**5. Clinical outcomes**

#### **5.1. Intermediary treatment outcomes (short‐term studies)**

In this section, intermediary treatment outcomes stand for surrogates of primary outcomes in assessing HDF safety, efficacy, and tolerance. For this purpose, we focus on cardiovascular stability and treatment tolerance, solute removal (phosphate, β2M), and inflammation, oxidative stress, and anemia.

#### *5.1.1. Cardiovascular stability and treatment tolerance*

of plasma proteins) [33], ideal modern equipment should be able to automatically address

Calcium: The target in the case of calcium is to maintain a neutral calcium balance, as an excessive calcium load has been associated with vascular calcification, whereas calcium depletion has been linked to worsening secondary hyperparathyroidism and decreased bone mass [34]. Here also, the modality of HDF affects the balance. Although calcium balance during postdilution online HDF does not differ from standard HD, it is usually recommended to increase the dialysate/substitution fluid calcium concentration by 0.25 mmol/L in predilution HDF mode in order [34]. In addition, discrepancies between expected and observed concen‐ trations in the dialysate/substitution fluid play an important role in the case of online HDF. In a volumetric system based on conductivity, the sodium for the bicarbonate dialysate/substi‐ tution fluid comes in part from a basic component and in part from an acidic component. In cases of a decrease in dialysate sodium with concomitant increase of bicarbonate, a lower proportion of the acid component will produce a lower than expected calcium level in the

Bicarbonate: a positive bicarbonate balance is targeted during the treatment session to neutralize the interdialytic accumulation of strong acid anions and to avoid starting the next session with metabolic acidosis. However, there is also a risk of postdialysis alkalosis. In a recent publication by Havlin et al. [36] analyzing 68 patients on postdilution HDF treated for 4 to 5 h with 80 to 90 mL/min of substitution fluid (19–27 L/session) with a dialysate bicarbonate concentration of 32 mmol/L, 34% of patients were acidotic at dialysis initiation, but 80% had metabolic alkalosis after dialysis. They speculated that this was due to an excessive elimination of retained and endogenous anions. According to the authors, this observation requires further investigation. In any case, several factors affect the mass balance. A significantly lower bicarbonate gain was observed in predilution HDF versus postdilution HDF [37]. As is true for all electrolytes, the difference in concentration between bicarbonate levels in the dialysate/ substitution fluid and in the blood at the initiation of the session is positively correlated to the mass transfer. Therefore, to maintain the same bicarbonate balance when moving from postdilution HDF to predilution HDF, dialysate bicarbonate concentration should be increased by 2 mmol/L. In fact, in predilution HDF, bicarbonate levels of the blood entering in the dialyzer increase, enhancing the loss across the membrane and reducing the normal gain by

In this section, intermediary treatment outcomes stand for surrogates of primary outcomes in assessing HDF safety, efficacy, and tolerance. For this purpose, we focus on cardiovascular stability and treatment tolerance, solute removal (phosphate, β2M), and inflammation,

sodium balance.

86 Advances in Hemodiafiltration

dialysate/substitution fluid [35].

diffusion from dialysis fluid to blood [33].

**5.1. Intermediary treatment outcomes (short‐term studies)**

**5. Clinical outcomes**

oxidative stress, and anemia.

In the short term, a significant reduction in the episodes of intradialytic hypotension was observed in HDF compared to conventional HD [16]. This has been ascribed to negative thermal balance (due to the infusion of relatively cool replacement fluid), a high sodium concentration of the substitution fluid, and/or removal of vasodilating mediators [4].

#### *5.1.2. Solute removal: phosphate, β2‐microglobulin*

Several controlled studies have confirmed enhanced clearance and mass removal of β2M with HDF (30–40% higher than high‐flux HD) accompanied by a 10 to 20% decline in circulating blood β2M concentrations [38, 39]. It must be reminded that reduction of circulating predialysis β2M takes time, as plasma levels reflect the equilibrium between production and elimination rates. Thus, 3 to 4 weeks are required to achieve a new steady‐state and a serum concentration change [40]. Recently, it has been calculated in a large cohort of incident ESKD patients that each additional 10 L convection volume was associated with a 0.8 mg/L reduction of β2M [24].

Phosphate mass removal and serum phosphate is a major concern in ESKD patients. RRT accounts for 60 to 70% of the total amount of phosphate removed to restore weekly phosphate mass balance. The other 30 to 40% needs to be eliminated by feces through the combined action of diet and phosphate binders. Although still a matter of controversy, high efficiency HDF has been shown to enhance the phosphate mass removed by 15 to 20% [41] with a subsequent predialysis serum phosphate level reduction of 6%. The percentage of patients reaching target pretreatment serum phosphorus levels with HDF was reported to increase from 64 to 74% in the Convective Transport Study (CONTRAST) [42].

Higher clearances of a number of other uremic compounds have also been documented with HDF: complement factor D (a proinflammatory mediator), leptin (16 kDa; involved in loss of appetite), FGF23 (30 kDa, implicated in metabolic bone disorders and vascular calcification), various cytokines, circulating advanced glycosylation end products (AGEs), and AGE precursors [43, 44].

#### *5.1.3. Inflammation, oxidative stress, and anemia*

Inflammation and oxidative stress profiles tend to be improved in patients treated by HDF. Several prospective studies have shown that levels of C‐reactive protein (CRP) and other sensitive biomarkers of inflammation (e.g., interleukin‐6) and/or proinflammatory cells are reduced. In this field, the Rischio cardiovascolare nei pazienti afferenti all'area vasta in dialisi (RISCAVID) study is certainly one of the more convincing studies, being conducted in a large cohort of dialysis patients [7, 45]. A meta‐analysis has recently reemphasized that the regular use of ultrapure dialysis fluid was the main driving force for such benefits [14].

The erythropoiesis‐stimulating agent (ESA) dose could be reduced in HDF, as reported in several clinical studies and summarized in a systematic review [46]. The benefit was attributed to the combined effects of the higher removal of middle‐sized toxins (erythropoietic inhibitor substances) and reduced inflammation due to the use of higher‐quality water and dialysis fluid [47, 48]. However, this effect was not confirmed in a recent meta‐analysis [49].

#### *5.1.4. Clinical benefits*

Several large cohort studies have indicated that the extended use of high‐flux membranes and convective therapies has a beneficial impact on the development of β2M amyloidosis in the long term, reducing the incidence of carpal tunnel syndrome and other related manifestations [50, 51]. This beneficial effect probably results from the regular use of ultrapure water and biocompatible materials, reducing inflammation, combined with convective modalities that enhance β2M removal [52].

#### **5.2. Endpoint outcomes (morbidity and mortality)**

In this section, endpoint outcomes are hard primary outcomes in assessing HDF long‐term efficacy. Consequently, the focus here is on mortality (all‐cause and cardiovascular) and morbidity (hospitalization, dialysis‐related pathology).

#### *5.2.1. Observational (cohort) studies on hemodiafiltration and clinical endpoints*

Locatelli et al. [50] conducted a retrospective observational study (Lombardy registry) of 6444 patients with ESKD who started RRT on HD, HDF, or hemofiltration (HF) between 1983 and 1995. A total of 1082 patients were treated with HDF or HF (first choice in the case of 188); the median follow‐up time was 29.7 months. Interestingly, after adjustment for age, gender, and comorbidities (including diabetes), the RR for carpal tunnel surgery and mortality was 42% (statistically significant) and 10% lower (not statistically significant) in patients treated with HDF or HF.

In 2006, Canaud et al. [53] reported results of a prospective, nonrandomized observational study from the Dialysis Outcomes and Practice Patterns Study (DOPPS) of 2165 patients followed between 1998 and 2001 in five European countries. Patients were stratified into four groups: low‐flux HD (n=1366), high‐flux HD (n=546), low‐efficiency HDF (n=156; substitution volume 5–14.9 L/treatment), and high‐efficiency HDF (n=97; substitution volumes 15–24.9 L/ treatment). Patient characteristics (including age and sex), 14 comorbidities, and time on dialysis were similar in each group. High‐efficiency HDF patients had lower crude mortality rates than low‐flux HD patients. After Cox regression analysis with adjustment, high‐efficiency HDF patients had a 35% significantly lower mortality risk than those receiving low‐flux HD.

Also in 2006, Jirka et al. published the results of an observational study of 2564 ESKD patients (394 on HDF) treated in Fresenius Medical Care clinics and followed for 12 months. Data were collected in the European Clinical Database (EuCliD) [54]. In this patient cohort, all‐cause mortality was reduced by 43% and unadjusted mortality was reduced by 35% in patients treated with HDF compared to HD. Convection volume was not reported.

In 2008, Panichi et al. [7] reported the results of a prospective observational study performed in the northwestern part of Tuscany that included 757 ESKD patients (RISCAVID study) who were followed for 30 months. Treatment with low‐ or high‐flux HD (n=424) was compared to treatment with HDF using substitution fluid delivered in bags (130 patients on low‐volume HDF with acetate‐free biofiltration (AFB) and 74 patients on HDF with convection volumes of 10–15 L/treatment) and treatment with online HDF (129 patients on HDF with convection volumes of 22–25 L/session). Cox proportional hazards regression analysis showed that online HDF and bag HDF patients had a significantly better survival than HD patients, having a 22% reduced risk of all‐cause mortality after adjustment.

*5.1.4. Clinical benefits*

88 Advances in Hemodiafiltration

enhance β2M removal [52].

HDF or HF.

**5.2. Endpoint outcomes (morbidity and mortality)**

morbidity (hospitalization, dialysis‐related pathology).

*5.2.1. Observational (cohort) studies on hemodiafiltration and clinical endpoints*

Several large cohort studies have indicated that the extended use of high‐flux membranes and convective therapies has a beneficial impact on the development of β2M amyloidosis in the long term, reducing the incidence of carpal tunnel syndrome and other related manifestations [50, 51]. This beneficial effect probably results from the regular use of ultrapure water and biocompatible materials, reducing inflammation, combined with convective modalities that

In this section, endpoint outcomes are hard primary outcomes in assessing HDF long‐term efficacy. Consequently, the focus here is on mortality (all‐cause and cardiovascular) and

Locatelli et al. [50] conducted a retrospective observational study (Lombardy registry) of 6444 patients with ESKD who started RRT on HD, HDF, or hemofiltration (HF) between 1983 and 1995. A total of 1082 patients were treated with HDF or HF (first choice in the case of 188); the median follow‐up time was 29.7 months. Interestingly, after adjustment for age, gender, and comorbidities (including diabetes), the RR for carpal tunnel surgery and mortality was 42% (statistically significant) and 10% lower (not statistically significant) in patients treated with

In 2006, Canaud et al. [53] reported results of a prospective, nonrandomized observational study from the Dialysis Outcomes and Practice Patterns Study (DOPPS) of 2165 patients followed between 1998 and 2001 in five European countries. Patients were stratified into four groups: low‐flux HD (n=1366), high‐flux HD (n=546), low‐efficiency HDF (n=156; substitution volume 5–14.9 L/treatment), and high‐efficiency HDF (n=97; substitution volumes 15–24.9 L/ treatment). Patient characteristics (including age and sex), 14 comorbidities, and time on dialysis were similar in each group. High‐efficiency HDF patients had lower crude mortality rates than low‐flux HD patients. After Cox regression analysis with adjustment, high‐efficiency HDF patients had a 35% significantly lower mortality risk than those receiving low‐flux HD.

Also in 2006, Jirka et al. published the results of an observational study of 2564 ESKD patients (394 on HDF) treated in Fresenius Medical Care clinics and followed for 12 months. Data were collected in the European Clinical Database (EuCliD) [54]. In this patient cohort, all‐cause mortality was reduced by 43% and unadjusted mortality was reduced by 35% in patients

In 2008, Panichi et al. [7] reported the results of a prospective observational study performed in the northwestern part of Tuscany that included 757 ESKD patients (RISCAVID study) who were followed for 30 months. Treatment with low‐ or high‐flux HD (n=424) was compared to treatment with HDF using substitution fluid delivered in bags (130 patients on low‐volume HDF with acetate‐free biofiltration (AFB) and 74 patients on HDF with convection volumes of

treated with HDF compared to HD. Convection volume was not reported.

One year later, Vilar et al. [55] reported the results of an observational study of 858 incident ESKD patients in the United Kingdom followed over 18 years. Patients treated with online HDF (n=232) received 79% of treatments with this modality and a mean filtration volume of 14.9 L/session. The control group was treated exclusively with high‐flux HD (n=626) and pure dialysis fluid. The mortality risk was significantly reduced in patients receiving predominantly HDF [hazard ratio (HR) 0.45; p<0.001] after adjustment for age, gender, body mass index, and comorbidities.

Imamovic et al. reported the Balkan experience in 2014. In this cohort study of 442 incident patients, the risk of death for HDF‐treated patients relative to high‐flux HD patients was 0.87 (nonsignificant) for low‐volume HDF and 0.29 (highly significant) for high‐volume HDF. After adjustment for covariates, the HR for patients on low‐volume HDF remained statistically not significant compared to high‐flux HD (HR 0.84; nonsignificant), whereas patients on high‐ volume HDF had a significantly lower HR (0.29; nonsignificant) than high‐flux HD. In the time‐dependent analysis, the mortality risk was not lower in high‐volume HDF compared to high‐flux HD (HR 0.48; nonsignificant), but this may be because 44% of the patients changed treatment modality during follow‐up [56].

In 2015, Siriopol et al. published a report on the Romanian experience with HDF. In this study, the group of 221 prevalent patients treated with online HDF (mean convection volume was 22.2 L) was propensity score matched to a group of 431 patients treated with HD [57]. Online HDF was associated with a reduced mortality risk (HR 0.62; statistically significant). A second cohort consisting of 265 incident patients on HDF were matched with 530 patients treated with HD. The mortality risk was significantly lower in patients treated with HDF (HR 0.22; highly statistically significant).

In another study, Canaud et al. reported results based on the EuCliD database involving 1590 incident patients in 12 European countries. The patient groups receiving high volume HDF (≥21 L/session) were propensity score matched to the group receiving HD. Patients were followed for 2 years (HD group) or 1.6 years (high‐volume HDF group) [58]. In this study, a nonsignificant survival advantage of HDF was found (HR 0.88; nonsignificant). Using inverse probability of censoring weighting (IPCW) to take bias due to the large amount of modality crossover during the follow‐up time into consideration (7% HDF patients switched to HD; 55% HD patients switched to high‐volume HDF), a statistically significant survival advantage of high‐volume HDF was found (odds ratio 0.501; highly significant).

Canaud et al. [24] recently published the results of a dose‐finding study exploring the optimal convection volume required to observe an increase in patient survival. This was a retrospective analysis involving 2293 incident HDF patients whose treatments were documented in the EuCliD database. Advanced statistical tools, including cubic spline analyses, were applied for the determination of the range of convection volume over which a survival benefit was observed. The relative survival rate of online HDF patients, adjusted for age, gender, comor‐ bidities, vascular access, albumin, CRP, and dialysis dose, was found to increase at approxi‐ mately 55L/week and to stay increased up to approximately 75L/week. Similar analysis of predialysis β2M concentrations found a nearly linear decrease in marker concentration, as convection volume increased from 40 to 75L/week. The analysis of log CRP levels showed a decrease over the same convection volume range.

Mercadal et al. reported a study using data from the French national Renal Epidemiology and Information Network (REIN) registry to assess the effects of HDF on mortality in the total population of incident dialysis patients (treatments were performed between 01/01/2008 and 31/12/2011 and patients were followed up to the end of 2012) [59]. Analyses were performed at both patient and facility levels. Here, 5526 out of 28,407 ESKD patients used HDF for a median of 1.2 years and 2254 of them used HDF exclusively. All‐cause and cardiovascular mortality associated with HDF use were significantly reduced (HR 0.84 and 0.73, respectively). In patients treated exclusively with HDF, the beneficial effects on all‐cause and cardiovascular mortality were more pronounced (HR 0.77 and 0.66, respectively). At the facility level, increasing the percentage of patients using HDF from 0 to 100% reduced the HR for all‐cause and cardiovascular mortality (HR 0.87 and 0.72, respectively). The authors concluded that, irrespective of whether analyzed at patient or facility level, HDF treatment was associated with better survival.

#### *5.2.2. Prospective RCTs*

Four prospective controlled or randomized studies comparing HDF and standard HD have been reported in the past which were not designed (short term <12 months) or sufficiently powered (<100 patients) to assess mortality differences between modalities [25, 38, 60, 61]. These will not be discussed in this chapter.

The Dutch CONTRAST was performed in 29 centers in The Netherlands (n=26), Canada (n=2), and Norway (n=1) [62]. Here, 714 patients were randomized between treatment with low‐flux HD and online postdilution HDF between 2004 and 2009, both with ultrapure dialysate. The primary endpoint was all‐cause mortality, and the main secondary endpoint was a composite of fatal and nonfatal cardiovascular events. Of 358 HDF patients, 121 discontinued treatment during the study due to transplantation, switch to another center or therapy, or other reasons. Of 356 HD patients, 118 patients discontinued the allocated treatment. The mean follow‐up was 36 months (range 5–79 months); during this period, 269 deaths occurred in 2170 person‐ years. All‐cause mortality (HR 0.95; nonsignificant) as well as cardiovascular events (fatal and nonfatal; HR 1.07; nonsignificant) were not affected by treatment modality. Although the target convection volume was set at 24 L/treatment, the mean volume achieved was 20.7 L/session and only one third of centers achieved at least 24 L/session. Post hoc analyses based on convection volume delivered (tertiles) showed a significantly lower mortality in patients receiving the highest convection volume (>21.95 L/session). In this subgroup, mortality risk was reduced by 39% compared to HD (HR 0.61; significant), which remained after extensive adjustments.

The Turkish Hemodiafiltration Study with 782 patients was conducted between 2007 and 2010 in 10 centers; patients were randomized to treatment with online postdilution HDF or high‐ flux HD [63]. Patients with central venous catheters, poor blood flow, and significant residual urine output were excluded. The primary endpoint was a composite of all‐cause mortality and first nonfatal cardiovascular event. Of 391 patients treated with HDF, 110 discontinued the study (28%), including 40 (10%) who terminated early due to vascular access problems. Of 391 patients randomized to HD, 90 patients (23%) dropped out. The mean follow‐up time was 23 months (range 1–38 months). The mean substitution volume was 17.2 L/session and the mean intradialytic weight gain was 2.4 L/treatment, corresponding to a mean total convection volume of 19.6 L/treatment. The RR of death and first cardiovascular event was 0.82 (non‐ significant), and the RR for all‐cause and cardiovascular mortality was 0.79 (nonsignificant) and 0.72 (nonsignificant), respectively. In a post hoc analysis, patients who achieved a convection volume above the median of 17.4 L/session had a significant lower risk of all‐cause and cardiovascular mortality (RR 0.54 and 0.29, respectively; both highly significant). This association remained after extensive adjustments for age, gender, comorbidity, and practice patterns.

The Catalonian Hemodiafiltration Study (ESHOL) included 906 Spanish dialysis patients in 27 units who were randomized to online postdilution HDF (n=456) and HD (n=450) [64]. In the HD group, 8% of the patients were treated with low‐flux membranes and 92% were treated with high‐flux membranes. The mean follow‐up time was 23 months. Here, 355 patients dropped out of the study for various reasons and were censored at the time of loss (36% in the HD group and 42% in the HDF group). Centers involved in the study received a short training course on how to achieve the targeted convection volume. The median convection volume in HDF treated patients was 22.9 to 23.9 L/treatment. Here, 207 events were observed in 1730 patient‐years. A significant 30% decrease in all‐cause mortality and 33% decrease in cardio‐ vascular mortality were observed in the HDF group. Interestingly, a post hoc analysis based on convection volume showed that the highest convection volume tertile (>25.4 L/treatment) was associated with a lower mortality risk (HR 0.55; highly significant) compared to HD patients.

#### *5.2.3. Meta‐analyses*

observed. The relative survival rate of online HDF patients, adjusted for age, gender, comor‐ bidities, vascular access, albumin, CRP, and dialysis dose, was found to increase at approxi‐ mately 55L/week and to stay increased up to approximately 75L/week. Similar analysis of predialysis β2M concentrations found a nearly linear decrease in marker concentration, as convection volume increased from 40 to 75L/week. The analysis of log CRP levels showed a

Mercadal et al. reported a study using data from the French national Renal Epidemiology and Information Network (REIN) registry to assess the effects of HDF on mortality in the total population of incident dialysis patients (treatments were performed between 01/01/2008 and 31/12/2011 and patients were followed up to the end of 2012) [59]. Analyses were performed at both patient and facility levels. Here, 5526 out of 28,407 ESKD patients used HDF for a median of 1.2 years and 2254 of them used HDF exclusively. All‐cause and cardiovascular mortality associated with HDF use were significantly reduced (HR 0.84 and 0.73, respectively). In patients treated exclusively with HDF, the beneficial effects on all‐cause and cardiovascular mortality were more pronounced (HR 0.77 and 0.66, respectively). At the facility level, increasing the percentage of patients using HDF from 0 to 100% reduced the HR for all‐cause and cardiovascular mortality (HR 0.87 and 0.72, respectively). The authors concluded that, irrespective of whether analyzed at patient or facility level, HDF treatment was associated with

Four prospective controlled or randomized studies comparing HDF and standard HD have been reported in the past which were not designed (short term <12 months) or sufficiently powered (<100 patients) to assess mortality differences between modalities [25, 38, 60, 61].

The Dutch CONTRAST was performed in 29 centers in The Netherlands (n=26), Canada (n=2), and Norway (n=1) [62]. Here, 714 patients were randomized between treatment with low‐flux HD and online postdilution HDF between 2004 and 2009, both with ultrapure dialysate. The primary endpoint was all‐cause mortality, and the main secondary endpoint was a composite of fatal and nonfatal cardiovascular events. Of 358 HDF patients, 121 discontinued treatment during the study due to transplantation, switch to another center or therapy, or other reasons. Of 356 HD patients, 118 patients discontinued the allocated treatment. The mean follow‐up was 36 months (range 5–79 months); during this period, 269 deaths occurred in 2170 person‐ years. All‐cause mortality (HR 0.95; nonsignificant) as well as cardiovascular events (fatal and nonfatal; HR 1.07; nonsignificant) were not affected by treatment modality. Although the target convection volume was set at 24 L/treatment, the mean volume achieved was 20.7 L/session and only one third of centers achieved at least 24 L/session. Post hoc analyses based on convection volume delivered (tertiles) showed a significantly lower mortality in patients receiving the highest convection volume (>21.95 L/session). In this subgroup, mortality risk was reduced by 39% compared to HD (HR 0.61; significant), which remained after extensive

decrease over the same convection volume range.

better survival.

90 Advances in Hemodiafiltration

adjustments.

*5.2.2. Prospective RCTs*

These will not be discussed in this chapter.

Several meta‐analyses comparing conventional HD and convective‐based therapies have been published over the last decade.

Rabindranath et al. from the Cochrane group performed two analyses in 2005 and 2006. The latter included 20 trials (657 patients). Mortality results were available only in 4 trials (336 patients) and different therapies were mixed (AFB, HF, and HDF). The authors found no difference in mortality risk for patients treated with convective‐based therapies [26, 65]. This systematic review was severely criticized due to its poor methodology [66, 67].

Susantitaphong et al. [49] compared convective‐based therapies to standard low‐flux HD. HF, HDF, AFB, and high‐flux HD were all included in the convective therapy group. This meta‐ analysis aggregated data of 12,182 patients. Convective therapies resulted in a nonsignificant decrease in all‐cause mortality (RR 0.88) and all‐cause hospitalization (RR 0.91); a significant decrease in therapy‐related hypotension (RR 0.55) and cardiovascular mortality (RR 0.84) was reported. The authors concluded that convective therapies were associated with improved clearance of uremic solutes, but the potential long‐term benefits of specific convective modal‐ ities could not be confirmed.

Wang et al. [68] conducted a systematic review and meta‐analysis that included 16 trials and 3220 patients treated with convective‐based therapies (HDF and HF) and with standard HD (low- and high-flux). Convection volume was not considered as a confounder in this analysis. On the one hand, the authors concluded that convective modalities did not significantly reduce the risk of cardiovascular events (RR 0.85) or all‐cause mortality (RR 0.83). On the other hand, they noted that convective modalities reduced intradialytic symptomatic hypotension (RR 0.49) and reduced serum β2M levels (‐5.95 mg/L).

Mostovaya et al. [69] compared exclusively HDF to HD (low- and high-flux) including 2402 patients. The meta‐analysis identified six RCTs. The convective arm consisted exclusively of HDF patients treated with different HDF modes (a mixture of postdilution, mid‐dilution, and predilution HDF, and of online HDF and HDF with bags) and achieving a specified minimum convection volume. All‐cause and cardiovascular mortality were reduced with HDF compared to HD (RR 0.8 and 0.73, respectively).

Nistor et al. [70, 71] from the Cochrane group updated the previous systematic review of 2005 and compared HD (low- and high-flux) to convective‐based modalities (HF, AFB, bag HDF, and online HDF) without considering convection volume. Thirty‐five trials (4039 patients) were included in this meta‐analysis and the effects on mortality were estimated. The convective group consisted of 1648 patients, but 227 of them were treated with low convection volumes. Within the limitations of the review (e.g., studies reviewed were partially old and referred to a diverse mixture of HDF modalities), the authors concluded that convective therapies had no significant effect on reducing all‐cause mortality (RR 0.87), cardiovascular mortality (RR 0.75), and intradialytic hypotension (RR 0.72) but had uncertain effects on nonfatal cardiovascular events (RR 1.14) and hospitalization (RR 1.21).

#### *5.2.4. Individual participant data meta‐analysis*

An alternative to the aggregated data meta‐analysis approach is to perform meta‐analysis of individual participant data in which the raw "individual level data" for each study are obtained and analyzed. The term "individual participant data" relates to the data recorded for each participant in a study. Individual participant data sets of four randomized trials were pooled and used to compare online HDF to HD. The four studies aggregated 2793 patients and were designed to examine the effects of HDF on mortality endpoints. Bias by informative censoring of patients was resolved. HRs comparing the effect of online HDF versus HD on all‐ cause and cause‐specific mortality were calculated using Cox proportional hazard regression models.

In the first part of this individual participant data meta‐analysis, Davenport et al. [72] analyzed the relationship between convection volume and patient outcomes. After a median follow‐up time of 2.5 years, 769 of the 2793 participants had died (292 cardiovascular deaths). Convection volumes were either not standardized or standardized to weight, body mass index, body surface area, and total body water. Data were analyzed by multivariable Cox proportional hazards modeling from 2793 patients. All‐cause mortality was reduced when the convective dose was unstandardized or standardized to body surface area or total body water; corre‐ sponding HRs were 0.65 (0.51–0.82), 0.74 (0.58–0.93), and 0.71 (0.56–0.93). Standardization by body weight or body mass index was not associated with significant survival advantages. Higher convection volumes were generally associated with greater survival benefit with online HDF, but results varied across the different ways of standardization for body size. Further studies should take body size into account when evaluating the impact of convection volume on mortality endpoints.

In the second part of this analysis, Peters et al. [73] also investigated the effects of convection volume on patient outcomes. HRs comparing the effect of online HDF versus HD on all‐cause and cause‐specific mortality were calculated using Cox proportional hazards regression models. The relationship between convection volume and the study outcomes was examined by delivered convection volume standardized to body surface area. Online HDF reduced the risk of all‐cause mortality by 14% and cardiovascular mortality by 23%. There was no evidence for a differential effect in predefined convection volume subgroups. The largest survival benefit was for patients receiving the highest delivered convection volume (>23 L/1.73 m2 body surface area per session), with a multivariable‐adjusted HR of 0.78 (95% confidence interval 0.62–0.98) for all‐cause mortality and 0.69 for cardiovascular disease mortality. This pooled individual participant analysis indicates that online HDF reduces the risk of mortality in ESKD patients. This effect holds across a variety of important convection volume subgroups of patients and is most pronounced for those receiving a higher convection volume normalized to body surface area.
