The percentage of patients who achieved a Hb level within the 10–13 g/dl target range increased from 60% to 80%, which was in accordance with the results obtained from the national data system « REIN 2008 ». This improvement may be due to the increase of erythropoietin doses and to the optimization of the iron status. In terms of mortality, meta-analyses did not show any statistically significant difference between higher and lower Hb level in hemodialysis patients 18 . This target allows for relative flexibility in medical decision making and takes into account variability between patients’ comorbidities, prognosis, functional status, and responsiveness to ESA therapy. In in-center hemodialysis patients, it is more difficult to achieve hemoglobin targets because of comorbidities, including numerous diseases with inflammatory state, responsible for ESA resistance. Accordingly, we found no correlation between erythropoetin doses and the percentage of conformity for anemia.

Our percentage of patients hemodialysed with catheters did not decrease. Indeed, the majority of incident patients did not start dialysis with AVF. Our conformity rate for vascular access was higher than 80% in 2009 when we excluded the incident patients who had initiated hemodialysis for less than three months, whereas it was lower than 80% when these patients were included in the analysis. Another reason is that there is a lack of suitable vessels to create AVFs because of an aging and diabetic population. Data from « REIN 2008 », that include both in-patients and out-patients, showed that 16.5% of 22852 patients were hemodialysed via catheters in France in 2008. In a cohort study of 78420 maintenance hemodialysis patients comprising approximately 26% of the US hemodialysis population, Lacson Jr et al. found a 39% increased risk of death with catheters compared with fistulas, making the vascular access the second most important actionable variable associated with mortality after albumin 17 . Thus, the first measure to improve our percentage of patients dialysed on AVF would be to avoid the late referral of the patient to the nephrologist, which has been associated with increased catheter use. The second would be to shorten the delay before the creation of the AVF after initiation of dialysis.

The prevalence of conformity for dialysis dose increased regularly, up to 98.4% in April 2010. The few patients who did not reach the Kt/V_BCM target had either a high urea distribution volume, or a dysfunction of their vascular access (catheter dysfunction or immature fistula). Results from « REIN 2008 » showed that 78.4% of 13451 patients had a Kt/V > 1.2. We used high blood and dialysate flow rates, membranes with a large surface area, and performed high-flux dialysis according to the results of the HEMO Study 19 . The prevalence of patients using hemodiafiltration increased during follow-up, which is a potential explanation for these results. Indeed, small solute dialysis dose delivered by hemodiafiltration is higher than that delivered by hemodialysis, because of increasing convective clearance. Canaud et al. showed that high-efficiency hemodiafiltration (infusion volume > 15 liters) had a positive impact on survival compared to patients treated by high-flux hemodialysis 20 . Articles suggest that dosing of dialysis should be based on the volume of blood cleared (Kt), rather than on Kt/V, which can lead to under-dialysis in women and small men by underestimating the hemodialysis dose 14 15 21 . Kt can easily be monitored during each treatment with the OCM device.

Hypertension is a major cardiovacular risk factor in ESRD patients. In our study, the prevalence of conformity for systolic blood pressure was not improved. However, hypertension guidelines in hemodialysis patients are not currently based upon evidence. KDOQI guidelines concerning blood pressure target ranges were extrapolated from the general population 5 : Predialysis and postdialysis BP goals should be <140/90 mm Hg and <130/80 mmHg respectively, provided that there is no substantial orthostatic hypotension and that these levels are not associated with substantial and symptomatic intradialytic hypotension. Tentori et al., in a retrospective analysis in 13792 incident hemodialysis patients, showed that following the guideline for predialysis blood pressure (BP) measurements was associated with increased mortality 5 . Zager et al. found a « U » curve relationship between post-dialysis SBP and cardiovascular mortality in more than 5400 hemodialysis patients. SBP < 110 mmHg and > 180 mmHg, and DBP > 90 mmHg were associated with poor outcomes 16 . Indeed, relative hypotension is a potent marker of mortality in ESRD patients, probably reflective of cardiac failure. Thus, it is unclear which BP target should be used as a quality-of-care indicator for hemodialysis patients. In the future, targets related to ambulatory BP measurements (ABPMs) or home BP measurements may be used. ABPMs were found to be superior to dialysis unit recordings in predicting outcomes 22 . In a prospective cohort study conducted in 150 chronic hemodialysis patients, self-measured and ambulatory systolic BP between 125 and 145 mmHg, and between 115 and 125 mmHg, respectively, were associated with a decreased risk of death 23 .

No improvement was observed in the mean serum albumin level and the percentage of conformity, which is around 50%. Our results were different from those reported in « REIN 2008 », where 64.9% of 24436 patients had an albumin > 35 g/l, presumably because of increased comorbidities in our in-center patients.

Albumin level showed the strongest association with mortality compared with other predictor variables in several large observational studies 24 . However, when adjusted to other major comorbidities, hypoalbuminemia was not significantly associated with mortality 25 . Since comorbid medical conditions may decrease albumin synthesis in the liver, hypoalbuminemia is a non-specific marker of denutrition and a difficult-to-modify patient factor, better associated with patient comorbidities than with poor quality of care 26 . In a randomized controlled study involving 180 patients with albuminemia below 37 g/l, a nutrition intervention tailored to patient-specific barriers resulted in modest improvements in albumin levels, regardless of levels on inflammatory markers 27 . In our study, albumin was measured by immunonephelometry, which is currently considered as the gold-standard method. The threshold used by the KDOQI (40 g/l) to detect hypoalbuminemia was not chosen because the bromocresol-green method used for KDOQI guidelines overestimates albuminemia 28 .

The conformity rate for serum bicarbonate increased from 50.8 to 81.1% during the study period, due to the individualized prescription of bicarbonate dialysate concentration. The EBPG guidelines recommend that the mid-week predialysis serum bicarbonate levels should be maintained at 20–22 mmol/l 13 . A target range between 20 and 24 mmol/l was used in this study. A U-shape relationship between serum bicarbonate and mortality or hospitalization has been demonstrated in hemodialysis patients in the DOPPS study 29 . After adjustment for numerous comorbidities and for nutritional markers, the lowest risk for mortality was observed with serum bicarbonate between 20.1 and 21 mmol/l and the lowest risk for hospitalization was observed for serum bicarbonate levels between 21.1 and 22 mmol/l. In this study, there was an inverse correlation between pre-dialysis serum bicarbonate and albuminemia, suggesting that the lower bicarbonate concentrations resulted from greater acid load caused by protein intake. Moderate predialysis acidosis may be associated with better nutritional status, which is strongly associated with increased survival in hemodialysis patients. Patients with persistent and severe metabolic acidosis may be treated by increasing the bicarbonate dialysate concentration or with oral sodium bicarbonate. Patients with serum bicarbonate levels higher than the upper target range can be treated by lowering the bicarbonate dialysate concentration, bearing in mind that intradialytic alkalosis is poorly tolerated with more hypotensive episodes during the dialysis sessions 30 .

Using dyslipidemia to define a quality-of-care indicator is unlikely to be clinically relevant, for several reasons. First, there is an inverse association of total cholesterol with mortality in dialysis patients 31 , probably due to the cholesterol-lowering effect of systemic inflammation and malnutrition. Secondly, the only two large trials that studied the effects of statins in haemodialysis patients did not show any beneficial effect on cardiovascular death 32 33 . In a post-hoc analysis of the 4D trial (Die Deutsche Diabetes Dialyse Study), high levels of LDL-cholesterol showed a tendency to increase the risk of cardiac endpoints and all-cause mortality. In patients with a baseline LDL-cholesterol greater than 1.45 g/l atorvastatin significantly decreased adverse fatal and non-fatal cardiac events and all-cause mortality, compared to placebo. As low serum cholesterol levels are associated with increased mortality, this indicator was not chosen as a « quality-of-care » indicator. Maybe cholesterol should be used to identify patients who need statins, ie patients with high LDL-cholesterol (> 1.45 g/l). Of course, statins should be continued in patients who already had this treatment before initiation of dialysis for a cardiac event 32 .

The conformity rate for serum calcium targets was stable around 60% during the study period despite the use of calcimimetics or low dialysate calcium concentration; it was above that observed in the national data system for bone metabolism (52%, Observatoire Photo-graphe ; n = 11172 patients) 34 . Recent KDIGO guidelines introduced new targets 12 . Despite a lack of evidence from randomized controlled trials demonstrating that attaining serum calcium targets impacts clinical outcome, large observational studies showed that the inflection point at which calcium becomes associated with an increased relative risk of all-cause mortality varies among studies, from 2.38 to 2.85 mmol/l 35 36 37 . In an observational study in incident hemodialysis patients, hypocalcemia < 2.20 mmol/l was independently associated with mortality (RR 2.10) 38 . Chronic hypocalcemia was significantly associated with both de novo and recurrent ischemic heart disease, and de novo and recurrent cardiac failure. Thus, the KDIGO lower and upper thresholds for calcium could be adapted, but one may argue that with this new target range, the mean serum calcium level will increase, and that high levels of Ca-P product may favor vascular calcification 39 .

In France, the conformity rate for serum phosphorus and the mean serum phosphate level were 52% and 1.56 mmol/l in June 2009, respectively (Observatoire Photo-graphe data). The mean serum phosphate level was lower in our study, probably because of the increased prevalence of malnutrition. The percentage of conformity for serum phosphorus improved during the follow-up, up to 72.3% in October 2009. The wider use of hemodiafiltration may partially explain this result, since we observed a trend towards lower serum phosphorus in hemodiafiltration patients. Two large prospective observational studies showed that this dialysis modality could improve phosphate control 40 41 . In the study of Lars Penne et al., the proportion of hemodiafiltration (HDF) patients with pre-dialysis phosphate concentrations < 1.78 mmol/l increased from 64% to 74% during the 6-month study period and was stable in hemodialysis patients. Nevertheless, two RCTs did not show any benefit of HDF on phosphate control, maybe because of the low baseline phosphate levels (1.58 mmol/l et 1.62 mmol/l) 42 43 . There is no evidence from RCTs that lowering serum phosphorus to a specific target range reduces mortality in hemodialysis patients. The target range for phosphorus was modified with the publication of the KDIGO clinical practice guidelines. The current target is between 0.9 and 1.5 mmol/l 12 . Our conformity rate will decrease with this new target range. Thus, more aggressive strategies will have to be adopted to lower serum phosphorus, such as adding a convective component to clearance with hemodiafiltration, lengthening dialysis session time or increasing dialysis frequency.

Because of the limited sunlight exposure in Normandy, our patients are systematically supplemented with 25OHvitD. Nevertheless, our conformity rate for vitamin D was not satisfactory, with important variations considering the mean vitamin D levels. Increasing evidence sugsests that Vitamin D deficiency is an independant risk factor for cardiovascular events and all-cause mortality in hemodialysis patients 44 45 . In the 2009 KDIGO guidelines, repeated measurements of 25OHvitD and therapeutic supplementation in case of deficiency are recommended 12 . A randomized controlled trial should be performed to clarify whether vitamin D supplementation can decrease adverse outcomes. Thus, vitamin D may become an important quality-of-care indicator in the near future.

In France, the conformity rate for PTH and the mean serum PTH level were 33% and 317 pg/ml in June 2009 respectively (Observatoire Photo-graphe data), whereas prevalence of conformity for PTH increased from 22% to 42.8% in our study. According to recent KDIGO guidelines PTH levels should be maintained in the range of approximately two to nine times the upper normal limit for the assay (45 to 490 pg/ml) 12 . Observational data demonstrated that the K/DOQI treatment goals were not easily achieved or maintained with traditional therapeutic options (phosphate binders, vitamin D analogs) for secondary hyperparathyroidism (SHPT). Moe et al. performed a secondary analysis of three large RCTs, demonstrating that cinacalcet effectively reduces PTH, calcium, and phosphorus to the K/DOQI target ranges in hemodialysis patients with SHPT 46 . In 2010, Block et al. showed in a prospectively designed observational study a significant survival benefit associated with prescribing cinacalcet for hemodialysis patients with evidence of SPHT and receiving i.v. vitamin D 47 . In contrast, the recently published EVOLVE study found no benefit to cinacalcet in hemodialysis patients and raised significant safety issues 48 .

An important improvement concerning the conformity rate for β2M was noticed, which increased from 28.2% to 48.5%. β2-microglobulin (β2M) is a marker for middle molecules in uremia and a potential target for adequacy in hemodialysis therapy. In 1704 patients from the HEMO study, pre-dialysis serum β2M levels > 27.5 mg/l were associated with all-cause-mortality 49 . Since we increased the use of hemodiafiltration with ultrapure dialysate, which is the most efficient therapy to reduce serum β2M levels, we expected a better conformity rate for β2M. Indeed, convective treatments are an established therapy to enhance uremic toxin removal over a wide molecular-weight spectrum. Maduell et al. found mean β2-microglobulin reduction rates were 75.4% for on-line post-dilution hemodiafiltration versus 60.1% for high-flux hemodialysis 50 . They also showed that short daily on-line hemodiafiltration was associated with lower pre-dialysis serum β2-microglobulin levels 51 . In addition, dialyzers with better β2M clearance may be a therapeutic option, although there may be an associated risk of greater albumin loss.

The percentage of dialysis sessions with intradialytic hypotension (IDH) remained low during the study period, around 12%. Mortality in patients with frequent IDH is significantly higher than in those without such events, but after adjustment for covariates, this association loses significance 52 . Thus, IDH may represent a marker of comorbid conditions. Our in-center dialysis study failed to demonstrate a substantial improvement for this indicator, despite many measures systematically followed to improve hemodynamic instability, perhaps because it was already very low. The dry weight was assessed by clinical examination with the help of bioimpedance measurements, using the BCM. We checked in those patients with frequent IDH that sodium restriction and timing of antihypertensive agents were well respected. More than 90% of the patients during the study period were treated with isothermic dialysis, that was shown to improve hemodynamic instability, as well as dialysis at cooler dialysate temperatures 9 53 . The percentage of patients on on-line post-dilution hemodiafiltration increased during the study. This convective technique was associated with a significant reduction of hypotensive episodes, predominantly related to decreased body temperature 54 . In the EBPG guidelines on hemodynamic instability, convective techniques are a possible alternative to cool dialysis. If these treatment options have failed, other available therapies are suggested: midodrine (level 1 evidence), blood volume controlled ultrafiltration, use of a dialysate calcium of 1.5 mmol/l, prolongation of dialysis time or increase in dialysis frequency (level 2 evidences), L-carnitine supplementation, and/or bicarbonate dialysis (level 3 evidences) 55 .

The importance of quality-of-care indicators is well established. Some indicators, as defined by international guidelines, rely on a solid scientific basis with clear associations with outcomes (hemoglobin, albumin, dialysis dose, vascular access). Nevertheless, the clinical advantage of meeting multiple treatment targets simultaneously remains to be established. In addition cost/benefit evaluations of each indicator, and especially of strategies based on multiple quality-of-care indicators should be performed. In hemodialysis patients, the association between several indicators, such as phosphate, BP or BMI, and mortality is characterized by a J-shaped curve. In this context, which quality-of-care indicator should be prioritized may be a matter of debate. In addition, regarding several routinely measured parameters, such as 25OHvitaminD or β2-microglobulin, evidence for specific targets and guidelines are lacking, suggesting that locally determined targets may be useful before RCTs and evidence-based guidelines are available.