Nancy Lew | Boston College (original) (raw)

Papers by Nancy Lew

Journal of the American Society of Nephrology : JASN, 1997

Despite the prevalent use of recombinant human erythropoietin (rhEPO), anemia is a frequent findi... more Despite the prevalent use of recombinant human erythropoietin (rhEPO), anemia is a frequent finding in hemodialysis patients. The goal of this study was to evaluate the impact of anemia on patient survival and characterize the determinants of hematopoiesis that may be amenable to therapeutic manipulation to enhance rhEPO responsiveness and reduce death risk. Patient characteristics and laboratory data were collected for 21,899 patients receiving hemodialysis three times per week in dialysis centers throughout the United States in 1993. Hemoglobin concentrations (Hb) < or =80 g/L were associated with a twofold increase in the odds of death (odds ratio = 2.01; P = 0.001) when compared with Hb 100 to 110 g/L. No improvement in the odds of death was afforded for Hb >110 g/L. Using multiple linear regression, variables of rhEPO administration (rhEPO dose and percentage of treatments that rhEPO was administered), variables of iron status (serum iron, transferrin saturation, and ferr...

American journal of kidney diseases : the official journal of the National Kidney Foundation, 1994

The mortality rate for hemodialysis patients in the United States is higher than in other industr... more The mortality rate for hemodialysis patients in the United States is higher than in other industrialized countries. Some attribute this to insufficient quantities of prescribed and delivered dialysis. A multicenter study in Dallas dialysis centers (Dallas Nephrology Associates) was begun in 1989 to assess the impact of increasing the delivered quantity of dialysis on mortality in subsequent years. Dialysis dose was measured by urea kinetic modeling. Kt/V, reflecting the fractional volume of body water clearance of urea during a dialysis treatment, was purposefully increased from 1.18 starting in 1989 to 1.46 in 1992. Additionally, the dialysis dose measured by the urea reduction ratio, the fractional reduction of blood urea nitrogen concentration caused by a dialysis treatment, increased from 63.0% to 69.6% between 1990 and 1992. Outcome analytical methods included both crude and standardized mortality rates and mortality ratios standardized to large end-stage renal disease database...

American journal of kidney diseases : the official journal of the National Kidney Foundation, 1994

Hemodialysis patients who had received intradialytic parenteral nutrition (IDPN) during 1991 were... more Hemodialysis patients who had received intradialytic parenteral nutrition (IDPN) during 1991 were identified. These patients were compared with unexposed controls after adjusting for demographic variables, baseline renal diagnosis, diabetic status, serum albumin (ALB), creatinine (CRE), and urea reduction ratio. At lower levels of ALB (< or = 3.4 g/dL), treatment with IDPN was associated with a reduction in the odds of death at 1 year, an effect that became stronger at lower levels of CRE (< or = 8.0 mg/dL). In contrast, treatment with IDPN in patients with normal ALB was associated with increased mortality. Time trend analyses of ALB and CRE demonstrated progressive increases toward pretreatment levels in IDPN recipients that were not evident in control subjects. These time trend data suggest that in undernourished hemodialysis patients, IDPN can effect the serum levels of valid biochemical surrogates of visceral and somatic protein nutrition. Albeit retrospective, the improv...

Kidney International, 1997

Development of a population specific regression equation to estimate total body water in hemodial... more Development of a population specific regression equation to estimate total body water in hemodialysis patients. We have previously shown that the impedance index (height corrected resistance) is a valid and reliable correlate of total body water (TBW) in hemodialysis patients. We estimated TBW by single frequency bioelectrical impedance analysis (BIA) in 3009 in-center hemodialysis patients, and developed an ESRDspecific TBW equation from routinely available demographic and anthropometric variables. The mean SD age was 60.5 15.5 years; 47% were female, 47% African-American, and 36% diabetic. Dialysis duration was 3.8 3.7 years. Mean TBW was 40.8 9.3 kg, 56 9% of body weight. A stepwise linear regression equation was fit on a two-thirds random sample, deriving significant parameter estimates for the variables age, gender, height, weight, diabetic status, weight squared, and the crossproducts of age and gender, age and weight, gender and weight, and height and weight. The equation was then validated in the remaining one-third sample, and compared with TBW estimates by the Watson and Hume-Weycr formulae. TBW estimated by our equation (40.6 8.6 kg) was not significantly different from the BIA TBW (40.5 9.3 kg). In contrast, TBW estimated by the Watson (37.0 7.6 kg) and Hume-Weyer (37.9 7.7 kg) formulae underestimated TBW by a mean of 3.5 and 2.6 kg, respectively. A population-specific equation provides superior prediction of TBW in hemodialysis patients. The use of formulae developed and validated in non-uremic populations may result in underestimates of TBW in patients with ESRD, and potentially, overestimates of dialysis dose approximated by the clearance-time to TBW ratio (Kt/V). TBW resides in the skeletal muscle, TBW may serve as a proxy for somatic protein stores [1]. In end-stage renal disease (ESRD), TBW is of additional importance, as it approximates the volume of distribution of urea. An estimate of TBW is frequently used to calculate the clearance time-product (Kt) to volume (V) ratio, a marker of dialysis adequacy. This ratio can be misinterpreted if V is over-or underestimated [21. Several means of estimating the urea volume of distribution are commonly used. The formal method uses

Kidney International, 2000

6% increase in the risk of death, all else equal. Longitudinal Vintage, nutritional status, and s... more 6% increase in the risk of death, all else equal. Longitudinal Vintage, nutritional status, and survival in hemodialysis paassessments of nutritional status, including body composition, tients. are required to better understand the natural history of wasting Background. The link between dialysis "vintage" (length of with ESRD and its implications for long-term survival. time on dialysis in months to years) and survival has been difficult to define, largely because of selection effects. Endstage renal disease (ESRD) is thought to be a wasting illness, but there are no published reports describing the associations The link between dialysis "vintage" (length of time on between vintage and body composition in hemodialysis patients. dialysis in months to years) and survival of hemodialysis Methods. We explored the relationships among vintage, nupatients has been difficult to define, largely because of tritional status, and survival in a 3009 patient cohort of prevalent hemodialysis patients. Body weight, total body water, body selection effects. Over time, patients leave the dialysis cell mass, and phase angle by bioelectrical impedance analysis realm either because of transplantation or death, so that were the body composition parameters of interest. We examthe steady-state (that is, prevalent) dialysis population ined vintage as an explanatory variable in multiple linear rediffers considerably from the incident population. Acgression analyses (adjusted for age, gender, race, and diabetes) cordingly, patients of long vintage might be expected using body composition parameters and biochemical indicators of nutritional status as dependent variables. Proportional hazto differ greatly from patients of short vintage. Several ards regression was used to evaluate the association of vintage reports have described clinical characteristics of "longand survival with and without adjustment for case mix and term dialysis survivors." [1-3]. In general, long-term surlaboratory variables. vivors tend to be younger in age at the time of dialysis Results. Dialysis vintage was 3.8 Ϯ 3.7 (median 2.6) years. initiation and African American rather than white. Pri-Body composition parameters tended to be lower after dialysis year 2. Linear estimates per year of vintage beyond year 2 mary renal diseases are more likely to be glomeruloneinclude Ϫ0.66 kg body wt (P Ͻ 0.0001), Ϫ0.17 kg total body phritis or polycystic kidney disease rather than diabetes water (P ϭ 0.0003), Ϫ0.14 kg body cell mass (P Ͻ 0.0001), mellitus, hypertensive nephrosclerosis, or other condiand Ϫ0.07 degrees phase angle (P Ͻ 0.0001). In unadjusted tions. analyses, vintage was not associated with survival, either as a Few investigators have explored the association of vinlinear or higher order term. The adjustment for case mix yielded a vintage term associated with an increased relative tage with nutritional status in end-stage renal disease risk (RR) of death (RR 1.04 (95% CI, 1.01 to 1.07 per year). (ESRD). Avram et al compared long-term (Ͼ10 years) A further adjustment for laboratory data yielded a RR of 1.06 with "average" (Ͻ5 years) survivors on dialysis, and (95% CI, 1.03 to 1.09 per year). found that the serum albumin and creatinine concentra-Conclusion. Dialysis vintage is related to nutritional status tions at dialysis enrollment were higher in the 86 patients in hemodialysis patients, with vintage of more than years associated with a significant decline in all measured nutritional pa-(58 hemodialysis and 28 peritoneal dialysis) of extended rameters. Cross-sectional analyses probably underestimate vintage (Ͼ10 years) [3]. Lowrie and Lew found an inverse these effects. A year accrued on dialysis is associated with a relationship between vintage and death risk in prevalent hemodialysis patients during 1988 [4] and 1989 [5]. Bloembergen et al found no association between vintage and

Kidney International, 1997

Bioimpedance norms for the hemodialysis population. More than 3,000 hemodialysis patients were ex... more Bioimpedance norms for the hemodialysis population. More than 3,000 hemodialysis patients were examined with single-frequency bioelectrical impedance analysis (BIA). Distributions of resistance, reactance, phase angle (PA), and estimates of total body water (TBW) and body cell mass (BCM) by BIA were determined, and compared with traditional laboratory markers of nutritional status. Bioimpedance parameters and body composition estimates differed significantly by age, sex, race, and diabetic status. PA and BCM correlated directly with serum creatinine, albumin, and prealbumin concentrations. Population-based norms for bioimpedance parameters and estimates of body composition are provided. Protein energy malnutrition afflicts a large fraction of hemodialysis patients [1-3] and is an important determinant of mortality and morbidity [4-8]. Several methods of nutritional assessment have been applied in this population, including estimates of dietary intake, anthropometry, and biochemical tests consisting of serum concentrations of creatinine, albumin, and prealbumin. These biochemical indicators have been repeatedly shown to predict survival in hemodialysis patients, although their levels can be confounded by other disease processes (such as liver disease), and they do not capture the entire realm of malnutrition. Body composition analysis has attracted some interest, however, most facilities.

Kidney International, 2001

Changing hemodialysis thresholds for optimal survival. Numerous studies have demonstrated an asso... more Changing hemodialysis thresholds for optimal survival. Numerous studies have demonstrated an association Background. The urea reduction ratio (URR), a measure between the amount of hemodialysis and mortality among quantitating solute removal during hemodialysis, is the fracpatients with end-stage renal disease (ESRD) [1-8]. In tional reduction of the blood urea concentration during a single general, patient mortality is higher when the hemodialyhemodialysis treatment. The URR is the principal measure of sis "dose" is low and is lower when the hemodialysis hemodialysis dose in the United States. Based on studies of patients dialyzed prior to 1994, a minimum URR value of 65% dose is high. The urea reduction ratio (URR) is a comwas recommended to optimize survival. Because of new hemomonly used measure of hemodialysis dose and is based dialysis technologies and evolving demographics of the hemoon the fractional reduction of blood urea nitrogen dialysis population, the relationship between the amount of (BUN) concentration during a single hemodialysis treathemodialysis and mortality was examined in contemporary cohorts. ment. It is calculated by dividing the fall of BUN (predial-Methods. This retrospective cohort included Ͼ15,000 patients ysis minus postdialysis BUN) by the predialysis and is per year receiving hemodialysis during 1994 through 1997. Each expressed as a percentage [1, 9-11]. Retrospective studpatient's URR was averaged for the three months prior to the ies of mortality for ESRD patients suggest that the odds beginning of each year. Mortality odds ratios were calculated of death increase progressively as the URR falls below for patients by URR. To determine the URR value above which no further improvement in mortality was seen ("thresh-60 to 65% [1-3]. Such findings and a professional consenold"), spline functions were tested in logistic regression models, sus have led three national organizations, including the both unadjusted and adjusted for case mix measures. The Health Care Financing Administration (HCFA), the strength of fit for URR, defined by a range of candidate threshprincipal payer of dialysis services, to advocate a URR olds from 55 to 75%, was evaluated in increments of 1% for of 65% as the threshold for "adequate" hemodialysis each year using spline functions. Results. The median URR was 63.2, 65.4, 67.4, and 68.1% and to profile providers accordingly [9, 10, 12-14]. for 1994 through 1997, respectively. The median length of hemo-Using URR as a measure of hemodialysis dose, substandialysis treatments increased only six minutes from the begintial improvement in the amount of hemodialysis has ocning to the end of the period of analysis. Using spline functions, curred in recent years. As reported in the 1998 National the threshold URR values were 61.1, 65.0, 68.0, and 71.0% for 1994 through 1997 in models adjusted for case mix. The ratio ESRD Core Indicators Report, a profile of patientof median URR to URR threshold decreased from 1.03 in specific dialysis practices, the mean URR for the United 1994 to 0.97 in 1997. States increased from 62.7% (deemed inadequate) to Conclusions. From 1994 to 1997, the median URR and the 68.0% (deemed adequate) from 1993 to 1997 [14]. Al-URR threshold for mortality benefit increased. Although an though mean URR provides a gauge of hemodialysis increased need in the amount of hemodialysis may be a consequence of changes in patients' demographic characteristics, the adequacy at the population level, the proportion of palikely explanation(s) is a change in the dialysis procedure and/or tients whose URR is Ն65% is of greater significance for blood sampling favoring higher URR values without changing optimizing individual patient survival. Using this clinical the amount of dialysis provided. The recommended minimum performance benchmark, the percentage of patients re-URR of 65% appears to be too low to confer an optimal mortality benefit in the context of current practices. ceiving a URR Ն65% rose from 43% in 1993 to 72% in 1997 [14]. The aforementioned outcome studies characterizing

Kidney International, 1999

Background. The current standard of adequacy for peritoor more [3]. Recognized weaknesses of thes... more Background. The current standard of adequacy for peritoor more [3]. Recognized weaknesses of these guidelines neal dialysis (PD) is to provide a weekly normalized urea include: (a) the lack of evidence to establish equivalency clearance (Kt/V) of 2.0 or more and a creatinine clearance between the renal and peritoneal contributions of small (C Cr) of 60 liter/1.73 m 2 or more. As native renal function is lost, it is important to determine the effectiveness of the available solute removal to clinical outcome; and (b) the uncertain therapeutic modalities in achieving these goals. value of urea versus creatinine as markers of uremia Methods. A model to assess our ability to provide a weekly [4, 5]. Most clinical outcome studies have included pa-Kt/V urea of 2.0 or more and a C Cr of 60 liter/1.73 m 2 or more tients at various stages of uremia therapy with different to anuric patients undergoing continuous ambulatory PD degrees of residual renal function (RRF). The most (CAPD) and automated PD (PD Plus) was developed. The body surface area (BSA) distribution was obtained from 38,768 quoted prospective study, CANUSA, was not designed patients undergoing dialysis during January 1997. The distributo evaluate the relative effect of RRF and the peritoneal tion of peritoneal transport rates (PTRs) was obtained from clearance on patient outcome [2, 6, 7]. Furthermore, 2531 peritoneal equilibration tests performed during 1996. The patients with significant RRF are more likely to achieve weekly K p t/V urea was calculated for the various PTR groups higher clearances of creatinine than anuric patients beand the range of BSA with four PD prescriptions: CAPD 8 liters, CAPD 10 liters, PD Plus 12 liters, and PD Plus 15 liters, cause of tubular secretion of creatinine in advanced renal using a previously validated kinetic program (PackPD). failure. Few prospective studies have been performed in Results. The predicted percentage of patients capable of anuric patients undergoing PD. However, Selgas et al achieving the adequacy goals for Kt/V and C Cr , respectively, studied patients who had a minimum of three years on

Kidney International, 2000

Prealbumin is as important as albumin in the nutritional assesstive risk (RR) of death between tw... more Prealbumin is as important as albumin in the nutritional assesstive risk (RR) of death between two-and tenfold or ment of hemodialysis patients. more, depending on the severity of PEM and the interac-Background. Although serum prealbumin is considered a tion of PEM with other factors (age, anemia, dialysis valid indicator of nutritional status in hemodialysis patients, vintage, etc.) [1-3]. Serum albumin has been the most there is relatively little evidence that its determination is of commonly employed marker of PEM, based largely on major prognostic significance. In this study, we aimed to determine the independent association of serum prealbumin with the statistical association between diminished serum alsurvival in hemodialysis patients, after adjusting for serum albumin, mortality, and morbidity. Indeed, the Health Care bumin and other indicators of protein energy nutritional status. Financing Administration's (HCFA) core clinical indica-Methods. Serum prealbumin was measured in more than 1600 tors for adequate dialysis treatment involves an audit of maintenance hemodialysis patients. We determined the correlations among prealbumin and other indicators of nutritional serum albumin concentration in dialysis facilities [4] costatus, including serum albumin, and bioimpedance-derived indiequal with anemia and measures reflecting the adequacy cators of body composition. The relationship between serum of dialysis. prealbumin and survival was determined using proportional The clinical approach to the patient with PEM within hazards regression.

Kidney International, 1999

malnutrition (manifest by a reduced V) that overcomes what-Exploring the reverse J-shaped curve b... more malnutrition (manifest by a reduced V) that overcomes what-Exploring the reverse J-shaped curve between urea reduction ever benefit might be derived from an associated increase in urea ratio and mortality. clearance. Identification of patients who achieve extremely Background. Although accepted worldwide as valid meahigh URR (Ͼ75%) or single-pooled Kt/V (Ͼ1.6) values using sures of dialysis adequacy, neither the Kt/V (urea clearance standard dialysis prescriptions should prompt a careful assessdetermined by kinetic modeling) nor the urea reduction ratio (URR) have unambiguously predicted survival in hemodialysis ment of nutritional status. Confounding by protein-calorie malpatients. Because the ratio Kt/V can be high with either high nutrition may limit the utility of URR or Kt/V as a population-Kt (clearance ϫ time) or low V (urea volume of distribution) based measure of dialysis dose. and V may be a proxy for skeletal muscle mass and nutritional health, we hypothesized that the increase in the relative risk of death observed among individuals dialyzed in the top 10 to Defining the dose of dialysis to optimize health and 20% of URR or Kt/V values might reflect a competing risk of malnutrition. functional status for patients with end-stage renal disease Methods. A total of 3,009 patients who underwent bioelec-(ESRD) has been among the principle goals of the netrical impedance analysis were stratified into quintiles of URR. phrology community for more than two decades. In a Laboratory indicators of nutritional status and two bioimpelandmark publication in 1985, Gotch and Sargent revoludance-derived parameters, phase angle and estimated total tionized the field of dialysis therapy with their reanalysis body water, were compared across quintiles. The relationship between dialysis dose and mortality was explored, with a focus of data from the National Cooperative Dialysis Study on how V influenced the structure of the dose-mortality rela-(NCDS) [1, 2], redefining the dose of dialysis in terms tionship. of the clearance (K) ϫ time (t) product normalized to Results. There were statistically significant differences in all the urea volume of distribution (V), a dimensionless nutritional parameters across quintiles of URR or Kt/V, indiquantity derived from the fractional reduction in urea cating that patients in the fifth quintile (mean URR, 74.4 Ϯ 3.1%) were more severely malnourished on average than panitrogen (URR) achieved over the course of a hemodialtients in all or some of the other quintiles. The relationship ysis treatment. These authors and others who followed between URR and mortality was decidedly curvilinear, resemthem were successful at identifying a minimum dose of bling a reverse J shape that was confirmed by statistical analysis. dialysis above which death and complications were less An adjustment for the influence of V on URR or Kt/V was performed by evaluating the Kt-mortality relationship. There likely, although the optimal dose of dialysis remained was no evidence of an increase in the relative risk of death unknown. Application of this methodology to other paamong patients treated with high Kt. Higher Kt was associated tient cohorts similarly showed that an increase in URR with a better nutritional status. or Kt/V was generally associated with reduced mortality Conclusion. We conclude that the increase in mortality oband less frequent complications [3-8]. Although it folserved among those patients whose URR or Kt/V are among the top 10 to 20% of patients reflects a deleterious effect of lowed that further increases in dialysis dose might be expected to lead to fewer complications and better outcomes, Owen et al initially showed that there appeared

Kidney International, 1999

ity in those patients who are in the lower 50th percentile for Body weight-for-height relationshi... more ity in those patients who are in the lower 50th percentile for Body weight-for-height relationships predict mortality in mainthis measurement. tenance hemodialysis patients. Background. Protein-energy malnutrition is a strong predictor of mortality in maintenance hemodialysis (MHD) patients. This association has generally been described for serum Many epidemiological studies have shown a correlachemistry measures of protein-energy malnutrition. We hytion between parameters of protein-energy nutritional pothesized that body weight-for-height relationships also predict survival in MHD patients. status and morbidity or mortality in maintenance dialysis Methods. During the last three months of 1993, data were patients. These observations, in general, have demonobtained on 12,965 men and women concerning clinical characstrated relationships between serum chemistry measures, teristics (height, postdialysis weight, age, gender, race, and dietary protein intake as indicated by the protein equivapresence or absence of diabetes mellitus) and laboratory mealent of protein nitrogen appearance (PNA, also called surements (predialysis serum albumin, creatinine and cholesterol, and the urea reduction ratio). Patient survival during the PCR), postdialysis body weight, and clinical outcome next 12 months was evaluated retrospectively. [1-6]. In maintenance hemodialysis (MHD) patients, the Results. In comparison to values for normal Americans deserum chemistry measurements which reflect nutritional termined from the National Health and Nutrition Evaluation status or nutrient intake and which correlate with mor-Survey II data, weight-for-height relationships tended to be bidity or mortality include predialysis serum albumin, slightly lower than normal in African American men and women and Caucasian men undergoing MHD and were normal creatinine, urea, cholesterol, potassium, and phosphorus or slightly greater in the taller Caucasian women. In both men [1, 2, 4, 5]. and women, the mortality rate decreased progressively as the These relationships may be particularly important for patients' weight-for-height increased. MHD patients who the following reasons: (a) Some of these nutritional paweighed more than normal had the lowest mortality rates. rameters, such as the serum albumin, strongly predict After adjustment for clinical characteristics and laboratory measurements, the inverse relationship between mortality rates outcome [2, 4, 5]. (b) Many MHD patients have evidence and weight-for-height percentiles was still highly significant for of protein-energy malnutrition [1-7]. (c) The mortality patients within the lower 50th percentile of body weight-forrate of MHD patients is very high [2, 4, 5, 8], averaging height. Serum albumin correlated directly with weight-for-height over 20% in the United States. (d) Certain nutritional in patients in the lower 50th percentile of weight-for-height. parameters are potentially modifiable, and there is a Serum creatinine and cholesterol correlated directly with weight-for-height in the entire population of men and women. possibility that altering these nutritional parameters may In contrast, the urea reduction ratio was inversely correlated improve outcome; this possibility has not yet been tested with weight-for-height. in randomized prospective clinical trials. Conclusions. These data indicate that weight-for-height is a Because body weight-for-height relationships predict strong predictor of 12-month mortality in male and female clinical outcome in healthy men and women [9-11], we MHD patients. Multivariate analyses indicate that body weight-for-height is an independent predictor of higher mortalassessed the relationship of weight-for-height to mortality in a retrospective analysis of data obtained from 12,965 MHD patients. Four primary questions were ad-Key words: protein-energy malnutrition, survival rate for hemodialysis, dressed: (a) How does the distribution of weight-fordialysis, albumin, creatinine, cholesterol, mortality prediction, maintenance hemodialysis. height and body mass index (BMI) in MHD patients compare with the normal population in the United

Kidney International, 1999

not) and also with increasing body size (whether adjusted for The urea {clearance ϫ dialysis time... more not) and also with increasing body size (whether adjusted for The urea {clearance ϫ dialysis time} product (Kt) as an outcome-Kt or not) for each estimate of size. Significant statistical interbased measure of hemodialysis dose. actions of Kt with gender, but not Kt with race, were observed Background. The normalized treatment ratio [Kt/V ϭ the ratio of the urea {clearance ϫ time} product to total body in all models. There were no statistical interactions, suggesting water] and the urea reduction ratio (URR) have become widely that higher Kt was routinely required with increasing body accepted measures of dialysis dose. Both are related to and size. Separate risk profiles for males and females suggested a derived from pharmacokinetic models of blood urea concentrahigher Kt threshold for males. tion during the dialysis cycle. Theoretical reconsideration of Conclusions. The urea {clearance ϫ time} is a valid outcomethe models revealed that the premise about V on which they based measure of dialysis dose and is not confounded by inrest (that is, that V is a passive diluent with no survival-associdexing it to an estimate of body size, which has outcomeated properties of its own) is flawed if the intended use of the associated properties of its own. Dialysis prescriptions for models is for profiling clinical outcome (for example, mortality) males and females should be regarded separately, but there rather than estimating urea concentration. As a proxy for body appears no need to make a distinction between the races. mass, V has survival-associated properties of its own. Thus, indexing {clearance ϫ time} to body size could create an offsetting combination whereby one measure favorably associated with survival (Kt) is divided by another (for example, V). The primary finding [1] and final report [2] of the Observed clinical paradoxes support that interpretation. For National Cooperative Dialysis Study (NCDS) were pubexample, patients with a low body mass have both higher URR lished more than 15 years ago. That study evaluated four and higher mortality than heavier patients. Increasing mortality groups of patients arranged in a 2 ϫ 2 factorial design is often observed at high URR, suggesting the possibility of "over-dialysis." Black patients tend to be treated at lower URR [3] where blood urea nitrogen (BUN) concentration, a than whites but enjoy better survival on dialysis. Therefore, proxy for "small molecule"-directed therapy, and the {clearance ϫ time} was evaluated as an outcome-based measure length of the dialysis treatment, a proxy for therapy of dialysis dose, not indexed to V, and various body size estidirected at larger molecules, were the factors [3, 4]. A mates were evaluated as separate and distinct measures. Methods. The retrospective sample included 17,141 black computer-assisted, single-pool, variable-volume urea kiand white hemodialysis patients treated three times per week. netic model [5-7] was used to facilitate the control of Logistic regression analysis was used to evaluate death odds in BUN at two levels (low and high) within two levels of age-, gender-, race-, and diabetes-adjusted models. Kt and five dialysis time (long and short). The urea kinetic equations body size estimates (total body water or V, body weight, body weight adjusted for height, body surface area, and body mass

Kidney International, 2001

It was stronger for the albumin Indicator (R 2 ϭ 11.6, 20.4, Association between prevalent care p... more It was stronger for the albumin Indicator (R 2 ϭ 11.6, 20.4, Association between prevalent care process measures and facil-21.8). The fractions of patients falling outside of the Indicator ity-specific mortality rates. guidelines tended to be higher in the highest SMR group. The Background. Medical communities often develop practice groups were not well separated, however, particularly for the guidelines recommending certain care processes intended to hematocrit and URR Indicators, and there was substantial promote better clinical outcome among patients. Conformance overlap between them. Finally, although the likelihood that a with those guidelines by facilities is then monitored to evaluate facility would be a member of the high or low SMR group was care quality, presuming that the process is associated with and associated with fractional variance from Core Indicator guidecan be used reliably to predict clinical outcome. Outcome is lines, the strengths of association were weak, and the probabiloften monitored as a facility-specific mortality rate (SMR) stanity that a facility would be a member of the high or low group dardized to the mix of patients treated, also presuming that could not be easily distinguished from the probability that it inferior outcome implies a suboptimal process. The U.S. Health would be a member of the middle group. Care Financing Administration monitors three practice guide-Conclusions. While there were statistical associations belines, called Core Indicators, in dialysis facilities to assist mantween SMR and the fraction of patients in facilities who were agement of its end-stage renal disease program: (1) patients' at variance with these guidelines, they were weak and variances hematocrit values should exceed 30 vol%, (2) the urea reducfrom the guidelines could not be used reliably to predict high or tion ratio (URR) during dialysis should equal or exceed 65%, low SMR. Such findings do not imply that measures reflecting and (3) patients' serum albumin concentrations should equal anemia, dialysis dose, or medical processes that influence serum or exceed 3.5 g/dL. albumin concentration are irrelevant to the quality of care. They Methods. The associations of a facility-specific SMR were evaldo suggest, however, that more attention needs be paid to these uated with the fractions of hemodialysis patients not conformand other associates and causes of mortality among dialysis ing to (that is, at variance with) the Core Indicators during three patients when developing care process indicator guidelines. successive years (1993 to 1995) in large numbers of facilities (394, 450, and 498) using one-variable and multivariable statistical models. Three related strategies were used. First, the association of the SMR with the fraction of patients not meeting Process measures of care, such as the Health Care Fithe guideline was evaluated. Second, each facility was classified nancing Administration's (HCFA) Core Indicators, are by whether its SMR exceeded the 80% confidence interval selected measures thought to affect the outcome of care above 1.0 (worse than 1.0, Group 3), was less than the inter-[1-3]. They are sometimes monitored within preselected val below 1.0 (better than 1.0, Group 1), or was within the interval (Group 2). The fraction of those patients who did not limits hoping to influence the quality and outcome of care meet the Indicator guidelines was then evaluated in each group. [4]. The Core Indicators were used for that purpose. Thus, Third, the ability of variance from Indicator guidelines to prethey should be strongly associated with and predict clinidict into which of the three SMR groups a facility would be cal outcome. That is particularly true if they are the main categorized was evaluated. or exclusive measures used to judge the care quality among Results. SMR was directly correlated with variance from the Indicator guidelines, but the strengths of the associations were facilities. It is necessary, therefore, to understand and demweak particularly for the hematocrit (R 2 ϭ 2.2%, 5.6, and 2.2 onstrate the nature and strength of association between for each of the 3 years) and URR Indicators (R 2 ϭ 2.6, 0.6, 3.3). outcome and those measures selected for monitoring in order to ensure that actions taken pursuant to monitoring will likely be effective and efficient in achieving the

Kidney International, 2001

Cardiac arrest and sudden death in dialysis units. mon cause of death, with nearly half of deaths... more Cardiac arrest and sudden death in dialysis units. mon cause of death, with nearly half of deaths attributed Background. For patients with end-stage renal disease and to myocardial infarction, cardiac arrest, or other cardiac their providers, dialysis unit-based cardiac arrest is the most causes [2]. Numerous studies have looked at predictors feared complication of hemodialysis. However, relatively little of cardiac mortality in maintenance hemodialysis pais known regarding its frequency or epidemiology, or whether tients. A history of ischemic heart disease, congestive a fraction of these events could be prevented. Methods. To explore clinical correlates of dialysis unit-based heart failure, left ventricular hypertrophy (LVH) and dyscardiac arrest, 400 reported arrests over a nine-month period function, peripheral vascular disease, hypertension or hyfrom October 1998 through June 1999 were reviewed in detail. potension, hypoalbuminemia, dyslipidemia, anemia, and Clinical characteristics of patients who suffered cardiac arrest inflammation (elevated C-reactive protein) have all been were compared with a nationally representative cohort of associated with increased cardiovascular mortality [3-9]. Ͼ77,000 hemodialysis patients dialyzed at Fresenius Medical Care North America-affiliated facilities. Using data from the U.S. Renal Data System Case Results. The cardiac arrest rate was 400 out of 5,744,708, Mix Adequacy Study, Bleyer, Russell, and Satko docucorresponding to a rate of 7 per 100,000 hemodialysis sessions. mented a significantly higher proportion of sudden Cardiac arrest was more frequent during Monday dialysis sesdeaths in hemodialysis patients on Monday and Tuesday sions than on other days of the week. Case patients were nearly twice as likely to have been dialyzed against a 0 or 1.0 mEq/L compared with other days of the week [10]. No other potassium dialysate on the day of cardiac arrest (17.1 vs. 8.8%). risk factors for sudden death were identified. The authors Patients who suffered a cardiac arrest were on average older postulated that the accumulation of electrolytes and flu-(66.3 Ϯ 12.9 vs. 60.2 Ϯ 15.4 years), more likely to have diabetes ids between dialysis sessions increased the susceptibility (61.8 vs. 46.8%), and more likely to use a catheter for vascular to arrhythmias, especially after a weekend respite from access (34.1 vs. 27.8%) than the general hemodialysis population. Sixteen percent of patients experienced a drop in systolic dialysis. No data were available on the timing of sudden pressure of 30 mm Hg or more prior to the arrest. Thirtydeath relative to the hemodialysis session. Presumably, seven percent of patients who suffered cardiac arrest had been the majority of sudden death cases occurred at home hospitalized within the past 30 days. Sixty percent of patients rather than during dialysis. died within 48 hours of the arrest, including 13% while in the dialysis unit. Relatively few studies have examined the epidemiol-Conclusions. Cardiac arrest is a relatively infrequent but ogy of in-dialysis unit cardiac arrest. Moreover, no prior devastating complication of hemodialysis. To reduce the risk investigation has had sufficient power to estimate its inof adverse cardiac events on hemodialysis, the dialysate precidence accurately or to identify patient-and hemodialyscription should be evaluated and modified on an ongoing sis-specific factors that might be associated with a higher basis, especially following hospitalization in high-risk patients. than baseline risk. We therefore undertook this study of cardiac arrest during hemodialysis using data obtained directly from dialysis units at the time of the event. We Hemodialysis patients have an age-adjusted mortality were most interested in identifying modifiable risk facrate that is 3.5 times higher than that of the general population [1]. Cardiovascular disease remains the most comtors so that some future events might be prevented.

Journal of Renal Nutrition, 1997

To determine the relation between phase angle by bioelectrical impedance analysis (BIA) and survi... more To determine the relation between phase angle by bioelectrical impedance analysis (BIA) and survival in hemodialysis patients. Design: Cohort analytic study. Setting: One hundred one free-standing outpatient dialysis units. Patients: Three thousand nine adult patients on thrice weekly hemodialysis. Patients with amputations above the transmetatarsal site were excluded from participation. Main Outcome Measure: Vital status, with follow-up to at least 1 year. Results: Mean phase angle was 4.8 2: 1.8 degrees. Patients with narrow (low) phase angle experienced an increased relative risk (RR) of death «3 degrees; RR 4.3; 95% confidence interval [CI], 2.9-6.2; and 3 to 4 degrees); RR 2.2; 95% CI, 1.6-3.2; compared with the 2::6 degrees reference). There were no significant differences in risk among patients with phase angle 4 to 5 degrees (RR 1.2; 95% CI, 0.8-1.8), 5 to 6 degrees (RR 1.1; 95% CI, 0.7-1.7), and 2::6 degrees, suggesting a nonlinear relation between phase angle and survival. The RRs for phase angle <4 degrees remained statistically significant after adjusting for age, gender, race, serum albumin and creatinine concentrations, and dialysis intensity «3 degrees, RR 2.2; 95% CI, 1.6-3.1, and 3 to 4 degrees, RR 1.3; 95% CI, 1.0-1.7, compared with all patients 2::4 degrees). Conclusions: In patients on hemodialysis, BIA-derived phase angle <4 degrees was associated with an increased RR of death, even after adjustment for case mix and several nutritional indicators. Further research is required to determine whether BIA can be used to monitor health status over time, or to gauge response to nutrition support or other clinical interventions in patients with end-stage renal disease.

Journal of Proteome Research, 2005

In this paper, we report an experimental setup and mathematical algorithm for determination of re... more In this paper, we report an experimental setup and mathematical algorithm for determination of relative protein abundance from directly labeled native protein samples applied to an array of antibodies. The application of the proposed experimental system compensates internally at each array element for a number of deficiencies in array experiments such as differential labeling efficiency in dual color assay systems, differential solubility of protein molecules in dual color assay systems, and differential affinity of capture reagents toward proteins labeled with two different fluorescent dyes. This system offers full compensation for variable amounts of capture reagents on separate array structures, as well as limited compensation for nonspecific interactions between capture reagents and analytes. The proposed experimental strategy enables the use of a large number of capture reagents to develop a true multiplex analysis system that will yield complete relative protein abundance information in two biological systems.

JAMA: The Journal of the American Medical Association, 1995

Hemodialysis International, 2003

Background: Consensus processes using the clinical literature as the primary source for informati... more Background: Consensus processes using the clinical literature as the primary source for information generally drive projects to draft clinical practice guidelines (CPGs). Most such literature citations describe special projects that are not part of an organized quality management initiative, and the publication/review/consensus process tends to be long. This project describes an initiative to develop and explore a flexible and dedicated data-driven paradigm for deciding new CPGs that could be rapidly responsive to changing medical knowledge and practice. Methods: Candidate Clinical Practice Monitoring Measures (CPMM) were selected using a large, national database according to the natures and strengths of their associations with mortality risk among patients during 1994. Thresholds above or below which risk of death increased were evaluated for each CPMM using risk profile charts and spline functions. The fractions of patients outside of those thresholds in each dialysis unit (the %Var) were determined for the years 1993, 1994, and 1995. A standardized mortality ratio (SMR) was also determined for each year for each facility. The associations between the %Var and SMR were evaluated in several single-variable and multivariable statistical models. Results: Eleven CPMM were selected and evaluated based on their associations with death risk. These included the urea clearance x dialysis time product (Kt); the concentrations of albumin, potassium, phosphate, bicarbonate, hemoglobin, neutrophils, and lymphocytes in the blood; the body weight/height ratio; diastolic blood pressure; and vascular access type. Even though the CPMM were strongly associated with death risk among patients, the %Var were weakly and inconsistently associated with SMR among facilities. Conclusions: The paradigm was flexible, easy to implement, quickly executed, and potentially able to accommodate evolving medical practice assuming the availability of large database systems such as this. The primary associates of death risk were easily identified and the thresholds easily adopted. The SMR and %Var from the CPMM were only weakly associated, however, suggesting that one cannot be reliably predicted from the other. As such, quality management programs should likely monitor both the processes and outcomes of care among dialysis facilities.

Archives of Internal Medicine, 1991

The annual mortality rate among patients receiving long-term hemodialysis has been rising over th... more The annual mortality rate among patients receiving long-term hemodialysis has been rising over the past decade. The prevalences of known risk factors such as older age, male sex, duration of dialysis, presence of diabetes, coronary artery disease, or hypertension do not seem to have changed during this time. However, evidence suggests that an increased body aluminum level may have an adverse effect on survival even in the absence of overt aluminum toxic reaction. Therefore, we evaluated the correlation between serum aluminum levels and mortality in 10 646 patients undergoing long-term hemodialysis. Mortalities were 18% higher for patients with serum aluminum levels between 1520 and 2220 nmol/L and progressively increased to 60% higher for patients with aluminum levels above 7410 nmol/L. Serum aluminum level was an important predictor of survival even after other known risk factors had been controlled. These data strongly suggest that patients undergoing long-term hemodialysis should have periodic surveillance of the serum aluminum levels, and in those patients who have plasma levels of 1520 to 2220 nmol/L or higher, one should seriously consider discontinuing aluminum salts and giving therapy to decrease body aluminum level if it is found to be increased.

Journal of the American Society of Nephrology : JASN, 1997

Despite the prevalent use of recombinant human erythropoietin (rhEPO), anemia is a frequent findi... more Despite the prevalent use of recombinant human erythropoietin (rhEPO), anemia is a frequent finding in hemodialysis patients. The goal of this study was to evaluate the impact of anemia on patient survival and characterize the determinants of hematopoiesis that may be amenable to therapeutic manipulation to enhance rhEPO responsiveness and reduce death risk. Patient characteristics and laboratory data were collected for 21,899 patients receiving hemodialysis three times per week in dialysis centers throughout the United States in 1993. Hemoglobin concentrations (Hb) < or =80 g/L were associated with a twofold increase in the odds of death (odds ratio = 2.01; P = 0.001) when compared with Hb 100 to 110 g/L. No improvement in the odds of death was afforded for Hb >110 g/L. Using multiple linear regression, variables of rhEPO administration (rhEPO dose and percentage of treatments that rhEPO was administered), variables of iron status (serum iron, transferrin saturation, and ferr...

American journal of kidney diseases : the official journal of the National Kidney Foundation, 1994

The mortality rate for hemodialysis patients in the United States is higher than in other industr... more The mortality rate for hemodialysis patients in the United States is higher than in other industrialized countries. Some attribute this to insufficient quantities of prescribed and delivered dialysis. A multicenter study in Dallas dialysis centers (Dallas Nephrology Associates) was begun in 1989 to assess the impact of increasing the delivered quantity of dialysis on mortality in subsequent years. Dialysis dose was measured by urea kinetic modeling. Kt/V, reflecting the fractional volume of body water clearance of urea during a dialysis treatment, was purposefully increased from 1.18 starting in 1989 to 1.46 in 1992. Additionally, the dialysis dose measured by the urea reduction ratio, the fractional reduction of blood urea nitrogen concentration caused by a dialysis treatment, increased from 63.0% to 69.6% between 1990 and 1992. Outcome analytical methods included both crude and standardized mortality rates and mortality ratios standardized to large end-stage renal disease database...

American journal of kidney diseases : the official journal of the National Kidney Foundation, 1994

Hemodialysis patients who had received intradialytic parenteral nutrition (IDPN) during 1991 were... more Hemodialysis patients who had received intradialytic parenteral nutrition (IDPN) during 1991 were identified. These patients were compared with unexposed controls after adjusting for demographic variables, baseline renal diagnosis, diabetic status, serum albumin (ALB), creatinine (CRE), and urea reduction ratio. At lower levels of ALB (< or = 3.4 g/dL), treatment with IDPN was associated with a reduction in the odds of death at 1 year, an effect that became stronger at lower levels of CRE (< or = 8.0 mg/dL). In contrast, treatment with IDPN in patients with normal ALB was associated with increased mortality. Time trend analyses of ALB and CRE demonstrated progressive increases toward pretreatment levels in IDPN recipients that were not evident in control subjects. These time trend data suggest that in undernourished hemodialysis patients, IDPN can effect the serum levels of valid biochemical surrogates of visceral and somatic protein nutrition. Albeit retrospective, the improv...

Kidney International, 1997

Development of a population specific regression equation to estimate total body water in hemodial... more Development of a population specific regression equation to estimate total body water in hemodialysis patients. We have previously shown that the impedance index (height corrected resistance) is a valid and reliable correlate of total body water (TBW) in hemodialysis patients. We estimated TBW by single frequency bioelectrical impedance analysis (BIA) in 3009 in-center hemodialysis patients, and developed an ESRDspecific TBW equation from routinely available demographic and anthropometric variables. The mean SD age was 60.5 15.5 years; 47% were female, 47% African-American, and 36% diabetic. Dialysis duration was 3.8 3.7 years. Mean TBW was 40.8 9.3 kg, 56 9% of body weight. A stepwise linear regression equation was fit on a two-thirds random sample, deriving significant parameter estimates for the variables age, gender, height, weight, diabetic status, weight squared, and the crossproducts of age and gender, age and weight, gender and weight, and height and weight. The equation was then validated in the remaining one-third sample, and compared with TBW estimates by the Watson and Hume-Weycr formulae. TBW estimated by our equation (40.6 8.6 kg) was not significantly different from the BIA TBW (40.5 9.3 kg). In contrast, TBW estimated by the Watson (37.0 7.6 kg) and Hume-Weyer (37.9 7.7 kg) formulae underestimated TBW by a mean of 3.5 and 2.6 kg, respectively. A population-specific equation provides superior prediction of TBW in hemodialysis patients. The use of formulae developed and validated in non-uremic populations may result in underestimates of TBW in patients with ESRD, and potentially, overestimates of dialysis dose approximated by the clearance-time to TBW ratio (Kt/V). TBW resides in the skeletal muscle, TBW may serve as a proxy for somatic protein stores [1]. In end-stage renal disease (ESRD), TBW is of additional importance, as it approximates the volume of distribution of urea. An estimate of TBW is frequently used to calculate the clearance time-product (Kt) to volume (V) ratio, a marker of dialysis adequacy. This ratio can be misinterpreted if V is over-or underestimated [21. Several means of estimating the urea volume of distribution are commonly used. The formal method uses

Kidney International, 2000

6% increase in the risk of death, all else equal. Longitudinal Vintage, nutritional status, and s... more 6% increase in the risk of death, all else equal. Longitudinal Vintage, nutritional status, and survival in hemodialysis paassessments of nutritional status, including body composition, tients. are required to better understand the natural history of wasting Background. The link between dialysis "vintage" (length of with ESRD and its implications for long-term survival. time on dialysis in months to years) and survival has been difficult to define, largely because of selection effects. Endstage renal disease (ESRD) is thought to be a wasting illness, but there are no published reports describing the associations The link between dialysis "vintage" (length of time on between vintage and body composition in hemodialysis patients. dialysis in months to years) and survival of hemodialysis Methods. We explored the relationships among vintage, nupatients has been difficult to define, largely because of tritional status, and survival in a 3009 patient cohort of prevalent hemodialysis patients. Body weight, total body water, body selection effects. Over time, patients leave the dialysis cell mass, and phase angle by bioelectrical impedance analysis realm either because of transplantation or death, so that were the body composition parameters of interest. We examthe steady-state (that is, prevalent) dialysis population ined vintage as an explanatory variable in multiple linear rediffers considerably from the incident population. Acgression analyses (adjusted for age, gender, race, and diabetes) cordingly, patients of long vintage might be expected using body composition parameters and biochemical indicators of nutritional status as dependent variables. Proportional hazto differ greatly from patients of short vintage. Several ards regression was used to evaluate the association of vintage reports have described clinical characteristics of "longand survival with and without adjustment for case mix and term dialysis survivors." [1-3]. In general, long-term surlaboratory variables. vivors tend to be younger in age at the time of dialysis Results. Dialysis vintage was 3.8 Ϯ 3.7 (median 2.6) years. initiation and African American rather than white. Pri-Body composition parameters tended to be lower after dialysis year 2. Linear estimates per year of vintage beyond year 2 mary renal diseases are more likely to be glomeruloneinclude Ϫ0.66 kg body wt (P Ͻ 0.0001), Ϫ0.17 kg total body phritis or polycystic kidney disease rather than diabetes water (P ϭ 0.0003), Ϫ0.14 kg body cell mass (P Ͻ 0.0001), mellitus, hypertensive nephrosclerosis, or other condiand Ϫ0.07 degrees phase angle (P Ͻ 0.0001). In unadjusted tions. analyses, vintage was not associated with survival, either as a Few investigators have explored the association of vinlinear or higher order term. The adjustment for case mix yielded a vintage term associated with an increased relative tage with nutritional status in end-stage renal disease risk (RR) of death (RR 1.04 (95% CI, 1.01 to 1.07 per year). (ESRD). Avram et al compared long-term (Ͼ10 years) A further adjustment for laboratory data yielded a RR of 1.06 with "average" (Ͻ5 years) survivors on dialysis, and (95% CI, 1.03 to 1.09 per year). found that the serum albumin and creatinine concentra-Conclusion. Dialysis vintage is related to nutritional status tions at dialysis enrollment were higher in the 86 patients in hemodialysis patients, with vintage of more than years associated with a significant decline in all measured nutritional pa-(58 hemodialysis and 28 peritoneal dialysis) of extended rameters. Cross-sectional analyses probably underestimate vintage (Ͼ10 years) [3]. Lowrie and Lew found an inverse these effects. A year accrued on dialysis is associated with a relationship between vintage and death risk in prevalent hemodialysis patients during 1988 [4] and 1989 [5]. Bloembergen et al found no association between vintage and

Kidney International, 1997

Bioimpedance norms for the hemodialysis population. More than 3,000 hemodialysis patients were ex... more Bioimpedance norms for the hemodialysis population. More than 3,000 hemodialysis patients were examined with single-frequency bioelectrical impedance analysis (BIA). Distributions of resistance, reactance, phase angle (PA), and estimates of total body water (TBW) and body cell mass (BCM) by BIA were determined, and compared with traditional laboratory markers of nutritional status. Bioimpedance parameters and body composition estimates differed significantly by age, sex, race, and diabetic status. PA and BCM correlated directly with serum creatinine, albumin, and prealbumin concentrations. Population-based norms for bioimpedance parameters and estimates of body composition are provided. Protein energy malnutrition afflicts a large fraction of hemodialysis patients [1-3] and is an important determinant of mortality and morbidity [4-8]. Several methods of nutritional assessment have been applied in this population, including estimates of dietary intake, anthropometry, and biochemical tests consisting of serum concentrations of creatinine, albumin, and prealbumin. These biochemical indicators have been repeatedly shown to predict survival in hemodialysis patients, although their levels can be confounded by other disease processes (such as liver disease), and they do not capture the entire realm of malnutrition. Body composition analysis has attracted some interest, however, most facilities.

Kidney International, 2001

Changing hemodialysis thresholds for optimal survival. Numerous studies have demonstrated an asso... more Changing hemodialysis thresholds for optimal survival. Numerous studies have demonstrated an association Background. The urea reduction ratio (URR), a measure between the amount of hemodialysis and mortality among quantitating solute removal during hemodialysis, is the fracpatients with end-stage renal disease (ESRD) [1-8]. In tional reduction of the blood urea concentration during a single general, patient mortality is higher when the hemodialyhemodialysis treatment. The URR is the principal measure of sis "dose" is low and is lower when the hemodialysis hemodialysis dose in the United States. Based on studies of patients dialyzed prior to 1994, a minimum URR value of 65% dose is high. The urea reduction ratio (URR) is a comwas recommended to optimize survival. Because of new hemomonly used measure of hemodialysis dose and is based dialysis technologies and evolving demographics of the hemoon the fractional reduction of blood urea nitrogen dialysis population, the relationship between the amount of (BUN) concentration during a single hemodialysis treathemodialysis and mortality was examined in contemporary cohorts. ment. It is calculated by dividing the fall of BUN (predial-Methods. This retrospective cohort included Ͼ15,000 patients ysis minus postdialysis BUN) by the predialysis and is per year receiving hemodialysis during 1994 through 1997. Each expressed as a percentage [1, 9-11]. Retrospective studpatient's URR was averaged for the three months prior to the ies of mortality for ESRD patients suggest that the odds beginning of each year. Mortality odds ratios were calculated of death increase progressively as the URR falls below for patients by URR. To determine the URR value above which no further improvement in mortality was seen ("thresh-60 to 65% [1-3]. Such findings and a professional consenold"), spline functions were tested in logistic regression models, sus have led three national organizations, including the both unadjusted and adjusted for case mix measures. The Health Care Financing Administration (HCFA), the strength of fit for URR, defined by a range of candidate threshprincipal payer of dialysis services, to advocate a URR olds from 55 to 75%, was evaluated in increments of 1% for of 65% as the threshold for "adequate" hemodialysis each year using spline functions. Results. The median URR was 63.2, 65.4, 67.4, and 68.1% and to profile providers accordingly [9, 10, 12-14]. for 1994 through 1997, respectively. The median length of hemo-Using URR as a measure of hemodialysis dose, substandialysis treatments increased only six minutes from the begintial improvement in the amount of hemodialysis has ocning to the end of the period of analysis. Using spline functions, curred in recent years. As reported in the 1998 National the threshold URR values were 61.1, 65.0, 68.0, and 71.0% for 1994 through 1997 in models adjusted for case mix. The ratio ESRD Core Indicators Report, a profile of patientof median URR to URR threshold decreased from 1.03 in specific dialysis practices, the mean URR for the United 1994 to 0.97 in 1997. States increased from 62.7% (deemed inadequate) to Conclusions. From 1994 to 1997, the median URR and the 68.0% (deemed adequate) from 1993 to 1997 [14]. Al-URR threshold for mortality benefit increased. Although an though mean URR provides a gauge of hemodialysis increased need in the amount of hemodialysis may be a consequence of changes in patients' demographic characteristics, the adequacy at the population level, the proportion of palikely explanation(s) is a change in the dialysis procedure and/or tients whose URR is Ն65% is of greater significance for blood sampling favoring higher URR values without changing optimizing individual patient survival. Using this clinical the amount of dialysis provided. The recommended minimum performance benchmark, the percentage of patients re-URR of 65% appears to be too low to confer an optimal mortality benefit in the context of current practices. ceiving a URR Ն65% rose from 43% in 1993 to 72% in 1997 [14]. The aforementioned outcome studies characterizing

Kidney International, 1999

Background. The current standard of adequacy for peritoor more [3]. Recognized weaknesses of thes... more Background. The current standard of adequacy for peritoor more [3]. Recognized weaknesses of these guidelines neal dialysis (PD) is to provide a weekly normalized urea include: (a) the lack of evidence to establish equivalency clearance (Kt/V) of 2.0 or more and a creatinine clearance between the renal and peritoneal contributions of small (C Cr) of 60 liter/1.73 m 2 or more. As native renal function is lost, it is important to determine the effectiveness of the available solute removal to clinical outcome; and (b) the uncertain therapeutic modalities in achieving these goals. value of urea versus creatinine as markers of uremia Methods. A model to assess our ability to provide a weekly [4, 5]. Most clinical outcome studies have included pa-Kt/V urea of 2.0 or more and a C Cr of 60 liter/1.73 m 2 or more tients at various stages of uremia therapy with different to anuric patients undergoing continuous ambulatory PD degrees of residual renal function (RRF). The most (CAPD) and automated PD (PD Plus) was developed. The body surface area (BSA) distribution was obtained from 38,768 quoted prospective study, CANUSA, was not designed patients undergoing dialysis during January 1997. The distributo evaluate the relative effect of RRF and the peritoneal tion of peritoneal transport rates (PTRs) was obtained from clearance on patient outcome [2, 6, 7]. Furthermore, 2531 peritoneal equilibration tests performed during 1996. The patients with significant RRF are more likely to achieve weekly K p t/V urea was calculated for the various PTR groups higher clearances of creatinine than anuric patients beand the range of BSA with four PD prescriptions: CAPD 8 liters, CAPD 10 liters, PD Plus 12 liters, and PD Plus 15 liters, cause of tubular secretion of creatinine in advanced renal using a previously validated kinetic program (PackPD). failure. Few prospective studies have been performed in Results. The predicted percentage of patients capable of anuric patients undergoing PD. However, Selgas et al achieving the adequacy goals for Kt/V and C Cr , respectively, studied patients who had a minimum of three years on

Kidney International, 2000

Prealbumin is as important as albumin in the nutritional assesstive risk (RR) of death between tw... more Prealbumin is as important as albumin in the nutritional assesstive risk (RR) of death between two-and tenfold or ment of hemodialysis patients. more, depending on the severity of PEM and the interac-Background. Although serum prealbumin is considered a tion of PEM with other factors (age, anemia, dialysis valid indicator of nutritional status in hemodialysis patients, vintage, etc.) [1-3]. Serum albumin has been the most there is relatively little evidence that its determination is of commonly employed marker of PEM, based largely on major prognostic significance. In this study, we aimed to determine the independent association of serum prealbumin with the statistical association between diminished serum alsurvival in hemodialysis patients, after adjusting for serum albumin, mortality, and morbidity. Indeed, the Health Care bumin and other indicators of protein energy nutritional status. Financing Administration's (HCFA) core clinical indica-Methods. Serum prealbumin was measured in more than 1600 tors for adequate dialysis treatment involves an audit of maintenance hemodialysis patients. We determined the correlations among prealbumin and other indicators of nutritional serum albumin concentration in dialysis facilities [4] costatus, including serum albumin, and bioimpedance-derived indiequal with anemia and measures reflecting the adequacy cators of body composition. The relationship between serum of dialysis. prealbumin and survival was determined using proportional The clinical approach to the patient with PEM within hazards regression.

Kidney International, 1999

malnutrition (manifest by a reduced V) that overcomes what-Exploring the reverse J-shaped curve b... more malnutrition (manifest by a reduced V) that overcomes what-Exploring the reverse J-shaped curve between urea reduction ever benefit might be derived from an associated increase in urea ratio and mortality. clearance. Identification of patients who achieve extremely Background. Although accepted worldwide as valid meahigh URR (Ͼ75%) or single-pooled Kt/V (Ͼ1.6) values using sures of dialysis adequacy, neither the Kt/V (urea clearance standard dialysis prescriptions should prompt a careful assessdetermined by kinetic modeling) nor the urea reduction ratio (URR) have unambiguously predicted survival in hemodialysis ment of nutritional status. Confounding by protein-calorie malpatients. Because the ratio Kt/V can be high with either high nutrition may limit the utility of URR or Kt/V as a population-Kt (clearance ϫ time) or low V (urea volume of distribution) based measure of dialysis dose. and V may be a proxy for skeletal muscle mass and nutritional health, we hypothesized that the increase in the relative risk of death observed among individuals dialyzed in the top 10 to Defining the dose of dialysis to optimize health and 20% of URR or Kt/V values might reflect a competing risk of malnutrition. functional status for patients with end-stage renal disease Methods. A total of 3,009 patients who underwent bioelec-(ESRD) has been among the principle goals of the netrical impedance analysis were stratified into quintiles of URR. phrology community for more than two decades. In a Laboratory indicators of nutritional status and two bioimpelandmark publication in 1985, Gotch and Sargent revoludance-derived parameters, phase angle and estimated total tionized the field of dialysis therapy with their reanalysis body water, were compared across quintiles. The relationship between dialysis dose and mortality was explored, with a focus of data from the National Cooperative Dialysis Study on how V influenced the structure of the dose-mortality rela-(NCDS) [1, 2], redefining the dose of dialysis in terms tionship. of the clearance (K) ϫ time (t) product normalized to Results. There were statistically significant differences in all the urea volume of distribution (V), a dimensionless nutritional parameters across quintiles of URR or Kt/V, indiquantity derived from the fractional reduction in urea cating that patients in the fifth quintile (mean URR, 74.4 Ϯ 3.1%) were more severely malnourished on average than panitrogen (URR) achieved over the course of a hemodialtients in all or some of the other quintiles. The relationship ysis treatment. These authors and others who followed between URR and mortality was decidedly curvilinear, resemthem were successful at identifying a minimum dose of bling a reverse J shape that was confirmed by statistical analysis. dialysis above which death and complications were less An adjustment for the influence of V on URR or Kt/V was performed by evaluating the Kt-mortality relationship. There likely, although the optimal dose of dialysis remained was no evidence of an increase in the relative risk of death unknown. Application of this methodology to other paamong patients treated with high Kt. Higher Kt was associated tient cohorts similarly showed that an increase in URR with a better nutritional status. or Kt/V was generally associated with reduced mortality Conclusion. We conclude that the increase in mortality oband less frequent complications [3-8]. Although it folserved among those patients whose URR or Kt/V are among the top 10 to 20% of patients reflects a deleterious effect of lowed that further increases in dialysis dose might be expected to lead to fewer complications and better outcomes, Owen et al initially showed that there appeared

Kidney International, 1999

ity in those patients who are in the lower 50th percentile for Body weight-for-height relationshi... more ity in those patients who are in the lower 50th percentile for Body weight-for-height relationships predict mortality in mainthis measurement. tenance hemodialysis patients. Background. Protein-energy malnutrition is a strong predictor of mortality in maintenance hemodialysis (MHD) patients. This association has generally been described for serum Many epidemiological studies have shown a correlachemistry measures of protein-energy malnutrition. We hytion between parameters of protein-energy nutritional pothesized that body weight-for-height relationships also predict survival in MHD patients. status and morbidity or mortality in maintenance dialysis Methods. During the last three months of 1993, data were patients. These observations, in general, have demonobtained on 12,965 men and women concerning clinical characstrated relationships between serum chemistry measures, teristics (height, postdialysis weight, age, gender, race, and dietary protein intake as indicated by the protein equivapresence or absence of diabetes mellitus) and laboratory mealent of protein nitrogen appearance (PNA, also called surements (predialysis serum albumin, creatinine and cholesterol, and the urea reduction ratio). Patient survival during the PCR), postdialysis body weight, and clinical outcome next 12 months was evaluated retrospectively. [1-6]. In maintenance hemodialysis (MHD) patients, the Results. In comparison to values for normal Americans deserum chemistry measurements which reflect nutritional termined from the National Health and Nutrition Evaluation status or nutrient intake and which correlate with mor-Survey II data, weight-for-height relationships tended to be bidity or mortality include predialysis serum albumin, slightly lower than normal in African American men and women and Caucasian men undergoing MHD and were normal creatinine, urea, cholesterol, potassium, and phosphorus or slightly greater in the taller Caucasian women. In both men [1, 2, 4, 5]. and women, the mortality rate decreased progressively as the These relationships may be particularly important for patients' weight-for-height increased. MHD patients who the following reasons: (a) Some of these nutritional paweighed more than normal had the lowest mortality rates. rameters, such as the serum albumin, strongly predict After adjustment for clinical characteristics and laboratory measurements, the inverse relationship between mortality rates outcome [2, 4, 5]. (b) Many MHD patients have evidence and weight-for-height percentiles was still highly significant for of protein-energy malnutrition [1-7]. (c) The mortality patients within the lower 50th percentile of body weight-forrate of MHD patients is very high [2, 4, 5, 8], averaging height. Serum albumin correlated directly with weight-for-height over 20% in the United States. (d) Certain nutritional in patients in the lower 50th percentile of weight-for-height. parameters are potentially modifiable, and there is a Serum creatinine and cholesterol correlated directly with weight-for-height in the entire population of men and women. possibility that altering these nutritional parameters may In contrast, the urea reduction ratio was inversely correlated improve outcome; this possibility has not yet been tested with weight-for-height. in randomized prospective clinical trials. Conclusions. These data indicate that weight-for-height is a Because body weight-for-height relationships predict strong predictor of 12-month mortality in male and female clinical outcome in healthy men and women [9-11], we MHD patients. Multivariate analyses indicate that body weight-for-height is an independent predictor of higher mortalassessed the relationship of weight-for-height to mortality in a retrospective analysis of data obtained from 12,965 MHD patients. Four primary questions were ad-Key words: protein-energy malnutrition, survival rate for hemodialysis, dressed: (a) How does the distribution of weight-fordialysis, albumin, creatinine, cholesterol, mortality prediction, maintenance hemodialysis. height and body mass index (BMI) in MHD patients compare with the normal population in the United

Kidney International, 1999

not) and also with increasing body size (whether adjusted for The urea {clearance ϫ dialysis time... more not) and also with increasing body size (whether adjusted for The urea {clearance ϫ dialysis time} product (Kt) as an outcome-Kt or not) for each estimate of size. Significant statistical interbased measure of hemodialysis dose. actions of Kt with gender, but not Kt with race, were observed Background. The normalized treatment ratio [Kt/V ϭ the ratio of the urea {clearance ϫ time} product to total body in all models. There were no statistical interactions, suggesting water] and the urea reduction ratio (URR) have become widely that higher Kt was routinely required with increasing body accepted measures of dialysis dose. Both are related to and size. Separate risk profiles for males and females suggested a derived from pharmacokinetic models of blood urea concentrahigher Kt threshold for males. tion during the dialysis cycle. Theoretical reconsideration of Conclusions. The urea {clearance ϫ time} is a valid outcomethe models revealed that the premise about V on which they based measure of dialysis dose and is not confounded by inrest (that is, that V is a passive diluent with no survival-associdexing it to an estimate of body size, which has outcomeated properties of its own) is flawed if the intended use of the associated properties of its own. Dialysis prescriptions for models is for profiling clinical outcome (for example, mortality) males and females should be regarded separately, but there rather than estimating urea concentration. As a proxy for body appears no need to make a distinction between the races. mass, V has survival-associated properties of its own. Thus, indexing {clearance ϫ time} to body size could create an offsetting combination whereby one measure favorably associated with survival (Kt) is divided by another (for example, V). The primary finding [1] and final report [2] of the Observed clinical paradoxes support that interpretation. For National Cooperative Dialysis Study (NCDS) were pubexample, patients with a low body mass have both higher URR lished more than 15 years ago. That study evaluated four and higher mortality than heavier patients. Increasing mortality groups of patients arranged in a 2 ϫ 2 factorial design is often observed at high URR, suggesting the possibility of "over-dialysis." Black patients tend to be treated at lower URR [3] where blood urea nitrogen (BUN) concentration, a than whites but enjoy better survival on dialysis. Therefore, proxy for "small molecule"-directed therapy, and the {clearance ϫ time} was evaluated as an outcome-based measure length of the dialysis treatment, a proxy for therapy of dialysis dose, not indexed to V, and various body size estidirected at larger molecules, were the factors [3, 4]. A mates were evaluated as separate and distinct measures. Methods. The retrospective sample included 17,141 black computer-assisted, single-pool, variable-volume urea kiand white hemodialysis patients treated three times per week. netic model [5-7] was used to facilitate the control of Logistic regression analysis was used to evaluate death odds in BUN at two levels (low and high) within two levels of age-, gender-, race-, and diabetes-adjusted models. Kt and five dialysis time (long and short). The urea kinetic equations body size estimates (total body water or V, body weight, body weight adjusted for height, body surface area, and body mass

Kidney International, 2001

It was stronger for the albumin Indicator (R 2 ϭ 11.6, 20.4, Association between prevalent care p... more It was stronger for the albumin Indicator (R 2 ϭ 11.6, 20.4, Association between prevalent care process measures and facil-21.8). The fractions of patients falling outside of the Indicator ity-specific mortality rates. guidelines tended to be higher in the highest SMR group. The Background. Medical communities often develop practice groups were not well separated, however, particularly for the guidelines recommending certain care processes intended to hematocrit and URR Indicators, and there was substantial promote better clinical outcome among patients. Conformance overlap between them. Finally, although the likelihood that a with those guidelines by facilities is then monitored to evaluate facility would be a member of the high or low SMR group was care quality, presuming that the process is associated with and associated with fractional variance from Core Indicator guidecan be used reliably to predict clinical outcome. Outcome is lines, the strengths of association were weak, and the probabiloften monitored as a facility-specific mortality rate (SMR) stanity that a facility would be a member of the high or low group dardized to the mix of patients treated, also presuming that could not be easily distinguished from the probability that it inferior outcome implies a suboptimal process. The U.S. Health would be a member of the middle group. Care Financing Administration monitors three practice guide-Conclusions. While there were statistical associations belines, called Core Indicators, in dialysis facilities to assist mantween SMR and the fraction of patients in facilities who were agement of its end-stage renal disease program: (1) patients' at variance with these guidelines, they were weak and variances hematocrit values should exceed 30 vol%, (2) the urea reducfrom the guidelines could not be used reliably to predict high or tion ratio (URR) during dialysis should equal or exceed 65%, low SMR. Such findings do not imply that measures reflecting and (3) patients' serum albumin concentrations should equal anemia, dialysis dose, or medical processes that influence serum or exceed 3.5 g/dL. albumin concentration are irrelevant to the quality of care. They Methods. The associations of a facility-specific SMR were evaldo suggest, however, that more attention needs be paid to these uated with the fractions of hemodialysis patients not conformand other associates and causes of mortality among dialysis ing to (that is, at variance with) the Core Indicators during three patients when developing care process indicator guidelines. successive years (1993 to 1995) in large numbers of facilities (394, 450, and 498) using one-variable and multivariable statistical models. Three related strategies were used. First, the association of the SMR with the fraction of patients not meeting Process measures of care, such as the Health Care Fithe guideline was evaluated. Second, each facility was classified nancing Administration's (HCFA) Core Indicators, are by whether its SMR exceeded the 80% confidence interval selected measures thought to affect the outcome of care above 1.0 (worse than 1.0, Group 3), was less than the inter-[1-3]. They are sometimes monitored within preselected val below 1.0 (better than 1.0, Group 1), or was within the interval (Group 2). The fraction of those patients who did not limits hoping to influence the quality and outcome of care meet the Indicator guidelines was then evaluated in each group. [4]. The Core Indicators were used for that purpose. Thus, Third, the ability of variance from Indicator guidelines to prethey should be strongly associated with and predict clinidict into which of the three SMR groups a facility would be cal outcome. That is particularly true if they are the main categorized was evaluated. or exclusive measures used to judge the care quality among Results. SMR was directly correlated with variance from the Indicator guidelines, but the strengths of the associations were facilities. It is necessary, therefore, to understand and demweak particularly for the hematocrit (R 2 ϭ 2.2%, 5.6, and 2.2 onstrate the nature and strength of association between for each of the 3 years) and URR Indicators (R 2 ϭ 2.6, 0.6, 3.3). outcome and those measures selected for monitoring in order to ensure that actions taken pursuant to monitoring will likely be effective and efficient in achieving the

Kidney International, 2001

Cardiac arrest and sudden death in dialysis units. mon cause of death, with nearly half of deaths... more Cardiac arrest and sudden death in dialysis units. mon cause of death, with nearly half of deaths attributed Background. For patients with end-stage renal disease and to myocardial infarction, cardiac arrest, or other cardiac their providers, dialysis unit-based cardiac arrest is the most causes [2]. Numerous studies have looked at predictors feared complication of hemodialysis. However, relatively little of cardiac mortality in maintenance hemodialysis pais known regarding its frequency or epidemiology, or whether tients. A history of ischemic heart disease, congestive a fraction of these events could be prevented. Methods. To explore clinical correlates of dialysis unit-based heart failure, left ventricular hypertrophy (LVH) and dyscardiac arrest, 400 reported arrests over a nine-month period function, peripheral vascular disease, hypertension or hyfrom October 1998 through June 1999 were reviewed in detail. potension, hypoalbuminemia, dyslipidemia, anemia, and Clinical characteristics of patients who suffered cardiac arrest inflammation (elevated C-reactive protein) have all been were compared with a nationally representative cohort of associated with increased cardiovascular mortality [3-9]. Ͼ77,000 hemodialysis patients dialyzed at Fresenius Medical Care North America-affiliated facilities. Using data from the U.S. Renal Data System Case Results. The cardiac arrest rate was 400 out of 5,744,708, Mix Adequacy Study, Bleyer, Russell, and Satko docucorresponding to a rate of 7 per 100,000 hemodialysis sessions. mented a significantly higher proportion of sudden Cardiac arrest was more frequent during Monday dialysis sesdeaths in hemodialysis patients on Monday and Tuesday sions than on other days of the week. Case patients were nearly twice as likely to have been dialyzed against a 0 or 1.0 mEq/L compared with other days of the week [10]. No other potassium dialysate on the day of cardiac arrest (17.1 vs. 8.8%). risk factors for sudden death were identified. The authors Patients who suffered a cardiac arrest were on average older postulated that the accumulation of electrolytes and flu-(66.3 Ϯ 12.9 vs. 60.2 Ϯ 15.4 years), more likely to have diabetes ids between dialysis sessions increased the susceptibility (61.8 vs. 46.8%), and more likely to use a catheter for vascular to arrhythmias, especially after a weekend respite from access (34.1 vs. 27.8%) than the general hemodialysis population. Sixteen percent of patients experienced a drop in systolic dialysis. No data were available on the timing of sudden pressure of 30 mm Hg or more prior to the arrest. Thirtydeath relative to the hemodialysis session. Presumably, seven percent of patients who suffered cardiac arrest had been the majority of sudden death cases occurred at home hospitalized within the past 30 days. Sixty percent of patients rather than during dialysis. died within 48 hours of the arrest, including 13% while in the dialysis unit. Relatively few studies have examined the epidemiol-Conclusions. Cardiac arrest is a relatively infrequent but ogy of in-dialysis unit cardiac arrest. Moreover, no prior devastating complication of hemodialysis. To reduce the risk investigation has had sufficient power to estimate its inof adverse cardiac events on hemodialysis, the dialysate precidence accurately or to identify patient-and hemodialyscription should be evaluated and modified on an ongoing sis-specific factors that might be associated with a higher basis, especially following hospitalization in high-risk patients. than baseline risk. We therefore undertook this study of cardiac arrest during hemodialysis using data obtained directly from dialysis units at the time of the event. We Hemodialysis patients have an age-adjusted mortality were most interested in identifying modifiable risk facrate that is 3.5 times higher than that of the general population [1]. Cardiovascular disease remains the most comtors so that some future events might be prevented.

Journal of Renal Nutrition, 1997

To determine the relation between phase angle by bioelectrical impedance analysis (BIA) and survi... more To determine the relation between phase angle by bioelectrical impedance analysis (BIA) and survival in hemodialysis patients. Design: Cohort analytic study. Setting: One hundred one free-standing outpatient dialysis units. Patients: Three thousand nine adult patients on thrice weekly hemodialysis. Patients with amputations above the transmetatarsal site were excluded from participation. Main Outcome Measure: Vital status, with follow-up to at least 1 year. Results: Mean phase angle was 4.8 2: 1.8 degrees. Patients with narrow (low) phase angle experienced an increased relative risk (RR) of death «3 degrees; RR 4.3; 95% confidence interval [CI], 2.9-6.2; and 3 to 4 degrees); RR 2.2; 95% CI, 1.6-3.2; compared with the 2::6 degrees reference). There were no significant differences in risk among patients with phase angle 4 to 5 degrees (RR 1.2; 95% CI, 0.8-1.8), 5 to 6 degrees (RR 1.1; 95% CI, 0.7-1.7), and 2::6 degrees, suggesting a nonlinear relation between phase angle and survival. The RRs for phase angle <4 degrees remained statistically significant after adjusting for age, gender, race, serum albumin and creatinine concentrations, and dialysis intensity «3 degrees, RR 2.2; 95% CI, 1.6-3.1, and 3 to 4 degrees, RR 1.3; 95% CI, 1.0-1.7, compared with all patients 2::4 degrees). Conclusions: In patients on hemodialysis, BIA-derived phase angle <4 degrees was associated with an increased RR of death, even after adjustment for case mix and several nutritional indicators. Further research is required to determine whether BIA can be used to monitor health status over time, or to gauge response to nutrition support or other clinical interventions in patients with end-stage renal disease.

Journal of Proteome Research, 2005

In this paper, we report an experimental setup and mathematical algorithm for determination of re... more In this paper, we report an experimental setup and mathematical algorithm for determination of relative protein abundance from directly labeled native protein samples applied to an array of antibodies. The application of the proposed experimental system compensates internally at each array element for a number of deficiencies in array experiments such as differential labeling efficiency in dual color assay systems, differential solubility of protein molecules in dual color assay systems, and differential affinity of capture reagents toward proteins labeled with two different fluorescent dyes. This system offers full compensation for variable amounts of capture reagents on separate array structures, as well as limited compensation for nonspecific interactions between capture reagents and analytes. The proposed experimental strategy enables the use of a large number of capture reagents to develop a true multiplex analysis system that will yield complete relative protein abundance information in two biological systems.

JAMA: The Journal of the American Medical Association, 1995

Hemodialysis International, 2003

Background: Consensus processes using the clinical literature as the primary source for informati... more Background: Consensus processes using the clinical literature as the primary source for information generally drive projects to draft clinical practice guidelines (CPGs). Most such literature citations describe special projects that are not part of an organized quality management initiative, and the publication/review/consensus process tends to be long. This project describes an initiative to develop and explore a flexible and dedicated data-driven paradigm for deciding new CPGs that could be rapidly responsive to changing medical knowledge and practice. Methods: Candidate Clinical Practice Monitoring Measures (CPMM) were selected using a large, national database according to the natures and strengths of their associations with mortality risk among patients during 1994. Thresholds above or below which risk of death increased were evaluated for each CPMM using risk profile charts and spline functions. The fractions of patients outside of those thresholds in each dialysis unit (the %Var) were determined for the years 1993, 1994, and 1995. A standardized mortality ratio (SMR) was also determined for each year for each facility. The associations between the %Var and SMR were evaluated in several single-variable and multivariable statistical models. Results: Eleven CPMM were selected and evaluated based on their associations with death risk. These included the urea clearance x dialysis time product (Kt); the concentrations of albumin, potassium, phosphate, bicarbonate, hemoglobin, neutrophils, and lymphocytes in the blood; the body weight/height ratio; diastolic blood pressure; and vascular access type. Even though the CPMM were strongly associated with death risk among patients, the %Var were weakly and inconsistently associated with SMR among facilities. Conclusions: The paradigm was flexible, easy to implement, quickly executed, and potentially able to accommodate evolving medical practice assuming the availability of large database systems such as this. The primary associates of death risk were easily identified and the thresholds easily adopted. The SMR and %Var from the CPMM were only weakly associated, however, suggesting that one cannot be reliably predicted from the other. As such, quality management programs should likely monitor both the processes and outcomes of care among dialysis facilities.

Archives of Internal Medicine, 1991

The annual mortality rate among patients receiving long-term hemodialysis has been rising over th... more The annual mortality rate among patients receiving long-term hemodialysis has been rising over the past decade. The prevalences of known risk factors such as older age, male sex, duration of dialysis, presence of diabetes, coronary artery disease, or hypertension do not seem to have changed during this time. However, evidence suggests that an increased body aluminum level may have an adverse effect on survival even in the absence of overt aluminum toxic reaction. Therefore, we evaluated the correlation between serum aluminum levels and mortality in 10 646 patients undergoing long-term hemodialysis. Mortalities were 18% higher for patients with serum aluminum levels between 1520 and 2220 nmol/L and progressively increased to 60% higher for patients with aluminum levels above 7410 nmol/L. Serum aluminum level was an important predictor of survival even after other known risk factors had been controlled. These data strongly suggest that patients undergoing long-term hemodialysis should have periodic surveillance of the serum aluminum levels, and in those patients who have plasma levels of 1520 to 2220 nmol/L or higher, one should seriously consider discontinuing aluminum salts and giving therapy to decrease body aluminum level if it is found to be increased.