset of the MONDO Initiative database, with at least multifrequency bioimpedance spectroscopy (MF-BIS) measurement in the year 2011 The last MF-BIS measurement of this year was used as an anchor point All MF-BIS measurements (including the last of 2011), dialysis characteristics, and laboratory values of the months prior to this last MF-BIS measurement were used to calculate the averages of baseline period The same parameters of the months after the last MF-BIS measurement in 2011 were used to calculate the averages of period A comparison of the study cohort and excluded patients revealed that excluded patients were younger (64 vs 66 years; P < 0.001), more frequently diabetic (19.9% vs 17.7%; P < 0.001), and had a higher prevalence of catheter as dialysis access (23.6% vs 15.6%, P < 0.001) The difference in body mass index was not statistically significant C-reactive protein (CRP) levels were higher (median: 6.4 mg/l [2.60–15.80] vs 5.2 mg/l [2.30–11.57]; P < 0.001) Predialysis fluid status was lower in the excluded patients (median ỵ1.57 L [0.592.62] vs ỵ1.71 L [0.88–2.62]; P < 0.001) For more details, see the Supplemental Material A, a comparison between the study cohort and the excluded patient cohort *No data available due to transfer to other clinics, kidney transplantation, modality switch, no reason for unavailable CRP, or MF-BIS measurements **Predialysis fluid depletion was defined as the predialysis “overhydration” compartment reported below –1.1 L 95% CI: 5.67–15.72) (Table and Figure 5) However, even after improvement of both parameters (FO–/I–), their risk of death was still significantly higher (HR: 3.28, 95% CI: 1.13– 9.52) compared with those who never experienced FO or inflammation The same holds true for patients with either FO (FOỵ/I, HR: 3.10, 95% CI: 1.386.96) or inammation (FO/Iỵ, HR: 2.24, 95% CI: 1.174.29) and with improvement in the second period (Table and Figure 5) When predictors of change were analyzed, we found that having catheter access was associated with a higher chance of switching to the inflammation subgroups, and diabetes was associated with a higher chance of switching to the FO groups (Supplementary Figure S1) The mortality trends were materially the same in a subset analysis of patients stratified by access type (Supplementary Material C) Additionally, the association between fluid and inflammation status with death, when assessed by a linear mixed-model with fluid and inflammation status as timevarying covariates, showed the same trend, increased risk of death with increasing levels of FO (data not shown) DISCUSSION The main results of the present study are, firstly, the incremental rise in mortality risk with increasing levels of pre- and Kidney International (2017) -, -–- postdialysis FO Secondly, we observed an increased mortality risk in patients with predialysis FD and a reduced mortality risk associated with postdialysis FD Thirdly, we found a substantially increased mortality risk in patients with concurrent FO and inflammation, which does not completely resolve after improvement of either FO or inflammation Lastly, our data suggest that FO may be a risk factor for the development of inflammation Previous observational studies showed a relationship between mortality and FO, as determined either by MF-BIS (with FO defined as a ratio of overhydration [OH] to extracellular water greater than 15% [approximately 2.5 L])11 or vector bioimpedance.3,12,13 However, these studies did not report a more granular stratification of predialysis FO, nor was the association between FD and mortality investigated In the present study we observed an increased mortality risk even with moderate FO This finding may be of clinical importance, because a significant percentage of patients fall into this category (38.8% in this study) An additional intriguing finding is the relationship between predialysis FD and increased mortality, whereas postdialysis FD was associated with a reduced mortality risk This corroborates the clinical acumen that dry weight needs to find the middle ground between FD and FO.12 Severe predialysis FD may Categories of predialysis fluid status Normovolemia Moderate FO Severe FO Extreme FO £ –1.1 L > –1.1 L to D1.1 L > D1.1 L to D2.5 L > D2.5 L to D5.0 L > D5.0 L P valuec –12.2 (–20.28 to –9.14) 239 (2.7) 60.9 (15.7) 18.8 3.17 (0.15 to 5.29) 2761 (31.1) 63.4 (15.1) 12.8 10.79 (8.75 to 12.85) 3444 (38.8) 64.4 (14.8) 17.9 17.78 (15.38 to 20.59) 2083 (23.4) 63.0 (14.01) 24.7 30.18 (27.08 to 35.68) 356 (4.0) 59.6 (13.6) 33.7 6.0 mg/l, according to previous studies.27–29 Only patients with at least predialysis fluid status with an OH > –1.1 L were included Categorization of confounders and predictors A subgroup analysis of postdialysis fluid status was performed; to that end, we calculated that postdialysis fluid status (l) ¼ (postdialysis weight [kg] – predialysis weight [kg] ỵ predialysis uid status [l]) Ultraltration volume (l) was calculated by subtracting postdialysis weight from predialysis weight UFR (ml/h/kg), defined as ultrafiltration volume (ml) / dialysis treatment time (h) / postdialysis weight (kg), was stratified into different groups (UFR < 10 ml/h/kg, 10–13 ml/h/kg, > 13 ml/h/kg) based on previous studies and used as a confounder.8 An nPCR below 1.1 g/kg/d was used as an indicator of malnutrition Treatment time was stratified above and below hours Statistical analysis Continuous variables are reported as mean Ỉ SD or median and the 25th to 75th percentile The continuous variables were compared via 1-way ANOVA or Kruskal-Wallis tests based on their distribution, and categorical variables were compared via chi-square test The primary outcome was all-cause mortality Cox proportional hazard models with adjustment for age, gender, region, dialysis vintage, body mass index, nPCR, access type (arteriovenous fistula vs catheter access), UFR, and comorbidities (diabetes mellitus, CHF, peripheral vascular disease and present malignancy) were developed to assess the association of the different categories with mortality To assess the association of a change between the different subgroups and all-cause mortality we developed Cox proportional hazards models with different levels of adjustment The first model was unadjusted; the second was adjusted for age, gender, region, vintage, access type, diabetes mellitus, and CHF; and the third was adjusted for the factors used in model plus cerebrovascular disease, peripheral vascular disease, malignancies, body mass index, UFR, treatment time (above and below hours), and nPCR These same models were used when the association between predialysis fluid status divided by normohydration weight on a patient-based level, stratified in quantiles, was used (data not shown) Lastly, we additionally used a linear mixed model to investigate the association between fluid and inflammation status and mortality for the different baseline periods With this approach the same trend as with the conventional Cox model was observed (data not shown) We report point estimates and 95% CIs of HRs Censoring events were transfer to another clinic, transplantation, dialysis modality change, or reaching the end of the study period (December 31, 2012) A P value of