Vous êtes sur la page 1sur 10

Association of Procalcitonin With Acute Pyelonephritis and Renal Scars in Pediatric UTI

WHAT S KNOWN ON THIS SUBJECT: Prompt, high-quality diagnosis of acute pyelonephritis and later identication of children with scarring are important to prevent future complications. Examination by dimercaptosuccinic acid scan is the current clinical gold standard but is not routinely performed. WHAT THIS STUDY ADDS: Procalcitonin demonstrated a more robust predictive ability, compared with C-reactive protein or white blood cell count, to selectively identify both children who had acute pyelonephritis during the early stage of urinary tract infections, as well as those with late scarring.
AUTHORS: Sandrine Leroy, MD, PhD,a,b Anna FernandezLopez, MD,c Roya Nikfar, MD,d Carla Romanello, MD,e Franois Bouissou, MD,f Alain Gervaix, MD,g Metin K. Gurgoze, MD,h Silvia Bressan, MD,i Vladislav Smolkin, MD,j David Tuerlinckx, MD,k Constantinos J. Stefanidis, MD,l Georgos Vaos, MD,m Pierre Leblond, MD,n Firat Gungor, MD,o Dominique Gendrel, MD,p and Martin Chalumeau, MD, PhDp,q
aCentre for Statistics in Medicine, University of Oxford, Oxford, United Kingdom; bEpidemiology of Emerging Disease Unit, Institut Pasteur, Paris, France; cDepartment of Pediatrics, Joan de Deu Hospital, Barcelona, Spain; dDepartment of Pediatrics, Abuzar Children Medical Center Hospital, Ahvaz, Iran; eDepartment of Pediatrics, University of Udine, Udine, Italy; fDepartment of Pediatrics, Childrens Hospital, Paul Sabathier University, CHU Purpan, Toulouse, France; gDepartment of Pediatrics, University Hospital of Geneva, Geneva, Switzerland; hFirat University Medicine Faculty, Department of Pediatrics, Division of Pediatric , Turkey; iDepartment of Pediatrics, University Nephrology, Elaz g of Padova, Padova, Italy; jDepartment of Pediatrics, Medical Center, Afula, Israel; kDepartment of Pediatrics, Cliniques Universitaires de Mont-Godinne, Universit Catholique de Louvain, Yvoir, Belgium; lDepartment of Pediatrics, P. and A. Kyriakou Childrens Hospital, Athens, Greece; mDepartment of Pediatric Surgery, Alexandroupolis University Hospital, Democritus University of Thrace School of Medicine, Thrace, Greece; nDepartment of Pediatrics, Jeanne de Flandre Hospital, Lille, France; oAkdeniz University, School of Medicine, Department of Nuclear Medicine, Antalya, Turkey; pDepartment of Pediatrics of Necker Hospital, Paris-Descartes University, Paris, France; and qInserm U953 Unit, Paris, France

abstract
BACKGROUND AND OBJECTIVE: Urinary tract infections (UTIs) are common childhood bacterial infections that may involve renal parenchymal infection (acute pyelonephritis [APN]) followed by late scarring. Prompt, high-quality diagnosis of APN and later identication of children with scarring are important for preventing future complications. Examination via dimercaptosuccinic acid scanning is the current clinical gold standard but is not routinely performed. A more accessible assay could therefore prove useful. Our goal was to study procalcitonin as a predictor for both APN and scarring in children with UTI. METHODS: A systematic review and meta-analysis of individual patient data were performed; all data were gathered from children with UTIs who had undergone both procalcitonin measurement and dimercaptosuccinic acid scanning. RESULTS: A total of 1011 patients (APN in 60.6%, late scarring in 25.7%) were included from 18 studies. Procalcitonin as a continuous, class, and binary variable was associated with APN and scarring (P , .001) and demonstrated a signicantly higher (P , .05) area under the receiver operating characteristic curve than either C-reactive protein or white blood cell count for both pathologies. Procalcitonin $0.5 ng/mL yielded an adjusted odds ratio of 7.9 (95% condence interval [CI]: 5.810.9) with 71% sensitivity (95% CI: 6774) and 72% specicity (95% CI: 6776) for APN. Procalcitonin $0.5 ng/mL was signicantly associated with late scarring (adjusted odds ratio: 3.4 [95% CI: 2.15.7]) with 79% sensitivity (95% CI: 7185) and 50% specicity (95% CI: 4554). CONCLUSIONS: Procalcitonin was a more robust predictor compared with C-reactive protein or white blood cell count for selectively identifying children who had APN during the early stages of UTI, as well as those with late scarring. Pediatrics 2013;131:870879

KEY WORDS acute pyelonephritis, children, procalcitonin, renal scarring, urinary tract infection ABBREVIATIONS APNacute pyelonephritis AUCarea under the curve CIcondence interval CRPC-reactive protein DCAdecision curve analysis DMSAdimercaptosuccinic acid LRlikelihood ratio ORodds ratio PCTprocalcitonin ROCreceiver operating characteristic UTIurinary tract infection VURvesicoureteral reux WBCwhite blood cell (Continued on last page)

870

LEROY et al

ARTICLE

Urinary tract infections (UTIs) are the most common invasive bacterial infections among young febrile children.1 UTIs can occur as simple bladder infections (lower UTI; bacteriuria only) but can also involve the kidneys (acute pyelonephritis [APN], in which bacteriuria is associated with infectious renal parenchymal involvement), leading to renal scarring.2 The belief that persisting APN effects followed by late renal scarring, sometimes with recurrences, may lead to future complications such as hypertension and/or end-stage renal failure has been the major driving force behind the aggressive investigation and treatment of rst-occurrence UTIs.3 The prompt and high-quality diagnosis of APN and differentiation from lower UTI is therefore of key importance. A dimercaptosuccinic acid (DMSA) scan is considered the gold standard in imaging for both renal parenchymal involvement during acute infection and for late renal damage left by the infection.4 However, DMSA scans are not performed in most children with UTI due to the limited availability of nuclear medicine departments compared with the high number of children with UTIs. Thus, a more practical and accessible tool that could assist clinicians in determining the presence of renal parenchymal involvement and/or late renal damage would be of great clinical value. Procalcitonin (PCT), a 116amino acid propeptide of calcitonin without hormonal activity, is an early, sensitive, and specic marker of bacterial infection.5,6 PCT is almost undetectable under physiologic conditions or during viral infections but rises in response to bacterial endotoxins; the extent of this increase seems to be proportional to the severity of the infection.6 However, its exact role, if any, in the inammatory response and in the cytokine cascade remains unknown.6
PEDIATRICS Volume 131, Number 5, May 2013

In febrile UTI, the predictive ability of high PCT concentrations for both APN and late renal scarring has been previously investigated by several teams. A review7 and a recent systematic review and meta-analysis8 showed that a serum PCT .0.5 ng/mL predicts early renal parenchymal involvement reasonably well (diagnostic odds ratio [OR]: 14.3 [95% condence interval (CI): 4.7 43.2]); however, heterogeneity made these results inconclusive.8 Moreover, results concerning late renal scarring were controversial, with no pooled measurements provided.7 Most of this heterogeneity and these discrepancies may be due to threshold effects because the initial studies chose different PCT cutoff values due to population variation; unfortunately, any effects from the latter could not be fully explored with only pooled data from the studies. Under these circumstances, the only way to analyze PCT as a continuous biomarker without a priori threshold choice, simultaneously control for potential individual-level confounders, and then provide robust conclusions concerning PCT as a predictor of APN and/or scarring would be to obtain individual data unaltered by thresholds.9 We thus aimed to perform an updated systematic review and meta-analysis on individual patient data to investigate PCT as a predictor for both APN and renal scarring in children with a febrile UTI. The most appropriate threshold values of PCT were simultaneously studied.

measurement, and a renal DMSA scintigraphy published between January 1993 and September 2011. The search methods are detailed in Fig 1. Ethics committees from each participating center approved the protocol for each initial study from which data were collected. All cohort studies of consecutively included children with a febrile UTI, a PCT measurement, and an early (ie, within 14 days) and a late (ie, repeated at least

FIGURE 1
Flow chart of the systematic review. The electronic search was conducted in Medline for all studies of UTI with a PCT measurement in children published from January 1993 (when PCT was rst described in relation to bacterial infection5) through November 2009, and updated in September 2011. The search strategy used medical subject heading terms and text words, including procalcitonin and children. The electronic search was enhanced by handsearching reference lists of all included articles, obtaining any identied articles, and also supplemented by a manual review of abstracts from the European Society of Pediatric Infectious Diseases, the European Society for Pediatric Nephrology, the International Pediatric Nephrology Association, the American Academy of Pediatrics, and the American Society of Nephrology and by discussion with experts in the eld. The electronic search was then validated, comparing the obtained list with the reference list of previous reviews on PCT and UTI7,8 to identify any potential systematic default. No language restriction was used. The search ended with 290 potentially eligible abstracts, among which 19 were considered for inclusion. One article was not included because of absence of DMSA scan data. Finally, 18 articles were included, representing 13 centers as follows: Afula (Israel),28 Ahvaz (Iran),33 Antalya (Turkey),22 Athens (Greece),27 Barcelona (Spain),19,20 Elazig (Turkey),26 Geneva (Switzerland),17,21,31 Lille (France),25 Padova (Italy),32 Thrace (Greece),29,30 Toulouse (France),16 Udine (Italy),23 and Yvoir (Belgium).24

METHODS
We performed a systematic review and meta-analysis on individual patient data, in accordance with international standards (Centre for Reviews and Dissemination guidelines,10 PRISMA,11 and STARD12). We electronically and manually searched for all cohort studies of children with UTI, a PCT

871

3 months later if available) renal DMSA scintigraphy were included. Febrile UTI was dened as fever ($38C) with a positive result on bacterial urine culture (thresholds for collection techniques are given in Table 1). We asked the authors to send us their study data les (duplicates were discarded if any), from which we extracted clinical (gender, age), laboratory (C-reactive protein [CRP], PCT, white blood cell [WBC] count), and radiologic (DMSA scan results, vesicoureteral reux [VUR] grade13 on cystography) data. Information concerning the standard operating procedures used for urine collection, PCT measurement, DMSA scanning (and timing), and cystography at each center was also collected. Methodologic study quality was assessed via a checklist (Supplemental Appendix Tables 1 and 2).14

872 Timing of Late DMSA Scan No. Included (for APN) APN, n (%) No. Included (for LRS) LRS, n (%) Male, n (%) Age (mo), Median (IQR) All-Grade VUR, n (%) Grade $3 VUR, n (%) 36 mo 6 mo 6 mo 3 mo 6 mo 6 mo 6 mo 64 33 61 174 76 80 57 100 61 100 42 72 91 62 (62) 21 (50) 52 (72) 68 (75) ,.001 613 (61) 0 0 61 59 23 (36) 21 (64) 25 (41) 92 (53) 34 (45) 77 (96) 27 (47) 63 (63) 48 (79) 0 23 59 0 76 59 57 79 52 12 mo 624 moc 4 (17) 10 (17) 12 (16) 34 (58) 12 (21) 18 (23) 18 (35) 14 (23) 13 (22) ,.001 525 135 (26) 22 (35) 2 (6) 27 (44) 83 (48) 28 (37) 32 (40) 13 (23) 31 (31) 13 (21) 19 (19) 12 (29) 31 (43) 19 (21) .008 332 (33) 14.0 (4.523.5) 48.0 (24.072.0) 0.6 (0.22.0) 4.8 (2.09.0) 30.0 (10.572) 8.9 (2.421.5) 16.0 (7.040.0) 8.0 (4.017.5) 37.2 (12.578.4) 8.0 (17.557.0) 13.5 (6.048.0) 4.5 (1.19.8) 20.0 (8.864.9) .001 10.0 (4.030.0) 1011 17 (27) 2 (33) 9 (15) 27 (23) 3 (4) 21 (28) 15 (29) 16 (18) 13 (22) 14 (58) 13 (18) 32 (36) ,.001 182 (23) 9 (14) 2 (33) 7 (12) 9 (5) 0 (0) 9 (12) 11 (22) 9 (10) 6 (10) 3 (13) 3 (4) 12 (14) ,.001 80 (10)

We analyzed the relationships between APN/renal damage and PCT, CRP, and WBC count, respectively, using different, backward, stepwise multilevel logisticregression models for each biomarker (center was a group-level variable), with fractional polynomial transformation for continuous variables if the model assumption of linearity was violated. The discriminative ability of each biomarker for APN and then late renal damage was evaluated by drawing receiver operating characteristic (ROC) curves, as well as by calculating sensitivity, specicity, predictive values, and likelihood ratios (LR) after dichotomization. In addition, we compared biomarker models by using decision curve analysis (DCA), a method for evaluating the clinical net benet of prediction models in which the benets (true-positives) are added and the harms (false-positives) are subtracted.15 Due to collinearity, no attempt was made to combine biomarkers. Statistical methods are detailed in the Appendix.

LEROY et al

TABLE 1 Population Characteristics According to Each Center

Centera

Urine Collection Techniques (Threshold of the Positive Bacteriuria)b

SA (any), UC (103) UC (104), CVM (105) SA (103), UC (104), CVM (105) SA (any), UC (5.104), CVM (105) UC (103), CVM (105) SA (103), UC (103), CVM (105) SA (any), UC (104), CVM (105) UC (105), CVM (105) SA (103), UC (5.104), CVM (105)

Centers using SA or UC Afula28 Antalya22 Athens27 Barcelona19,20 Elazig26 Geneva17,21,31 Thrace29,30 Udine23 Yvoir24 Centers using SB Ahvaz33 Lille25 Padova18,32 Toulouse16 Between-center variability (P) Total

SB (105), CVM (105) SB (105) SB (105) SB (105), CVM (105)

References for the articles corresponding to each center are presented in the Supplemental Appendix. CVM, clean-voided midstream; IQR, interquartile range; LRS, late renal scars; SA, suprapubic aspiration; SB, sterile bag; UC, urethral catheterization. a Classied according to the urine collection technique in nontoilet-trained children. b In colony-forming units per milliliter. c Median of 9 months.

ARTICLE

RESULTS
Study Characteristics Following the aforementioned criteria, we retrieved 227 abstracts by electronic searching; 19 were potentially suitable (Fig 1). After full text review, 1 study was not included because of absence of DMSA scan data, leaving 18 articles to be included.1633 The 13 corresponding study authors were contacted; all agreed to participate and send data. A total of 1011 (97.9%) patients fully met the inclusion criteria. All studies had a high methodologic quality (Supplemental Appendix Tables 1 and 2). Nine (69.2%) centers performed both early and late DMSA scans. All centers performed the early DMSA scan within 7 days; 5 of 9 centers performed the late scan at 6 months, and the other centers varied between 3 and 24 months (Table 1). Among the 9 centers performing late DMSA scans, late scanning was systematically conducted regardless of early-scan results in only 1 (11.1%) center (Elazig, Turkey). Nine (69.2%) centers collected urine samples following highquality standard operating procedures (suprapubic aspiration, urethral catheterization for nontoilet-trained children, and clean-voided midstream for the other patients). All centers measured PCT by using the LUMItest PCT immunoluminometric assay or the BRAHMS PCT-Q semiquantitative rapid test (BRAHMS, Hennigsdorf, Germany). All centers included hospitalized children with UTI. No adverse events had been reported in performing PCT measurement, DMSA scanning, or cystography. Table 1 provides details on the characteristics of each center s population. Analysis of APN and late renal scars involved 1011 and 525 patients, respectively. APN by grade was analyzed in 357 patients. PCT as a continuous variable involved only 883 (87.3%) patients,
PEDIATRICS Volume 131, Number 5, May 2013

as PCT was measured by the PCT-Q semiquantitative test for 128 patients. Analysis of CRP and WBC count involved 959 (94.9%) and 962 (95.2%) patients, respectively. VUR was examined in 772 (76.4%) patients. Predicting APN APN was demonstrated in 613 children (60.6%) of the 1011 patients included. The mean 6 SD age of the children was 25.2 6 32.8 months (median: 10.5; interquartile range: 4.332.3); 332 (32.8%) were boys. VUR was diagnosed in 182 (23.6%), and VUR $3 was found in 80 children (10.4%). PCT as a continuous, class, or binary variable was signicantly associated with APN (Table 2, Fig 2). The strength of the association increased when the PCT category (when ordinal variable) increased (Table 2). PCT $0.5 ng/mL (current threshold used in the literature) yielded an adjusted OR of 7.9 (95% CI: 5.810.9). CRP and WBC count were also signicantly related to APN, with similar OR values for CRP as previously described; however, lower OR values were obtained for WBC count (Table 2). PCT as a continuous variable offered an area under the ROC curve (AUC ROC) of 0.82 (95% CI: 0.790.84), after adjusting according to the chosen model. The AUC ROCs for CRP and WBC count were signicantly lower (P , .0001): 0.72 (95% CI: 0.69 0.76) and 0.62 (95% CI: 0.570.65), respectively, once adjusted by using the model (Fig 3). The DCA demonstrated that PCT provided a more statistically robust test than CRP, WBC count, or extreme systematic strategies (ie, DMSA for all or no patients) for all threshold probabilities. A PCT threshold $0.3 ng/ mL (median of nondiseased patient distribution) demonstrated 88% sensitivity (95% CI: 8590), with 47% specicity (95% CI: 4252) (Table 2); interestingly, PCT $0.5 ng/mL offered a higher specicity of 72% (95% CI: 67 76) with a 71% sensitivity (95% CI: 67

74) (Table 3). PCT remained strongly associated with APN when assessed by clinical grade, as did CRP and WBC count (Supplemental Appendix Table 3). Predicting Late Renal Scars Late scars were demonstrated in 135 (25.7%) of the 525 children included. The mean 6 SD patient age was 26.6 6 33.8 months (median: 11.0; interquartile range: 436); 162 (31%) were male. VUR was present in 107 (22.0%) of the 486 patients who underwent cystography; VUR $3 was diagnosed in 51 (10.5%) children. PCT as a continuous and binary variable was signicantly associated with renal scars (Table 2, Fig 2). PCT $0.5 ng/mL yielded an adjusted OR of 3.4 (95% CI: 2.15.7). CRP and WBC count were also signicantly related to renal scarring (Table 2, Fig 2). PCT as a continuous variable resulted in an AUC ROC of 0.75 (95% CI: 0.700.80) once adjusted according to the model built and was signicantly higher (P = .02) than those values observed for CRP and WBC count (0.70 [95% CI: 0.650.76] and 0.66 [95% CI: 0.600.72], respectively) (Fig 3). According to DCA, PCT was better than CRP, WBC count, and both extreme systematic strategies (ie, DMSA for everyone or no one) (Fig 3). PCT $0.5 ng/mL had a 79% of sensitivity (95% CI: 7185), with a 50% specicity (95% CI: 4554) (Table 3).

DISCUSSION
We demonstrated that the measurement of serum PCT can provide considerable predictive value for the development of APN and renal scars, and that this predictive capacity is better than that provided by either CRP or WBC count regardless of considered thresholds. Because the related medical decision process is binary (to perform or not to perform a DMSA scan), our goal was to provide an alternative
873

TABLE 2 Relationship Between APN or Late Renal Scars and PCT, CRP, and WBC Count
Variables APN PCT (ng/mL)a PCT as a continuous variable PCT as a class variable ,0.13 0.130.3 0.30.6 .0.6 PCT as a class variable ,0.5 0.52 210 .10 PCT as a binary variable ($0.3) PCT as a binary variable ($0.5) CRP (mg/L)b CRP as a continuous variable CRP as a class VARIABLE ,10 1020 2060 .60 CRP as a binary variable ($20) CRP as a binary variable ($30) WBC count (cell/mm3)c WBC count as a continuous variable WBC count as a class variable ,10 000 10 00015 000 15 00020 000 .20 000 WBC count, binary variable ($15 000) Late renal scars PCT (ng/mL)d PCT as a continuous variable PCT as a binary variable ($0.5) CRP (mg/L)e CRP as a continuous variable CRP as a binary variable ($20) CRP as a binary variable ($30) WBC count (cell/mm3)f WBC count as a continuous variable WBC count, binary variable ($15 000) Crude OR (95% CI) P Adjusted OR (95% CI) P

2.6 (2.23.1) 1 2.5 (1.34.9) 3.1 (1.65.8) 16.9 (9.330.6) 1 3.5 (2.55.0) 25.3 (14.843.2) 132.5 (17.9979) 5.5 (3.87.8) 7.5 (5.510.3) 2.7 (2.33.3) 1 2.7 (1.45.5) 8.0 (4.215.2) 26.0 (13.649.7) 8.6 (5.513.3) 6.3 (4.49.0) 1.0 (1.01.0) 1 1.8 (1.23.1) 3.1 (1.95.1) 4.9 (2.98.3) 2.4 (1.83.3) 1.8 (1.62.1) 3.8 (2.36.1) 2.3 (1.82.9) 4.8 (2.59.4) 4.6 (2.58.4) 1.0 (1.01.0) 2.1 (1.43.3)

,.0001

2.7 (2.33.1) 1 2.3 (1.24.6) 3.0 (1.65.6) 17.0 (9.330.9) 1 3.7 (2.65.3) 26.3 (15.345.0) 146.3 (19.71085) 5.8 (4.08.3) 7.9 (5.810.9) 2.7 (2.33.3) 1 2.9 (1.45.9) 8.2 (4.315.5) 26.5 (13.850.8) 8.6 (5.513.3) 6.3 (4.49.1) 1.0 (1.01.0) 1 1.8 (1.13.0) 3.0 (1.85.0) 4.9 (2.98.3) 2.4 (1.83.3) 1.7 (1.52.0) 3.4 (2.15.7) 2.2 (1.72.9) 4.6 (2.39.4) 4.2 (2.37.9) 1.0 (1.01.0) 2.1 (1.33.3)

,.0001

.006 .001 ,.0001 ,.0001 ,.0001 ,.0001 ,.0001 ,.0001 ,.0001

.01 .001 ,.0001 ,.0001 ,.0001 ,.0001 ,.0001 ,.0001 ,.0001

.005 ,.0001 ,.0001 ,.0001 ,.0001 ,.0001

.003 ,.0001 ,.0001 ,.0001 ,.0001 ,.0001

assay that could contribute to diagnosis and investigate the optimal threshold by which these clinical decisions could be made. PCT $0.5 ng/mL seemed to offer the optimal compromise of sensitivity and specicity for both APN and late renal scars: 71% sensitivity (95% CI: 6774) with a 72% specicity (95% CI: 6776) for APN; 79% sensitivity (95% CI: 7185) with a 50% of specicity (95% CI: 4554) for late scarring. Furthermore, DCA showed that PCT offered the best benet/harm balance irrespective of the chosen threshold, compared with CRP, WBC count, or systematic strategies (DMSA for everyone or no one) for the selective identication of children who might benet from a DMSA scan. Our ndings add evidence to those of Mantadakis et al.8 Together, they suggest a reasonably strong predictive value based on PCT levels; however, data from studies using pooled estimates lead to a more cautious interpretation due to the signicant heterogeneity found within these study pools. We avoided these issues by working with individual data, adjusting for intercenter variability modeling with multilevel regressions, as well as accounting for all covariables of interest at the individual level. Moreover, the study design (a metaanalysis of individual patient data) allowed us to study the impact of different threshold levels, to perform DCA, and to draw conclusions without the usual threshold effect that often affects diagnostic accuracy assessed by meta-analysis, thus confounding results. Our approach was complementary to that of Zaffanello et al,7 who performed a nonsystematic review of the potential of PCT to predict late renal scarring, without computing pooled estimates, as they were confronted with different studies and cutoffs. With our systematic meta-analysis,

.01 ,.0001 ,.0001 ,.0001 ,.0001 ,.0001 ,.0001 ,.0001 ,.0001 ,.0001 ,.0001

.01 ,.0001 ,.0001 ,.0001 ,.0001 ,.0001 ,.0001 ,.0001 ,.0001 ,.0001 ,.0001

A model was separately derived for each biomarker (PCT, CRP, or WBC count). Univariate and multivariate analysis used a multilevel regression model. The nal multivariate model was built by using a backward stepwise reduction procedure. Each multivariate model for each biomarker was signicantly better than the reduced model based on the maximum likelihood test (P , .05). a The analysis included 883 patients for PCTas a continuous or class variable, 937 patients for PCT dichotomized according to the 0.3 ng/mL threshold, and 1011 patients for PCT dichotomized according to the 0.5 ng/mL threshold (because some patients had a PCT-Q test [see text]). The nal multivariate model was based on PCT and age (as a continuous variable, after linear transformation); continuous PCT was transformed as follows to assess the model linearity assumption: ln(PCT/100) + 3.33732173. b The analysis included 959 patients. The nal multivariate model was based on CRP and age (as a continuous variable, after linear transformation); CRP was transformed as follows to assess the model linearity assumption: ln(CRP/100) + 0.3965495066. c The analysis included 962 patients. The nal multivariate model was based on WBC count and gender. d The analysis included 479 patients. The nal multivariate model was based on PCTand high-grade VUR; PCTwas transformed as follows to assess the model linearity assumption: ln(PCT/100) + 3.293186643. e The analysis included 478 patients. The nal multivariate model was based on CRP and high-grade VUR; CRP was transformed in the same purpose as above into: ln(CRP/100) + 0.439036841. f The analysis included 478 patients. The nal multivariate model was based on WBC count and high-grade VUR.

874

LEROY et al

ARTICLE

FIGURE 2

Distribution of PCT, CRP, and WBC count values according to the presence of APN or late renal scars. (AC) Plots of PCT, CRP, and WBC count for APN. (DF) Plots of PCT, CRP, and WBC count for late renal scars. In each graphic, the horizontal line is the proposed threshold for dichotomizing classication via the biomarker.

we offer further evidence to support their results, updating the review in a systematic manner and providing pooled estimates of the predictive ability of PCT, leading to a robust conclusion. The use of imaging in this eld has been largely debated in the last decade. However, the decision as to which tests, if any, should be routinely conducted in children with UTIs necessarily depends on many factors. The top-down approach uses early DMSA scanning as a screening test.34 Although children with a negative acute-phase DMSA scan are unlikely to develop scarring, DMSA scans are expensive, invasive, and expose children to radiation. However, the top-down strategy raises 2 concerns: rst, it requires DMSA scan availability across countries and settings, which is not currently the case, and second, the identication of late renal scarring
PEDIATRICS Volume 131, Number 5, May 2013

results in only a more careful follow-up of affected children. Therefore, PCT may occupy an intermediate position useful for identifying children at high risk for APN and renal scarring, and for whom a DMSA scan can be selectively proposed to conrm parenchymal involvement. The reported sensitivity and specicity values may not appear very convincing (70%). However, PCT is not meant to replace DMSA scanning, which remains the gold standard for assessing parenchymal involvement (APN or scarring). PCT could be used as an intermediate strategy, based on a single biomarker, easier to set up than a nuclear imaging process, which can help discriminate between lower UTI and APN, even in settings in which DMSA scans are not available. Interestingly, PCT offered the best benet/harm balance irrespective of the chosen threshold, compared with systematic

strategies (DMSA for everyone or no one) for the selective identication of children who might benet from a DMSA scan. Later in the imaging evaluation, a cystography could be proposed for children with a proven APN, to diagnose or rule out VUR, and treat it if necessary. Moreover, PCT may also be helpful when choosing between oral or intravenous antibiotic treatments during the early infectious phase, depending on the severity of the UTI (lower UTI or APN). PCT could nd a place in the debated process of UTI imaging and treatment, as a key point test in the decisional owchart. There are several potential limitations to our study that should be addressed. First, despite the extensive electronic and hand searches performed, a publication bias is possible, especially because test accuracy studies are more
875

FIGURE 3
ROC curve and DCA for PCT, CRP, and WBC count. (A) ROC curves of PCT, CRP, and WBC count for APN. (B) ROC curves of PCT, CRP, and WBC count for late renal scars. (C) The DCA of PCT, CRP, and WBC count for APN. One line represents 1 biomarker (PCT, CRP, and WBC count); the line all represents the benet/harm curve if everyone is investigated with a DMSA scan, whereas the line none represents the corresponding curve if no one undergoes examination. (D) The DCA of PCT, CRP, and WBC count for late renal scars. One line represents 1 biomarker (PCT, CRP, and WBC count); the line all represents the benet/harm curve if everyone is investigated with a DMSA scan, whereas the line none represents the corresponding curve if no one undergoes examination.

easily conducted and abandoned than randomized controlled trials, and are then particularly susceptible to publication bias.35 However, our current knowledge about the precise effects of publication bias on meta-analytic esti-

mates, as well as how to assess the extents of these possible limitations, are limited.36 Therefore, due to the complexity of accurately assessing this issue, we can provide no estimates of the effect of a probable

publication bias. Secondly, a participation bias related to the response and voluntary participation of the centers also might be possible but seems unlikely because all authors contacted responded positively to our

TABLE 3 Diagnostic Accuracy of PCT, CRP, and WBC Count for APN and Late Renal Scars
Variable For APN PCT $0.3 ng/mL PCT $0.5 ng/mL PCT $1 ng/mL CRP $20 mg/L CRP $30 mg/L WBC count $15 000/mm3 For late renal scars PCT $0.3 ng/mL PCT $0.5 ng/mL PCT $1.0 ng/mL CRP $20 mg/L CRP $30 mg/L WBC count $15 000/mm3
Data are presented as % (95% CI).

Sensitivity 88 (8590) 71 (6774) 65 (6169) 87 (8489) 74 (7077) 63 (5967) 91 (8595) 88 (8391) 74 (6681) 85 (7890) 78 (7185) 68 (6075)

Specicity 47 (4252) 72 (6776) 87 (8390) 41 (3747) 54 (4959) 55 (5060) 30 (2635) 50 (4554) 67 (6271) 36 (3241) 47 (4152) 51 (4656)

Positive Predictive Value 74 (7177) 79 (7683) 90 (8692) 70 (6774) 72 (6976) 71 (6774) 31 (2736) 51 (4656) 43 (3750) 32(2737) 34 (2939) 33 (2738)

Negative Predictive Value 69 (6374) 61 (5766) 60 (5564) 66 (5971) 56 (5161) 46 (4251) 91 (8495) 87 (8291) 88 (8492) 87 (8192) 86 (8190) 82 (7786)

Positive LR 1.6 (1.51.8) 2.5 (2.13.0) 4.9 (3.76.5) 1.5 (1.31.6) 1.6 (1.41.8) 1.4 (1.21.6) 1.3 (1.21.4) 1.6 (1.41.8) 2.3 (1.92.6) 1.3 (1.21.5) 1.5 (1.31.7) 1.4 (1.21.6)

Negative LR 0.3 (0.20.3) 0.4 (0.40.5) 0.4 (0.40.5) 0.3 (0.30.4) 0.5 (0.40.6) 0.7 (0.60.8) 0.3(0.20.5) 0.4 (0.30.6) 0.4 (0.30.5) 0.4 (0.30.6) 0.5 (0.30.7) 0.6 (0.50.8)

876

LEROY et al

ARTICLE

requests for patient data sets. Thirdly, the possibility of a classication bias seems unlikely because PCT was measured by using validated techniques (immunoluminometric assay or semiquantitative PCT-Q assay), while blinded to the outcome. Fourthly, we assumed that patients who had a normal DMSA had no late lesions even if late DMSA was not performed. However, this assumption is commonplace in the literature,37 and we veried this assumption in the only center (Elazig, Turkey) in which all patients systematically underwent both late and early DMSA: none of the negative early DMSA cases were followed by a positive late DMSA. This outcome gives an indication on the robustness of our assumption. Fifth, we addressed heterogeneity issues due to data pooling from different centers (including different time frames for the late DMSA scan) by analyzing them as hierarchical data and using multilevel modeling. We chose to analyze the data set as a meta-analysis of individual patient data because this method provides the least biased and most reliable means of addressing the questions at hand.9 Sixth, technical concerns, such as the collection of urine from nontoilet-trained children in sterile bags at 4 of the selected centers (not a recommended method) could have led to selection bias.38,39 This procedure could have increased the number of false-positive results for UTI but without consequences to the relationship between APN or late scarring and PCT or other biomarkers.

In addition, the inclusion of only hospitalized children by the centers might have led to another selection bias, due to the inclusion of only the sickest children. However, all children with febrile UTI were systematically hospitalized in the included centers. Seventh, the absence of previous negative DMSA scintigraphy results might also have introduced a selection bias. Even if all the centers conrmed having included exclusively or mostly children with a rst febrile UTI, it could not guarantee that previous UTI with persistent scarring had not occurred in the included patients, as correct diagnosis depends on the clinical evaluation performed during previous febrile episodes. Moreover, not all UTIs in children are accompanied by fever, which is the main clinical reason for obtaining urine cultures in infants; therefore, previous UTIs unaccompanied by fever may not have been clearly identied. Lastly, the delay between the rst indications of infection and PCT level measures was not taken into account, and this might have introduced a bias in the results but only by underestimating the relationship between APN or late scarring and PCT, because this marker increases as early as 6 hours postinfection and also decreases just as quickly at the end of infection.

stages of UTI and those that developed later renal scarring. The use of serum PCT measurements has the potential to aid the clinical decision-making process regarding the appropriate acute management of children with UTI. In particular, due to limited resources and technical availability, it may be helpful to use such an assay to selectively identify children who may benet from a DMSA scan at the early and late stages of infection. The impact of PCT measurements on the currently debated practice of UTI examinations needs to be evaluated by a welldesigned impact study and may lead to possible renement of the decisional process.

CONCLUSIONS
We demonstrated that PCT has a robust predictive ability to selectively identify children who had APN in the early

ACKNOWLEDGMENTS The authors thank Dr Gardikis and Dr Deftereos (Department of Pediatric Surgery and Radiology, Alexandroupolous University Hospital, Democritus University of Thrace School of Medicine, Thrace, Greece), Dr Galetto-Lacour (Department of Pediatrics, University Hospital of Geneva, Geneva, Switzerland), Dr Ellero (Department of Pediatrics, University of Udine, Udine, Italy), Professor Da Dalt (Department of Pediatrics, University of Padova, Italy) for participation in data collection/sharing, Dr Bacchetta (Department of Pediatric NephrologyReference Centre for Rare Renal Diseases, Femme Mre Enfant Hospital, University of Lyon, Lyon, France) for helpful discussions, and Melissa Laird (Inserm U818, Institut Pasteur, Paris, France) for pertinent comments and corrections.

REFERENCES
1. Mrild S, Jodal U. Incidence rate of rst-time symptomatic urinary tract infection in children under 6 years of age. Acta Paediatr. 1998;87(5):549552 2. Vernon SJ, Coulthard MG, Lambert HJ, Keir MJ, Matthews JN. New renal scarring in children who at age 3 and 4 years had had normal scans with dimercaptosuccinic acid: follow up study. BMJ. 1997;315(7113):905908 3. Williams G, Fletcher JT, Alexander SI, Craig JC. Vesicoureteral reux. J Am Soc Nephrol. 2008;19(5):847862 4. Rushton HG, Majd M. Dimercaptosuccinic acid renal scintigraphy for the evaluation of pyelonephritis and scarring: a review of experimental and clinical studies. J Urol. 1992;148(5 pt 2):17261732 5. Assicot M, Gendrel D, Carsin H, Raymond J, Guilbaud J, Bohuon C. High serum procalcitonin concentrations in patients with sepsis and infection. Lancet. 1993;341(8844):515518

PEDIATRICS Volume 131, Number 5, May 2013

877

6. Simon L, Gauvin F, Amre DK, Saint-Louis P, Lacroix J. Serum procalcitonin and C-reactive protein levels as markers of bacterial infection: a systematic review and meta-analysis. Clin Infect Dis. 2004;39(2): 206217 7. Zaffanello M, Brugnara M, Franchini M, Fanos V. Is serum procalcitonin able to predict long-term kidney morbidity from urinary tract infections in children? Clin Chem Lab Med . 2008;46(10): 1358 1363 8. Mantadakis E, Plessa E, Vouloumanou EK, Karageorgopoulos DE, Chatzimichael A, Falagas ME. Serum procalcitonin for prediction of renal parenchymal involvement in children with urinary tract infections: a meta-analysis of prospective clinical studies. J Pediatr. 2009;155(6): 875881, e1 9. Stewart LA, Parmar MK. Meta-analysis of the literature or of individual patient data: is there a difference? Lancet. 1993;341 (8842):418422 10. Centre for Review Dissemination. Systematic reviews: CRDs guidance for undertaking reviews in health care. Available at: www.yorkacuk/inst/crd/SysRev/!SSL!/WebHelp/ SysRev3htm. Accessed December 1, 2012 11. Moher D, Liberati A, Tetzlaff J, Altman DG; PRISMA Group. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. BMJ. 2009;339: b2535 12. Bossuyt PM, Reitsma JB, Bruns DE, et al; Standards for Reporting of Diagnostic Accuracy. Towards complete and accurate reporting of studies of diagnostic accuracy: the STARD initiative. BMJ. 2003;326(7379): 4144 13. Lebowitz RL, Olbing H, Parkkulainen KV, Smellie JM, Tamminen-Mbius TE; International Reux Study in Children. International system of radiographic grading of vesicoureteric reux. Pediatr Radiol. 1985; 15(2):105109 14. Whiting P, Rutjes AW, Reitsma JB, Bossuyt PM, Kleijnen J. The development of QUADAS: a tool for the quality assessment of studies of diagnostic accuracy included in systematic reviews. BMC Med Res Methodol. 2003;3(25):113 15. Vickers AJ, Elkin EB. Decision curve analysis: a novel method for evaluating prediction models. Med Decis Making. 2006;26 (6):565574 16. Belhadj-Tahar H, Coulais Y, Tafani M, Bouissou F. Procalcitonin implication in renal cell apoptosis induced by acute pyelonephritis in children. Infect Drug Resist. 2008;1:1720

17. Benador N, Siegrist CA, Gendrel D, et al. Procalcitonin is a marker of severity of renal lesions in pyelonephritis. Pediatrics. 1998;102(6):14221425 18. Bressan S, Andreola B, Zucchetta P, et al. Procalcitonin as a predictor of renal scarring in infants and young children. Pediatr Nephrol . 2009;24(6): 1199 1204 19. Fernndez Lopez A, Luaces Cubells C, Garca Garca JJ, Fernndez Pou J; Spanish Society of Pediatric Emergencies. Procalcitonin in pediatric emergency departments for the early diagnosis of invasive bacterial infections in febrile infants: results of a multicenter study and utility of a rapid qualitative test for this marker. Pediatr Infect Dis J. 2003;22(10): 895903 20. Fernndez Lpez A, Luaces Cubells C, Valls Tolosa C, et al. Procalcitonin in the early diagnosis of invasive bacterial infection in febrile infants [in Spanish]. An Esp Pediatr. 2001;55(4):321328 21. Gervaix A, Galetto-Lacour A, Gueron T, et al. Usefulness of procalcitonin and C-reactive protein rapid tests for the management of children with urinary tract infection. Pediatr Infect Dis J. 2001; 20(5):507511 22. Gven AG, Kazdal HZ, Koyun M, et al. Accurate diagnosis of acute pyelonephritis: How helpful is procalcitonin? Nucl Med Commun. 2006;27(9):715721 23. Pecile P, Miorin E, Romanello C, et al. Procalcitonin: a marker of severity of acute pyelonephritis among children. Pediatrics. 2004;114(2)e249e254. Available at: www.pediatrics.org/cgi/content/full/ 114/2/e249 24. Tuerlinckx D, Vander Borght T, Glupczynski Y, et al. Is procalcitonin a good marker of renal lesion in febrile urinary tract infection? Eur J Pediatr. 2005;164(10): 651652 25. Bigot S, Leblond P, Foucher C, Hue V, D Herbomez M, Foulard M. Usefulness of procalcitonin for the diagnosis of acute pyelonephritis in children [in French]. Arch Pediatr. 2005;12(7):1075 1080 26. Grgze MK, Akarsu S, Yilmaz E, et al. Proinammatory cytokines and procalcitonin in children with acute pyelonephritis. Pediatr Nephrol. 2005;20(10): 14451448 27. Karavanaki K, Haliotis FA, Sourani M, et al. DMSA scintigraphy in febrile urinary tract infections could be omitted in children with low procalcitonin levels. Infect Dis Clin Pract. 2007;15(6):377381

28. Smolkin V, Koren A, Raz R, Colodner R, Sakran W, Halevy R. Procalcitonin as a marker of acute pyelonephritis in infants and children. Pediatr Nephrol. 2002;17(6): 409412 29. Kotoula A, Gardikis S, Tsalkidis A, et al. Comparative efcacies of procalcitonin and conventional inammatory markers for prediction of renal parenchymal inammation in pediatric rst urinary tract infection. Urology. 2009;73(4):782 786 30. Kotoula A, Gardikis S, Tsalkidis A, et al. Procalcitonin for the early prediction of renal parenchymal involvement in children with UTI: preliminary results. Int Urol Nephrol. 2009;41(2):393399 31. Galetto-Lacour A, Zamora SA, Gervaix A. Bedside procalcitonin and C-reactive protein tests in children with fever without localizing signs of infection seen in a referral center. Pediatrics. 2003;112(5):1054 1060 32. Andreola B, Bressan S, Callegaro S, Liverani A, Plebani M, Da Dalt L. Procalcitonin and C-reactive protein as diagnostic markers of severe bacterial infections in febrile infants and children in the emergency department. Pediatr Infect Dis J. 2007;26(8): 672677 33. Nikfar R, Khotaee G, Ataee N, Shams S. Usefulness of procalcitonin rapid test for the diagnosis of acute pyelonephritis in children in the emergency department. Pediatr Int. 2010;52(2):196198 34. Preda I, Jodal U, Sixt R, Stokland E, Hansson S. Normal dimercaptosuccinic acid scintigraphy makes voiding cystourethrography unnecessary after urinary tract infection. J Pediatr. 2007;151(6):581584, 584, e1 35. Song F, Khan KS, Dinnes J, Sutton AJ. Asymmetric funnel plots and publication bias in meta-analyses of diagnostic accuracy. Int J Epidemiol. 2002;31(1):88 95 36. Deeks JJ, Macaskill P, Irwig L. The performance of tests of publication bias and other sample size effects in systematic reviews of diagnostic test accuracy was assessed. J Clin Epidemiol. 2005;58(9):882 893 37. Shaikh N, Ewing AL, Bhatnagar S, Hoberman A. Risk of renal scarring in children with a rst urinary tract infection: a systematic review. Pediatrics. 2010;126(6): 10841091 38. American Academy of Pediatrics. Committee on Quality Improvement. Subcommittee on Urinary Tract Infection. Practice parameter: the diagnosis, treatment,

878

LEROY et al

ARTICLE

and evaluation of the initial urinary tract infection in febrile infants and young children. Pediatrics. 1999;103(4 pt 1):843 852

39. Downs SM; The Urinary Tract Subcommittee of the American Academy of Pediatrics Committee on Quality Improvement. Technical report: urinary tract

infections in febrile infants and young children. Pediatrics. 1999;103(4):e54. Available at: www.pediatrics.org/cgi/content/ full/103/4/e54

(Continued from rst page) Dr Leroy designed the study, helped in collecting the data, analyzed the data, interpreted results, drafted the manuscript, and approved the article version to be published; Drs Fernandez-Lopez, Nikfar, Romanello, Bouissou, Gurgoze, Bressan, Smolkin, Leblond, Mr Vaos, and Mr Gervaix contributed to the study design, acquired all the data in their individual centers, revised the paper for important intellectual content, and approved the article version to be published; Drs Tuerlinckx and Gungor contributed to the study design, acquired all the data in their centers, and revised the paper for important intellectual content; Mr Stefanidis had great input in the conception of the study, realized the data collection in his center, performed in-depth revision of the manuscript, and approved the article version to be published; and Ms Gendrel and Mr Chalumeau contributed to the study design and interpretation of results, critically revised the manuscript, and approved the article version to be published. www.pediatrics.org/cgi/doi/10.1542/peds.2012-2408 doi:10.1542/peds.2012-2408 Accepted for publication Jan 24, 2013 Address correspondence to Sandrine Leroy, MD, PhD, Centre for Statistics in Medicine, University of Oxford, Wolfson College Annexe, Linton Rd, Oxford OX2 6UD, United Kingdom. E-mail: sandrin.leroy@gmail.com PEDIATRICS (ISSN Numbers: Print, 0031-4005; Online, 1098-4275). Copyright 2013 by the American Academy of Pediatrics FINANCIAL DISCLOSURE: The authors have indicated they have no nancial relationships relevant to this article to disclose. FUNDING: Dr Leroy was funded by a postdoctoral fellowship grant from the French Society of Nephrology and from the Fondation Bettencourt.

PEDIATRICS Volume 131, Number 5, May 2013

879