Mixed-effects Gaussian process functional regression models with application to dose-response curve prediction.

We propose a new semiparametric model for functional regression analysis, combining a parametric mixed-effects model with a nonparametric Gaussian process regression model, namely a mixed-effects Gaussian process functional regression model. The parametric component can provide explanatory information between the response and the covariates, whereas the nonparametric component can add nonlinearity. We can model the mean and covariance structures simultaneously, combining the information borrowed from other subjects with the information collected from each individual subject. We apply the model to dose-response curves that describe changes in the responses of subjects for differing levels of the dose of a drug or agent and have a wide application in many areas. We illustrate the method for the management of renal anaemia. An individual dose-response curve is improved when more information is included by this mechanism from the subject/patient over time, enabling a patient-specific treatment regime.

Optimizing erythropoietin therapy in hemodialysis patients.

The European Best Practice Guidelines for the management of anemia in patients with chronic renal failure recommend the percentage of hypochromic red blood cells (%HRCs) as the best measure of iron use by erythropoietic tissues. They suggest that "sufficient iron should be administered to attain: serum ferritin 100 ng/mL, HRCs <10%. In practice, to achieve these minimum criteria will mean aiming for optimal levels of serum ferritin 200-500 ng/mL, HRCs <2.5%." We increased prospectively the delivered dose of iron supplements to a large (n = 228) unselected hemodialysis cohort with a sustained (24-month) hemoglobin (Hb) outcome meeting the UK Renal Association minimum standard of 85%, greater than or equal to 10.0 g/dL. This was managed through a computer-aided decision support system for erythropoietin (EPO) and intravenous iron sucrose therapy. Hb outcome was maintained with medians between 11.3 and 11.8 g/dL. Median red blood cell hypochromia (%HRCs) decreased from 8% (interquartile range [IQR], 3 to 15) to 4% (IQR, 2 to 8; P < 0.001, U-Mann Whitney test). Serum ferritin level increased from a median of 188 (IQR, 115 to 256) to 480 ng/mL (IQR, 397 to 595; P < 0.001, U-Mann Whitney test). Median EPO dose decreased from 136 (IQR, 83 to 216) to 72 IU/kg/wk (IQR, 33 to 134), which strongly correlated with median %HRCs through the range less than 10% (Spearman's correlation, 0.73; P < 0.01). These data suggest that EPO responsiveness continues to improve toward the normal range for %HRCs (<2.5%) and aspiring to values much less than 10% is cost-effective. The ferritin outcome required to achieve these lower values for %HRC outcome is greater than the current recommended range, although in steady state, the mean iron treatment dose is similar to that in previous studies (ie, approximately 60 mg/wk).

Intravenous iron for CAPD populations: proactive or reactive strategies?

BACKGROUND: The European best practice guideline [Nephrol Dial Transplant 1999; 14 (Suppl 5)] (5A) for the management of anaemia suggests that > 85% of the CAPD population should have a haemoglobin level of > 11.0 g/dl. METHODS: We developed and implemented an outpatient-based protocol for intravenous iron sucrose (IV Fe) and erythropoietin (Epo) in CAPD patients showing iron deficiency despite oral iron therapy. We managed a total of 103 patients over 13 months of study. All CAPD patients were included, regardless of co-morbidity. Treatment developed in two phases: in phase 1 (reactive) (months 1-8), patients with markers of iron deficiency (ferritin < 100 ng/ml or ferritin 100-500 and percentage hypochromic red cells (%HRC) > or =5) were converted from oral iron to IV Fe (300 mg) and reviewed after 4-8 weeks according to haemoglobin (Hb). In phase 2 (proactive) (months 9-13), the criteria for iron therapy were extended: ferritin < 150 ng/ml or ferritin 150-500 and %HRC > or = 2. Patients then received IV Fe (200 mg) and were reviewed after 4 weeks according to Hb. RESULTS: The median haemoglobin increased from 11.0 (Inter quartile range, IQR, 10.1-12.6) g/dl to 11.7 (11.0-12.7) g/dl (P = 0.06). The proportion of patients with absolute iron deficiency (ferritin < 100 ng/ml) decreased from 24 to 2%. The percentage of hypochromic red cells (%HRC) decreased from 4 (2-7) to 1 (1-4) (P < 0.01). CONCLUSIONS: An integrated Epo and IV Fe policy increased the number of patients reaching the European guideline from 50 to 75% with no increase in the population median Epo requirements (42 (IQR, 25-95) IU/kg/week vs 45 (27-101) (P = NS)). This study demonstrates the benefit of early (proactive) intervention in achieving population compliance within current guidelines for renal anaemia.

The bradykinin response and early hypotension at the introduction of continuous renal replacement therapy in the intensive care unit.

We assessed the relationship of certain clinical variables (including bradykinin [BK] release and dialysis membrane) to initial mean arterial pressure (MAP) reduction in 47 patients requiring continuous renal replacement therapy (CRRT) in an intensive care unit. The pretreatment MAP was 84 +/- 14 mm Hg for the group as a whole. The initial MAP reduction was 11.5 (7-20) mm Hg, occurring 4 to 8 min after connection. MAP reduction was 9 (6-15) mm Hg with polyacryonitrile (PAN) membranes versus 14 (5-19) mm Hg with polysulfone (PS) (not significant). There were positive correlations between MAP reduction and BK concentration at 3 (BK3; r = 0.58, p < 0.01) and 6 (BK6; r = 0.67, p < 0.001) min with PAN but not with PS. A greater reduction in MAP was seen in patients who were not receiving inotropic support (Mann-Whitney test, p < 0.01). BK3 and BK6 values for the PAN and PS groups were not significantly different. However, BK concentrations greater than 1,000 pg/ml were only seen with PAN (6 patients, MAP reduction 27 [17-31] mm Hg). There were positive (albumin) and negative (age; acute physiology, age, and chronic health evaluation score; C-reactive protein [CRP]; calcium) correlations with BK3/BK6 in the PAN and PS groups, some of which (albumin, CRP) reached statistical significance. In summary, MAP reduction at the start of CRRT correlates with BK concentration. The similarity of response with PAN and PS suggests an importance for other clinical factors. In this study, hemodynamic instability was more likely in patients with evidence of a less severe inflammatory or septic illness.

Intervention thresholds and ceilings can determine the haemoglobin outcome distribution in a haemodialysis population.

We have explored the consequences of setting different thresholds and ceilings for erythropoietin dose changes in two randomized controlled studies of renal anaemia management based on an established algorithm. STUDY 1: A large (n=236) unselected haemodialysis cohort was randomized to monthly intervention (increased erythropoietin (Epo)) at haemoglobin (Hb) levels falling below either 10.5 g/dl (group A) or 11.5 g/dl (group B) and followed for 6 months. The mean Hb was 0.6 g/dl higher in the 11.5 g/dl threshold group (11.1 g/dl vs 11.7 g/dl) at 6 months (P=0.001++). The Epo dose did not differ between them (median 133 IU/kg/week, Interquartile range (IQR) 86-217 and 140, IQR 74-227 respectively) (P=NS(**)). STUDY 2: A large (n=211) unselected haemodialysis cohort was randomized to a reduction in Epo dose at Hb levels above either 12.0 g/dl (group C) or 13.0 g/dl (group D). The Hb outcome at 8 months differed between group C (mean 11.5 g/dl, SD 1.4) and group D (12.2, SD 2.1) (P=0.03++). The Epo dose did not significantly differ between groups C and D (median 60 IU/kg/week, IQR 32-142 and 71, IQR 38-117 respectively) (P=NS(**)). Study 1 showed that an intervention threshold of 11.0 g/dl with a mean Hb outcome of 11.6 g/dl and SD 1.6 g/dl would produce the desired UK Renal Association Standards outcome of 85% Hb > or =10.0 g/dl. Study 2 demonstrated that a ceiling of 12.0 g/dl narrowed the range of Hb values (P:<0.001##), achieving a SD of 1.37 g/dl, and reduced the number of patients with a Hb >13.0 g/dl from 25 to 12%. This narrowing of the distribution has cost implications for reaching minimum standards in a haemodialysis population. Formal use of threshold and ceiling values for intervention within an anaemia management system enabled the haemodialysis population outcome mean and SD to be literally prescribed.

Modeling the postdialysis rebound: the reconciliation of current formulas.

Three approaches are currently used in kinetic models (UKMs) to account for the postdialysis rebound in urea concentration, and thereby accurately measure the hemodialysis dose, KT/V (where K, T, V denote dialyzer clearance, dialysis duration, and urea distribution volume, respectively). The approach developed by Smye uses an intradialytic sample to predict the postdialysis equilibrium concentration, Ce, which is then used in a single pool UKM to give KT/V. A second approach developed by Tattersall introduces a patient clearance time, tp. The true dialysis dose is then given by T/(T + tp) x apparent dose, and tp is estimated to be 36 minutes. The Daugirdas analysis uses an empiric regression equation to give the true dose; KT/V)true from the single pool value, KT/V)sp; KT/V)true = KT/V)sp - (36/T)(KT/V)sp + 0.03. The analysis confirms the equivalence of all three formulas, which arises from the observation that during the later stages of dialysis, the urea concentration decreases as a single exponential. The formulas are independent of whether a flow or diffusion model is used to describe the kinetics of urea removal. The original analysis assumed constant volumes, but the effect of ultrafiltration volume u on C(e) may be accounted for by multiplying by (1 + u/V). The Smye equation is more vulnerable to error in practice, because small errors in the intradialytic sample give larger errors in the equilibrium concentration estimate, whereas dose estimates based on the Tattersall and Daugirdas equations are less affected by sampling errors. However, unlike the Smye approach, these two formulas would need adaptation for use with other solutes. The advent of continuous urea monitoring should permit more accurate, prospective estimates of equilibrium concentrations and dialysis dose.

Assessment of ultrasound guided percutaneous ethanol injection and parathyroidectomy in patients with tertiary hyperparathyroidism.

BACKGROUND: Tertiary hyperparathyroidism continues to cause significant morbidity in patients with chronic renal failure. This is frequently resistant to medical management and may ultimately require a surgical parathyroidectomy. Recent studies have reported upon the technique of percutaneous ethanol ablation for both primary and tertiary hyperparathyroidism. In this study we report on a 5 year experience using ethanol injection and compare the results with surgical parathyroidectomy. METHODS: A prospective study in 39 patients with tertiary hyperparathyroidism, 25 were dialysis dependent and 14 had a functioning renal allograft. Twenty-two patients underwent percutaneous fine needle ethanol injection (PFNEI) and 17 underwent surgical parathyroidectomy. RESULTS: A > 30% reduction in intact parathyroid hormone (iPTH) was achieved in 11 of 22 patients undergoing PFNEI after a mean of 1.8 +/- 1.4 injections per gland. In four patients, symptomatic hyperparathyroidism recurred and they required further PFNEI or surgical parathyroidectomy at 17, 28, 46, and 48 months later. There was no significant reduction in iPTH in 11 patients following PFNEI after a mean of 2.5 +/- 1.3 injections per gland. They all required a subsequent surgical parathyroidectomy for symptomatic hyperparathyroidism. Four patients developed a laryngeal nerve palsy following PFNEI, two of which were permanent. Seventeen patients underwent successful surgical parathyroidectomy as a primary procedure. CONCLUSION: Whilst PFNEI is successful in primary hyperparathyroidism, when typically only one adenoma is present, the effectiveness of PFNEI is unpredictable and the long term results are poor compared with those of surgical parathyroidectomy in tertiary hyperparathyroidism. The procedure is not without complications and makes subsequent surgery more difficult. Therefore it can only be recommended for patients with a known single parathyroid gland such as patients in whom hyperparathyroidism has recurred following a previous surgical subtotal parathyroidectomy and who are unsuitable for further surgery.

Percentage hypochromic red cells and the response to intravenous iron therapy in anaemic haemodialysis patients.

INTRODUCTION: Iron deficiency is commonly encountered in haemodialysis (HD) patients and may be overcome by i.v. iron therapy. We have examined the percentage hypochromic red cells (%HRC) for predicting response to i.v. iron in subjects with a low serum ferritin. METHODS: Prospective study of i.v. iron saccharate (trivalent iron 200 mg/week for 8 weeks) in anaemic (Hb < 10 g/dl) HD patients with serum ferritin < 100 microg/l despite oral iron therapy. Response to i.v. iron was assessed by comparing Hb at 0 and 8 weeks according to %HRC at baseline (0-3%, 4-9%, > or = 10%). Results are mean+/-1 SD. RESULTS: For all subjects (n=82), Hb and ferritin increased between 0 and 8 weeks (8.9+/-1.0 to 10.1+/-1.4, P<0.0001; 55+/-24 to 288+/-126, P<0.0001). Patients were stratified into three groups according to %HRC at baseline (0-3%, 4-9%, > or = 10%). Hb increased significantly in all three groups. The mean increase in Hb was greater (0-3%, 0.6+/-1.2; 4-9%, 1.2+/-1.0; > or = 10%, 1.6+/-1.4; P=0.02) and the proportion of patients showing a > or = 1 g/dl increase in Hb was greater (0-3%, 27%; 4-9%, 57%; > or = 10%, 67%; P=0.02) in those with the largest %HRC pre-treatment. CONCLUSION: Intravenous iron therapy is effective in improving Hb in anaemic HD patients with a low ferritin. However, the magnitude of this response and the proportion of patients responding is related to the percentage hypochromic red cells prior to treatment.

Hemodynamics and survival of patients with acute renal failure treated by continuous dialysis with two synthetic membranes.

Synthetic membranes are not identical and have specific interactions that may be harmful or beneficial. We have investigated the incidence of hypotension and the outcome of acute renal failure (ARF) in ventilated patients treated by continuous venovenous dialysis with 2 different synthetic membranes. In Study 1, the mean arterial pressure (MAP) and systemic vascular resistance (SVR) were monitored during the first 12 min of dialysis with polyacrylonitrile (PAN). In Study 2, the MAP and survival rates were compared in patients randomly assigned to either PAN or polysulfone. No subjects were receiving angiotensin converting enzyme inhibitors. In Study 1, the MAP decreased due to a reduction in the SVR during the first 6 min of dialysis but returned to the baseline value by 12 min in 22 patients during 27 dialysis treatments. In Study 2, the MAP was lower than the baseline value at 6 min during 233 dialysis treatments in 133 patients randomly assigned to PAN or polysulfone membranes (PAN group, 81.5 +/- 15 to 78.7 +/- 15.6 mm Hg, p = 0.001; and polysulfone group, 81.3 +/- 15.4 to 80.0 +/- 15.7 mm Hg, p = 0.06). Severe reductions in the MAP were seen during 13.2% of the PAN and 7.2% of the polysulfone treatments (chi 2, p = NS). The age, APACHE II score, MAP, inotrope requirement, and primary diagnosis did not differ according to membrane material in a total of 197 consecutive patients (PAN, n = 97; polysulfone, n = 100). Patients survival was 29% (PAN) and 27% (polysulfone). In multivariate analysis, APACHE II score, inotrope requirement, and liver failure were significant determinants of survival. In conclusion, PAN and polysulfone membranes were not different with respect to hypotensive reactions or survival in critically ill patients undergoing continuous venovenous hemodialysis.