Introduction

Postdiarrheal (Shiga-toxin mediated) hemolytic uremic syndrome (HUS) is the most frequent cause of acute renal failure in infants and young children [1]. Some patients who survive diarrheal HUS (D+HUS) never recover renal function, and others experience secondary decline in renal function years after apparent recovery [2–7]. Although the presence and duration of oliguria and anuria during the acute phase of HUS have been shown to correlate with long-term outcome, there is still a need for noninvasive ways to quantify the risk at time of hospital discharge. Being able to predict future renal sequelae at the time of hospital discharge would allow parents, patients, and doctors the opportunity to plan and prepare for future monitoring and treatment.

Methods

We utilized our computerized HUS registry to identify patients with post-D+HUS from 1970 to 2003. A case of classic D+HUS was defined as a child younger than 18 years of age with a prodrome of gastroenteritis (usually with hemorrhagic colitis) followed by microangiopathic hemolytic anemia, thrombocytopenia, and acute nephropathy [2]. For this study, oliguria was defined as urine output <240 ml/m2 per day. Thus patients that were anuric were also included in the oliguria patient cohort. Anuria was defined as urine output <15 ml/day. Cases were not excluded because of a normal platelet count (i.e. > 150 × 109/L) because thrombocytopenia has not been documented during hospitalization in all cases of D+HUS. Thirteen patients who displayed end-stage renal disease (ESRD) at time of discharge and those who died (17 patients) during the acute phase of the disease were excluded from analysis. Of the 357 patients identified, 159 (77 males, 82 females) had at least 1 year of follow-up data; 90 of these had at least 1 day of oliguria. Sixty-nine patients, a subset of those with at least 1 day of oliguria, also had at least 1 day of anuria.

Patient information from our follow-up database was included only if it had been collected at least 1 year following hospital discharge. Individuals were separated into experimental groups based on absence, presence, and duration of oliguria and anuria, and the percentage of those with renal-related sequelae within these groups was determined. Sequelae of interest included proteinuria, glomerular filtration rate (GFR), a combination of proteinuria and low GFR, and hypertension. Proteinuria was defined as ≥1+ by dipstick analysis [8–10]. GFR was calculated by the Schwartz formula, with <90 ml/min per 1.73 m2 being classified as an abnormal value [10–12]. Hypertension was defined as blood pressure >95th percentile for age and gender [9, 10, 13]. Urine measurements at follow-up were made using first-morning specimens.

Statistical analysis

Categorical variables were compared by Yates correct chi-square and Fischer exact tests. Continuous variables were compared by the Mann–Whitney tests. Analysis to compare those with follow-up information and those lost to follow-up was performed using the appropriate statistical tests for continuous and categorical variables. All statistical analysis was performed using SPSS 15.0 (SPSS, Inc., Chicago, IL, USA). To better define the contribution of the duration of oliguria and anuria to renal sequelae, binary logistic regression models were developed for each renal sequelae. Patients were divided into one of the following four cohorts: (1) no oliguria or anuria, (2) 1–5 days, (3) 6–10 days, (4) >10 days. The same acute-phase variables used in the ungrouped analysis were also entered into the logistic models. The models were adjusted to evaluate for differences in ascertainment or therapy by controlling for the decade of treatment and the length of follow-up. Two sided p values < 0.05 were considered significant. The reported odds ratios (ORs) and 95% confidence intervals (95% CI) were calculated in reference to patients with no oliguria or anuria unless otherwise noted.

Results

One hundred and fifty-nine cases that met the criteria for inclusion in this study were identified from the computerized registry. The mean interval from hospitalization until study enrollment was 8.75 (range 1–30) years. The mean duration of oliguria was 9.2 days [standard deviation (SD) 8.8 days; range 1–57 days], and for anuria 6.0 days (SD 4.9 days, range 1–24 days). There were no statistically significant differences for variables associated with hospital admission by decade of hospitalization (1970s, 1980s, 1990s, 2000s). No significant differences in acute-phase variables between those lost to follow-up and those included in analysis were detected. A moderately elevated incidence of sequelae was noted in patients with extended periods of follow-up. Follow-up duration was therefore entered in all multivariate models.

Table 1 provides demographic characteristics and selected clinical features of the study population during the acute phase of HUS. Statistically significant differences were noted in white blood cell count (WBC), hematocrit (HCT), blood urea nitrogen (BUN), serum creatinine concentration, and platelet counts in patients with no oliguria versus those with ≥ 1 day of oliguria. Statistically significant differences were also noted in the number of patients who received hemodialysis and those who received peritoneal dialysis. A relatively larger proportion of male patients suffered at least 1 day of oliguria (52.2%) in comparison with those with no oliguria (43.5%), though this difference failed to reach statistical significance. Table 2 shows the incidence of renal-related sequelae summarized by gender. Male patients exhibited an increased risk of low GFR in comparison with female patients. This difference persisted in multivariate models when controlling for age at admission and for length of follow-up.

Table 1 Demographic and clinical characteristics of patient population groups during acute-phase hospitalization
Full size table
Table 2 Incidence of renal-related sequelae summarized by gender
Full size table

Figure 1 shows the incidence of long-term renal-related sequelae and the relationship to oliguria duration. Figure 2 shows the incidence of long-term renal-related sequelae and the relationship to anuria duration. Patients with longer oliguria duration (> 10 days) exhibited higher incidence of all renal-related sequelae. The same trend was noted among patients with more than 5 days of anuria, though the overall incidence of sequelae was higher than that seen in patients with oliguria.

Fig. 1
figure 1

Incidence of long-term renal-related sequelae versus days of oliguria. A large increase in the incidence of all sequelae occurs at > 10 days of oliguria

Full size image
Fig. 2
figure 2

Incidence of long-term renal-related sequelae versus days of anuria. Patients with longer durations of anuria exhibit higher incidence of renal-related sequelae

Full size image

One or more sequelae (e.g. proteinuria, low GFR, hypertension) were seen in 25 (36.2%) of those who had no recorded oliguria and 34 (37.8%) of those with no recorded anuria. Tables 3 and 4 provide ORs of renal-related sequelae and their relationship to oliguria (Table 3) and anuria (Table 4) duration.

Table 3 Relationship between oliguria duration (in days) and the risk of long-term renal-related sequelae
Full size table
Table 4 Relationship between anuria duration (in days) and the risk of long-term renal-related sequelae
Full size table

In the oliguric group, the incidence of patients with one or more sequelae remained fairly constant in patients with 1–5 days (39.5%; OR 1.2, 95% CI 0.5–2.6) and 6–10 days (36.4%; OR 1.0, 95% CI 0.4–2.7). A dramatic increase in the incidence of one or more sequelae to 88.7% (OR 11.4, 95% CI 3.6–36.5) occurred among patients with > 10 days of oliguria, however. Oliguria duration of > 10 days was also found to be an independent predictor of all renal-related sequelae (p < 0.001). These findings persisted when controlling for age at admission, gender, and length of follow-up in multivariate logistic regression models.

Among patients with anuria (Table 4), there was a step-wise increase in the incidence of all sequelae as the duration of anuria increased. This was most notable once anuria exceeded 5 days in duration, at which time the incidence of all sequelae, with the exception of hypertension, increased substantially. In one example, the incidence of at least one sequel in those with a history of 1–5 days of anuria rose from 42.1% to 68.2% if the anuria persisted for 6–10 days. The incidence of proteinuria likewise increased from 13.2% to 31.8%, low GFR from 26.3% to 50.0%, and the combination of proteinuria and low GFR from 7.9% to 44.4% (see Table 4 for ORs and 95% CIs). Increase in the incidence of low GFR and proteinuria in patients with >10 days of anuria was associated with an OR of 9.8 (95% CI 1.7–57.5); it rose to 57.8 (95% CI 141–544) after controlling for age at hospital admission, decade of treatment, gender, and length of follow-up.

Hypertension, the incidence of which was similar in those with a history of 1–5 days (15.8%) and 6–10 days (13.6%) of anuria, increased markedly to 55.6% if anuria persisted for >10 days. The adjusted ratio for hypertension in patients with >10 days of anuria was 13.1 (95% CI 2.8–60.7) when controlling for age at admission, decade of treatment, gender, and length of follow-up.

Results of multivariate logistic regression analysis

The most significant predictor of sequelae (anuria duration) was entered into the multivariate regression equation first and its effect controlled in comparison with other clinical variables from the acute phase of the illness. Variables from the acute phase included WBC, HCT, creatinine, and BUN at admission. After the effect of anuria was controlled for, none of the other variables achieved statistical significance. This finding may reflect the fact that the other variables have a high degree of correlation with one another. Similar results were seen when oliguria was entered into the equation first and its effect was controlled for.

Discussion

Although the majority of studies [2, 6, 9, 14–18], including our own [2, 10], conclude that oliguria and anuria duration are major prognostic indicators, there have been few attempts to specifically quantify this relationship. It should be noted that larger studies with longer periods of follow-up [9, 18, 19] are generally necessary to obtain statistically significant information. Studies with smaller numbers of patients [3, 16, 20] and/or shorter periods of follow-up have often resulted in conclusions markedly different from those with larger sample sizes. This study describes a relatively large number of patients (159) with a mean follow-up that is substantial for such a large cohort (8.75 years). Another long-term follow-up study, but with a markedly smaller sample size, illustrates the importance of long-term follow-up to determine the true prevalence of renal-related sequelae [3].

Anuria served as a better predictor for most chronic renal sequelae than did oliguria. Furthermore, longer duration of either oliguria (> 10 days) or anuria (> 5 days) correlated with higher incidence in all sequelae measured. Of particular importance were the dramatic increases seen in the prevalence of proteinuria, low GFR, and low GFR combined with proteinuria once anuria or oliguria exceeded 10 days in duration. Such a strongly correlated relationship provides evidence of a link between the severity of initial renal injury and eventual outcome.

It is noteworthy that a substantial number (approximately 36%) of children with no recorded oliguria or anuria were left with sequelae. Moreover, proteinuria, a recognized sign of hyperfiltration injury, is seen in about 10% of those with no recorded oliguria or anuria. Although these abnormalities have so far remained mild, there is concern about the adverse effects of pregnancy and normal glomerular obsolescence as these survivors move through adulthood.

The tabulated models presented here may serve as quantitative predictors of long-term renal-related sequelae and may help identify patients who, at the time of hospital discharge, require extended follow-up. The predictive value for long-term sequelae was particularly strong in individuals who experienced anuria. This general trend provides evidence of the predictive value of oliguria or anuria duration and is one of the few noninvasive quantitative predictors available.

Based on our collective experience, we suggest that all patients be evaluated yearly during the first decade following D+HUS then, if normal, every 2 years during the second decade, and if normal then, every 5 years for life. These evaluations should include a measurement of GFR, a calculation of albumin/creatinine ratio using a first morning urine specimen, and careful blood pressure measurements. Those with signs of hyperfiltration injury (microalbuminuria and/or overt proteinuria) need to be evaluated more often during pregnancies and as they progress through the midlife years, a time when normal glomerular obsolescence might further stress the remaining nephron population.

This admittedly is a cautious approach but one we feel is reasonable until we have had the opportunity to track a sizable cohort throughout their lives. Of major concern is the long-term outlook for those with chronic kidney disease, including reduced GFR and persistent proteinuria. This group is at high risk for eventual advancement through the stages of chronic kidney disease and therefore should be evaluated at more frequent intervals. Based on the findings of this study, patients with > 10 days of oliguria or > 5 days of anuria warrant more careful monitoring.