THE CIRCUMCISION REFERENCE LIBRARY
J. C. Craig, MBChB, DCh, FRACP, J. F. Knight MBBS, FRACP, P. Sureshkumar, BSc. F. Mantz, MRCP. and L. P. Roy, MBBS, FRACP
From the Department of Nephrology, Royal Alexandra Hospital for Children, Sydney, Australia
Supported by the National Health and Medical Research Council, the Children's Hospital Research Fund, and the Manildra Group of Companies.
Submitted for publication June 27, 1995; accepted Aug. 31, 1995. Reprint requests; J. C, Craig, MBChB, DCH, FRACP, Department of Nephrology Royal Alexandra Hospital for Children, PO Box 3515, Parramatta, New South Wales 2124, Australia. Copyright (C) 1996 by Mosby-Year Book, Inc. 0022-3476/96/$5.0O + 0 9/20/68991
Objective: To determine whether circumcision decreases the risk of symptomatic urinary tract infection (UTI) in boys less than 5 years of age.
Study design: A case-control study (1993 to 1995) in the setting of a large ambulatory pediatric service. Case subjects and control subjects were drawn from the same population. One hundred forty-four boys less than 5 years of age (median age, 5.8 months) who had a microbiologically proven symptomatic UTI (case subjects), were compared with 742 boys (median age, 21.0 months) who did not have a UTI (control subjects). The proportion of case and control subjects who were circumcised in each group was compared with the use of the chi-square test, with the strength of association between circumcision and UTI expressed in terms of an odds ratio. To determine whether age was a confounder or an effect-modifier, we stratified the groups by age (<1 year; 1-5 years) and analyzed by the method of Mantel-Haenszel.
Results: Of the 144 preschool boys with UTI, 2(1.4%) were circumcised, compared with 47 (6.3%) of the 742 control subjects (chi-square value = 5.6; p = 0.02; odds ratio, 0.21; 95% confidence intervals, 0.06 to 0.76). There was no evidence that age was a confounder or modified the protective effect of circumcision on the development of UTI (Mantel-Haenszel chi-square value = 6.0; p = 0.01; combined odds ratio, 0.18; 95% confidence intervals, 0.05 to 0.71; Breslow-Day test of homogeneity chi-square value = 0.6; p = 0.4).
Conclusions: Circumcision decreases the risk of symptomatic UTI in preschool boys. The protective effect is independent of age.
(J PEDlATR 1996; 128:23-7)
Approximately 1% to 2% of boys can be expected to have a urinary tract infection by 10 years of age.1,2 Since 1982 it has been recognized that there is an association between UTI and the uncircumcised state.3 Cross-sectional,3 ecologic,4 before-and-after,5,6 case-control,7,8 and retrospective cohort studies5,9,10 have consistently demonstrated that an uncircumcised state is associated with a higher incidence of UTI.
These studies have been criticized on methodologic grounds, and there is not universal acceptance that circumcision prevents the development of UTI.11 All have been conducted in populations with a high prevalence of circumcision, and apart from a retrospective case-control study in a sexually transmitted disease clinic for adult men,8 the effect of circumcision on the development of UTI has not been examined in subjects more than 1 year of age. To determine whether the protective effect of circumcision extends to older children, we undertook a case-control study of boys less than 5 years of age attending a large ambulatory pediatric department in Australia, a country with a low prevalence of circumcision.
Case selection. To identify cases of UTI in boys less than 5 years of age, we prospectively reviewed all urine cultures that were obtained from eligible boys attending the ambulatory pediatric department of the Royal Alexandra Hospital for Children, Sydney, from March 1, 1993, to Dec. 17, 1994. All urine samples were innoculated onto blood agar and MacConkey medium with a 1 ul calibrated loop. Our case criteria were a colony-forming unit count of more than 106/L from urine obtained from a suprapubic tap or catheter sample, or more than 108/L from a midstream voided urine sample of a single organism, in combination with symptoms and/or signs of UTI. Meatal cleansing was not performed before the collection of midstream urine samples.12 Boys who had a previously diagnosed UTI, or a known renal tract, neurologic, or skeletal predisposing cause, were excluded. Investigators were unaware of the circumcision status of case patients at the time of diagnosis.
As part of a prospective cohort study of 304 children less than 5 years of age, and to evaluate associated renal tract ahnormalities, we obtained a technetium Tc 99m dimercaptosuccinic acid scan, a radiologic micturating cystourethrogram, and renal tract sonograms within 4 weeks of onset. Institutional ethics approval had been obtained for the study, and informed consent was obtained from the patents of the children who were study subjects.
Selection of control subjects. To identify control subjects, we ascertained, from April 5, 1995, to May 2, 1995 (28 days), the circumcision status and diagnosis of all boys less than 5 years of age who attended the ambulatory pediatric center of the Children's Hospital. It was mandatory for relevant patient details to be recorded into a computer database for all children who came for medical attention. During the study period, circumcision status was included in the required fields for all boys less than 5 years of age. The information was elicited by questioning of the parents, or when the parents were uncertain or did not have the required English-language proficiency, by direct inspection of the penis. Boys who fulfilled the criteria for the diagnosis of UTl were excluded.
Statistical analysis. Comparison between case patients and control subjects was made by means of the chi-square test for categorical variables and the Mann-Whitney U test for continuous variables, with the strength of association between circumcision and UTI expressed in terms of the odd ratio. To determine whether age was a confounder, we stratified the groups by age (<1 year; >=1 year) and analyzed by the Mantel-Haenszel method. The confidence limits for the combined odds ratio were generated by the test-based method. The Breslow-Day test of homogeneity was used to determine whether age was an effect modifier in the association between circumcision and UTI Data analyses were carried out with the SAS (SAS Institute Inc.. Carry, N.C.) and Minitab statistical packages (Minitab Inc., State College, Pa.).
Clinical findings. One hundred forty-four cases of UTI were identified. The method of urine collection and of counting colony-forming units are given in Table I. Escherichia coli was the pathogenic organism in 124 (86.1%), Proteus sp. in Il (7.6%), Klebsiella sp. and Escherichia faecalis in four each (2.8%), and Cornebacterium minutissimum in a single case (0.7%). The median age was 5.8 months, with a range of 0.3 to 54.5 months (Table II). Fever, the commonest initial symptom, was present in 111 boys (77.1%). Seventy-five boys (52.1%) were hospitalized for at least one night, and 7 (9.2%), of 76 boys from whom blood culture specimens were obtained, had concomitant bacteremia. There were no episodes of meningitis and no deaths.
Table I. Colony-forming unit count of cases by method of urine collection Voided catheter Suprapubic tap CFU No.(%) No. (%) No. (%) Total (%) >10 8/L 56 25 28 109 (75.6) 10 7/L-10 8/L 18 13 31 (21.6) 1O6/L-1O7/L - 3 4 (2.8) Total (%) 56 (38.9) 44 (30.6) 44 (30.6) 111 (100) CFU, Colony-forming unit.
Imaging. Cortical scintigraphy, using 99m Tc-dimercaptosuccinic acid, was undertaken for 139 boys (96.5%) at a median time of 7 days after diagnosis (range, 1 to 35 days), and demonstrated parenchymal abnormality in 55 boys (39.6%). Vesicoureteric reflux was present in 33 (24.1%) of the 137 boys who underwent micturating cystourethrography. All boys underwent renal ultrasonography; renal tract abnormalities, apart from vesicoureteric reflux were detected in 12(8.3%). Obstructive uropathy was present in 3 boys, primary megaureter in 6, a horseshoe malformation in 2, and a duplex collecting system in 1.
Control subjects. During the 28-day period for the selection of control subjects, 747 boys less than 5 years of age were treated at the ambulatory pediatric center. Five boys were excluded (all uncircumcised) because symptomatic urinary tract infection was the primary diagnosis. The median age of control subjects was 21.0 months (range, 0.1 to 59.9 months). Respiratory tract-related illness (271), gastrointestinal disease (110), accidents (93), other infections (76), and dermatologic problems (36) were the most common diagnoses.
Comparison of case subjects with control subjects. Of 742 boys, 47 (6.3%; 95% confidence intervals, 4.6% to 8.1%) in the control group were circumcised compared with 2 (1.4%) of 144 case patients (95% confidence intervals, -0.5% to 3.3%; odds ratio = 0.21; 95% confidence intervals, 0.06 to 0.76; chi-square value = 5.6; p = 0.02).
Table II Age distribution of case and control subjects Control subjects case subjects No. (%) No. (%) Total (%) Age (yr) 0-1 231 110 341 (38.5) 1-2 178 10 188 (21.2) 2-3 136 8 144 (10.3) 3-4 114 7 121 (13.7) 4-5 83 9 92 (10.3) TOTAL (%) 742 (83.7) 144 (16.3) 886(100) Median ages: control subjects, 21.0 months; ease subjects, 5.8 months; p <0.001 (Mann-Whitney U test).
Case patients had a lower median age (5.8 months) than control subjects had (21.0 months) (Table II). To determine whether age was a significant covariate in the effect of cir-cumcision on UTI, we compared the ages of circumcised and uncircumcised control subjects (Table III). There was no evidence of an association between age and circumcision status, and therefore no evidence that the difference in ages between case subjects and control subjects biased the association between circumcision and UTI. Age is known to be strongly associated with UTI in boys, so stratification by age (<1 year; 1 to 5 years) was performed so that the effect of circumcision in groups more homogeneous for age could be examined (Table IV). The combined odds ratio was 0.18 (95% confidence intervals, 0.05 to 0.70; chi-square value [Mantel-Haenszel method] = 6.0;p = 0.01); that is, there was evidence of an association between circumcision and UTI after allowance for age. There was no evidence that the patient's age modified the protective effect of circumcision status on the development of UTI (chi-square value = 0.6; p = 0.4).
Table III. Frequency distribution of circumcision in control boys by age circumcised uncircumcised No. (%) Age (yr) 0-1 215 16 6.9 1-2 165 13 7.3 2-3 128 8 5.9 3-4 110 4 3.5 4-5 77 6 7.2 TOTAL 695 47 6.3 Median ages: uncircumcised, 21.0 months; circumcised, 20.0 months; p = 0.8 Mann-whitney U test).
Table IV. Frequency of circumcision in case and control subjects stratified by age Case subjects Control subjects children <1 yr of age* Uncircumcised 108 215 circumcised 2 16 TOTAL 110 231 children >=1 yr of age** Uncircumcised 34 480 Circumcised 0 31 TOTAL 34 511 Combined odds ratio (Mantel-Haenszel test) = 0.18 (95% confidence intervals, 0.05 to 0.70); chi-square value=6.0; p=0.01. Breslow-Day lest for homogeneity chi-squane value = 0.6; p = 0.4. *n = 341, chi-square value = 3.9; p = 0.05; odds ratio = 0.03; 95% confidence intervals, 0.06 to 1.1. **n = 545. chi-squane value = 2.2; p = 0.1; odds ratio = 0.2; 95% confidence intervals 0.01 to 3.7.
The association between circumcision and UTI was first recognized by Ginsberg and McCracken3 in 1982 in a cross-sectional study of 100 hospitalized infants with UTI. Despite differences in study populations and design, all subsequent studies have consistently demonstrated an association between an uncircumcised state and UTI.
The strongest evidence of a causal link between an intact foreskin and UTI comes from the series of retrospective cohort studies from Wiswell and colleagues.5,9,10 When the incidence of UTI in the uncircumcised and the circumcised groups was compared,13 the relative risk was found to be 0.10. The attributable risk percentage* can be calculated to be 90.2%, and the number required to be circumcised to prevent one UTI is 99. With a relative risk of 0.1, it is unlikely that any confounding factor would significantly alter the strength of this association. Nonblinded classification and the inclusion of selected outcomes (hospitalized infants with UTI) are significant potential sources of bias in these studies, and no attempt was made to determine whether the protective effect of circumcision extended beyond 1 year of age.
Only one previous study has examined the effect of circumcision on UTI beyond infancy. Spach et al.8 published a retrospective case-control study of adult men from a sexually transmitted disease clinic. Both case and control subjects had urinary symptoms, but case subjects had microbiologically proven UTI was a significant association between circumcision and UTI, but the generalizability of this study to an ambulatory pediatric population could be questioned. The biologic plausibility of an association between an intact foreskin and circumcision has been demonstrated. With electron microscopy, uropathogens have been shown to bind avidly to the mucosal surface of the foreskin,14 but not to the stratified squamous epithelium, and two studies have shown a higher rate of colonization with these organisms in the periurethral area of uncircumcised boys compared with circumcised boys.15,16 Because most available evidence points to the ascending route as the mechanism by which urinary pathogens gain access to the renal tract, it is plausible that removing the foreskin should reduce the frequency of UTI.
Our data confirm the findings of previous studies in North American populations with relatively high rates of circumcision, and demonstrate that UTIs also occur more commonly in uncircumcised boys than circumcised boys beyond the age of 1 year, and in Australia, where only 16% to 19% of males are now circumcised (data obtained from Medicare Estimates and Statistics Section, Australia, February 1994). It is biologically implausible for the association between circumcision status and UTI to cease at 1 year of age. The relatively low frequency of UTI after the first year of age in boys, and the methods used to select subjects, account for the absence of an observed effect beyond the first year of life in previous publications. In our study, 79.2% of the risk of UTI in boys less than 5 years of age can be attributed to circumcision status.* The age distribution of our case subjects, prevalence of associated renal tract abnormalities, and causative urinary pathogens were similar to those in other studies.1,2,17 Only 52.5% in our series were hospitalized.
There are potential limitations with our study design. Selection of case and control subjects was sequential rather than simultaneous. However, the assumption that the population from which case and control subjects were drawn did not change significantly in the intervening period is reasonable; the intervening period was only 3 months, and there was no evidence of significant changes in the prevalence of circumcision with patient age or changes in the incidence of UTI. Because UTI is more common in infant boys, it was not surprising that case subjects were significantly younger than control subjects. Because no association between age and circumcision status was demonstrated, age was not acting as a confounder in these data.
Previous studies have been used as an argument for18 and against11 routine neonatal circumcision. Proponents emphasize the short- and long-term morbidity and mortality rates associated with UTI.18 In our series of 144 consecutive cases of UTI in boys less than 5 years of age, the acute illness caused significant short-term morbidity, but death and meningitis did not occur. On the basis of the number-needed-to-treat analysis calculated from the data of Wiswell et al.,10 99 boys would need to be circumcised to prevent one UTI. The true complication rate of circumcision is not known, but the reported rate is 0.19%10,19 to 5%,20 To further quantify this risk, one can calculate that, in a cohort of 100 boys circumcised to prevent one UTI, 0.2 to 5 boys will have complications, albeit usually of a minor nature, in addition to the cost and discomfort of the procedure. The incidences of permanent renal damage, hypertension, and chronic renal failure in children who have had a UTI are not known.
We conclude that circumcision is associated with a lower incidence of symptomatic UTI in a population with a low prevalence of circumcision, and beyond the first year of life.
We thank Professor Geoffrey Berry for assistance with the statistical analysis: Ralph Hanson, Audrey Chan, and Peter Greenacre for technical assisiance; and associate professor Les Irwig for comments on earlier drafts.
*Attributable risk percentage =(Odd ratio - 1/Odds ratio) x 100
[CIRP Note: Craig later participated in a systematic review of the literature in which the authors concluded that prevention of UTI could not be used to justify non-therapeutic (routine) circumcision of the newborn. See Circumcision for the prevention of urinary tract infection in boys: A systematic review of randomized trials and observational studies.]
Return to CIRP library