Why does prevention of recurrent UTI not result in less renal scarring? A Deeper Dive Into Randomized Intervention for Children with Vesicoureteral Reflux (RIVUR) Trial

Abstract

Background:

RIVUR reported antibiotic prophylaxis(AP) reduced recurrent UTI(rUTI), but AP was not associated with decreased new renal scarring(NRS). However, the original reports did not assess the relationship between rUTI, NRS, and AP in detail. We sought to investigate the relationship between AP, UTI, and NRS.

Methods:

We included subjects with DMSA study within 6 months of enrollment, and at least 1 follow-up DMSA scan from RIVUR. The primary outcome was rUTI-associated NRS, defined as rUTI AND new changes on DMSA. Due to low number of events, propensity score was used to adjust for confounders. Multivariate logistic regression was fitted to investigate the associations between the covariates and outcome.

Results:

489 patients(91% female, mean age 20.3 months) were included. Any NRS was more common among those with rUTI(OR=4.1(2.0–8.5), p<0.01) after adjustment of age, sex, index UTI, duplication, bowel-bladder dysfunction (BBD), and AP.

rUTI-associated NRS occurred in 5/244(2%) with AP and 13/245(5%) with placebo. When comparing to AP, placebo was associated with higher risk of rUTI-associated NRS(OR=3.1(1.0–8.8),p=0.04), after adjusting with age, sex, race, index UTI, BBD, duplication, hydronephrosis, VUR grade, and baseline renal scarring. There were no differences in scar severity at final DMSA scan(p=0.88) or change from baseline(p=0.53) between AP and placebo.

Conclusions:

Recurrent UTI was associated with NRS in RIVUR. When limited to rUTI-associated NRS, AP was associated with a decreased risk of this outcome. It remains unclear why a proportion of RIVUR subjects without rUTI developed NRS. The results should be carefully interpreted due to the inherent limitations.

INTRODUCTION

Significant controversy and variability exists regarding vesicoureteral reflux (VUR) management.¹ In particular, the effectiveness of antibiotic prophylaxis (AP) in VUR children has been highly debated.² In 2014, the multi-institutional Randomized Intervention for Children with Vesicoureteral Reflux (RIVUR) trial was published³, demonstrating that AP had strong protective effect on recurrent urinary tract infection (rUTI) in VUR. However, the incidence of new renal scarring (NRS) was not significantly different between AP and placebo. While the lack of impact on NRS prompted many to conclude that AP is not an effective intervention in primary VUR, this result is puzzling, if our underlying understanding of reflux nephropathy is correct. If VUR-associated pyelonephritis leads to renal injury through an inflammatory cascade, why did a 50% reduction in rUTI not result in less new scarring? Possible explanations include the relatively low baseline prevalence of scarring, the short study duration (2 years), and early treatment of rUTI (and thus prevention of scar development) among both AP and placebo subjects in the clinical trial setting. Others argue that other factors such as innate immunity⁴ or bowel bladder dysfunction (BBD)⁵ are more significant contributors.

None of these explanations, however, address the fundamental paradox of the RIVUR results: rUTI was significantly more common among those in the placebo group, and NRS was significantly more common among subjects with rUTI, and yet NRS was not more common in the placebo group. This seeming discrepancy prompted us to dig deeper as we sought to better understand the relationship between prophylaxis, recurrent infection, and renal scarring. Specifically, we sought to determine the relationship of AP with the incidence of NRS specifically occurring after rUTI, and we hypothesized that rUTI-associated NRS would be more common in patients on placebo than on AP.

MATERIALS and METHODS

Data source and Cohort Selection

RIVUR was a multicenter, randomized, double-blinded, placebo-controlled trial designed to determine whether daily antimicrobial prophylaxis is superior to placebo in preventing rUTI in children with VUR. The trial cohort and rationale were published previously.³ Briefly, the eligibility included (1) age at randomization between 2-months and 6-year, (2) a diagnosed first/second febrile or symptomatic index UTI within sixteen weeks prior to randomization, and (3) presence of VUR (from VCUG). Patients were followed for 2 years with primary outcome of recurrence of febrile and/or symptomatic UTI.

Per the trial protocol, study participants were scheduled for three DMSA renal scans. The baseline scan was obtained within 2 weeks of randomization and within sixteen weeks of the index UTI. A second DMSA scan was obtained within 21 days of the 12-month follow-up visit. The third DMSA scan was targeted within 10 days of the study exit visit at 24 months after randomization. For all children with treatment failure, the outcome DMSA scan was obtained approximately 4 months after meeting criteria for treatment failure. The DMSA scan review protocol by central readers was previously described.⁶ In brief, each kidney was divided into 12 zones. The severity (grade) of the renal scarring was categorized as mild (1–2 segments affected), moderate (3–4 segments affected), severe (>4 segments affected), and global atrophy.

We reviewed the RIVUR data from the National Institute of Diabetes and Digestive and Kidney Diseases data repository. We included subjects with a baseline DMSA study within 6 months of enrollment (to reduce possible bias toward more baseline scarring), and at least 1 follow-up DMSA scan. Those whose first DMSA scan was later than 6 months after the index UTI, those with rUTI prior to first DMSA scan (and therefore no baseline DMSA scan available), and those with rUTI occurring after their last DMSA scan (and therefore no DMSA scan to reflect the rUTI impact), were excluded (Figure 1).

Figure 1-. Flow Chart of Cohort Selection.

Covariates and Outcome Definition

The primary outcome was rUTI-associated NRS, defined as rUTI and NRS detected after rUTI event. NRS was defined within the RIVUR protocol as a change in number of renal segments with scarring between from initial DMSA and last DMSA. Subjects with NRS but no rUTI, with rUTI but no NRS, and those with neither were defined as negative for the primary outcome.

Covariates included demographic factors (age, gender, race), bowel-bladder dysfunction (among toilet-trained children, defined by dysfunctional voiding scale per protocol), index UTI presentation (number of episodes, fever, symptoms), ultrasound results (hydronephrosis, hydroureter, bladder wall thickening), VUR grade/laterality.

RIVUR used stringent criteria for UTI.⁷ For both index or recurrent UTI, the event must have met all of the following criteria: pyuria on urinalysis, culture-proven infection with single organism (≥50,000 CFU/mm3 for catheterized or suprapubic aspirated specimen; ≥100,000 CFU/mm3 for clean voided specimen), fever (≥38°C) or UTI symptoms within 24 hours of urine collection (suprapubic, abdominal, flank pain/tenderness, urgency, frequency, hesitancy, dysuria, foul-smelling urine, or failure-to-thrive, dehydration, hypothermia in infants ≤ 4 months old).

Statistical Analysis and Model Development

Bivariate analyses were performed to compare potential predictors between those with rUTI-associated NRS and those without. We used the Chi-Square test and Fisher’s exact test, as appropriate based on data characteristics and distribution.

Due to low number of NRS events and to account for patient characteristics associated with targeted exposure such as treatment arm (AP vs placebo), a propensity score was built to account for possible confounding effects in the final multivariate model. For treatment arms, patient characteristics were relatively even between AP and placebo group (Appendix 1). However, to ensure minimal residual confounding effect, we built the propensity score for treatment arms with variables including age, sex, race, prior UTI counts/type, BBD, duplication, hydronephrosis, VUR grade, and baseline renal scarring. Similarly, propensity score was developed for rUTI with age, sex, index UTI count, duplication, BBD, and antibiotic prophylaxis. Two multivariate logistic regression model was then fitted: one with treatment arm and the propensity scores as independent variables and rUTI-associated NRS (new scar that occurred in the setting of rUTI) as outcome; the other with rUTI and the propensity scores as independent variables, and all NRS (any new scar, whether rUTI occurred or not) as outcome. Additionally, in order to investigate possible discrepancy in the severity of scarring, we looked at the breakdown of scarring severity by rUTI, as well as NRS grade by treatment arm.

An alpha of 0.05 and 95% confidence intervals were used as criteria for statistical significance. All analyses were performed using SAS 9.4 (SAS Institute, Cary, NC).

RESULTS

Demographics and Cohort Selection

We identified 489 participants (244 with AP; 245 with placebo) with initial DMSA scan performed within 6 months of enrollment as baseline and a follow-up DMSA scan. The overall RIVUR outcomes for this cohort were similar to those reported in the primary RIVUR data, with the findings again showing as significant association of AP with rUTI, and of rUTI with any NRS, but not of AP with any NRS (Table 1). This suggests that the subset is representative of the broader RIVUR cohort.

Table 1.

Associations of the major RIVUR outcomes: recurrent UTI (rUTI) and any new renal scarring (NRS) with antibiotic prophylaxis (AP), and with each other.

	rUTI(%) N=89	No rUTI(%) N=400	p value*
Treatment arm			<0.01
AP	28(11.5%)	216(88.5%)
Placebo	61(24.9%)	184(75.1%)
	Any NRS(%) N=37	No NRS(%) N=452	p value*
Recurrent UTI			<0.01
Yes	18(20.2%)	71(79.8%)
No	19(4.7%)	381(95.3%)
	Any NRS(%) N=37	No NRS(%) N=452	p value*
Treatment arm			0.73
AP	17(7.0%)	227(93.0%)
Placebo	20(8.2%)	225(91.8%)
	rUTI-associated NRS(%) N=18	No rUTI-associated NRS(%) N=471	p value*
Treatment arm			0.06
AP	5(2.0%)	239(98.0%)
Placebo	13(5.3%)	232(94.7%)

^*p-values were generated by Fisher’s exact tests for all categorical comparisons

The general characteristics of the cohort are presented in Table 2 (detailed in Appendix 2–4). Mean age was 20.3 months. Female patients constituted 91% of the overall cohort. Most (92%) had history of a single index UTI. A minority (23%) of patients were toilet-trained at enrollment; of these, 51% had BBD. Approximately half (48%) of patients presented with grade 3–4 VUR. Baseline renal scarring was relatively rare (3.6%). 18% (89/489) had recurrent UTI. Any NRS was found in 7.5% (37/489) patients, 18/489 (3.7%) had rUTI-associated NRS.

Table 2.

Subject characteristics and their association with recurrent UTI (rUTI), any new renal scarring (NRS), and rUTI-associated NRS.

Characteristics	Total(%) N=489	rUTI(%) N=89	p value	Any NRS(%) N=37	p value	rUTI-associated NRS (%) N=18	p value*
Age group			0.01		<0.01		0.04
2–11 months	239(49)	40(45)		11(30)		7 (39)
12–23 months	80(16)	11(12)		6(16)		0
24–35 months	68(14)	8(9)		3(8)		3(17)
36–72 months	103(21)	30(34)		17(46)		8(44)
Gender			0.06		0.35		0.34
Female	447(91)	86(97)		36(97)		17(94)
Male	42(9)	3(3)		1(3)		1(6)
Race			0.12		0.70		0.72
White	391(80)	77(87)		33(89)		16(89)
Black	18(4)	0		0		0
Mixed	33(7)	4(4)		2(5)		0
Other	41(8)	8(9)		2(5)		2(11)
Index UTI			0.03		0.11		0.18
First episode	449(92)	76(85)		31(84)		15(83)
Second episode	40(8)	13(15)		6(16)		3(17)
Index UTI presentation			0.24		0.77		0.85
Febrile	260(53)	22(25)		10(27)		11(61)
Symptomatic	156(32)	13(15)		5(14)		5(28)
Both	73(15)	54(61)		22(59)		2(11)
Bladder & Bowel dysfunction			0.14		0.78		0.27
Yes	54(11)	19(21)		8(22)		6(33)
No	51(10)	11(12)		6(16)		2(11)
Ultrasound result
Duplication	28(6)	8(9)	0.14	5(14)	0.05	2(11)	0.28
Hydronephrosis	23(5)	7(8)	0.16	5(14)	0.02	2(11)	0.16
Highest VUR grade			0.10		<0.01		0.04
1	47(10)	6(7)		4(11)		1(6)
2	201(41)	29(33)		8(22)		4(22)
3	194(40)	39(44)		13(35)		8(44)
4	43(9)	12(13)		12(32)		5(28)
Baseline scarring			0.01		0.52		0.25
None	471(96)	81(91)		36(97)		17(94)
Mild	1(<1)	0		0		0
Moderate	5(1)	2(2)		1(3)		1(6)
Severe	6(1)	4(4)		0		0
Global atrophy	6(1)	2(2)		0		0

^*p-values were generated by Fisher’s exact tests for all categorical comparisons

Any New Renal Scarring and Recurrent Urinary Tract Infection

The univariate associations of these patient characteristics with outcomes are presented in Table 2. Compared with children without rUTI, children with rUTI were older (23.6 months vs 19.6 months, p=0.01), more likely to have had two (vs. one) index UTI’s (14.6 vs 6.8%, p=0.03), and more baseline scarring (8.9 vs 2.5%, p=0.01). Children with any NRS were also more likely to be older (median 26 vs 11 months, p<0.01), have hydronephrosis (14% vs 4%, p=0.02) or grade IV VUR (32% vs 7%, p<0.01).

On multivariate analysis incorporating propensity score, recurrent UTI (OR=4.1(95 CI: 2.0–8.5), p<0.01) remained independently and significantly associated with any NRS after adjustment of age, sex, index UTI count, duplication, bowel-bladder dysfunction, and antibiotic prophylaxis.

Recurrent UTI-Associated New Renal Scarring

Table 2 shows factors associated with rUTI-associated NRS. As with any NRS, rUTI-associated NRS was significantly associated with older age group (median 26 months vs 12 months, p=0.04) and grade 4 VUR (28% vs 8% grade 4, p=0.04). On the other hand, rUTI-associated NRS were not found to be significantly associated with sex, race, index UTI symptoms/presentations, BBD, ultrasound results (hydronephrosis, duplication), and baseline DMSA findings.

In contrast to the effect of AP when looking at any NRS, rUTI-associated NRS was more common in the placebo than in the AP arm, occurring in 5.3% of the placebo and 2% of the AP arm (OR=2.7, 95%CI: 0.9–7.6, p=0.06). After adjusting for age, sex, race, prior UTI counts/type, BBD, duplication, hydronephrosis, VUR grade, and baseline renal scarring, placebo remained associated with higher risk of rUTI-associated NRS (OR=3.0, 95%CI: 1.0–8.8, p=0.04) compared to AP.

There were roughly equal numbers of patients with rUTI-associated NRS (18 (3.7%) patients, of whom 13 were VUR Grade3–4) compared with those who had NRS despite no documented rUTI (19(3.9%) patients, of whom 12 were VUR Grade3–4). The distributions of VUR grades for these two groups were similar (p=0.6). In terms of final NRS grade, the NRS grade were similar between rUTI-associated versus non-rUTI associated NRS (Table 3). Further breakdown by treatment showed more severe (moderate/severe) NRS were distributed similarly between AP and placebo. On the other hand, the placebo group was found to have milder NRS especially those with rUTI (10/14 versus 2/9 in AP, p=0.04).

Table 3.

Breakdown of NRS patients by renal scarring grades in final DMSA

Comparison of final renal scarring grade between NRS patients with or without recurrent UTI
Final Scarring Grade	Mild	Moderate	Severe	Total
NRS with rUTI	12	2	4	18
NRS without documented rUTI	11	6	2	19

Comparison of final rUTI-associated new renal scarring severity between patients who received AP versus placebo
NRS with rUTI (+)
Study End Renal Scarring Grade	Mild	Moderate	Severe	Total	p value*
AP	2	1	2	5	0.3
Placebo	10	1	2	13

Comparison of final new renal scarring severity (without new recurrent UTI) between patients who received AP versus placebo
NRS with rUTI (−)
Study End Renal Scarring Grade	Mild	Moderate	Severe	Total	p value*
AP	7	3	2	12	0.67
Placebo	4	3	0	7

^*p-values were generated by Fisher’s exact tests for all categorical comparisons

DISCUSSION

This study was prompted by the seeming paradox within the RIVUR findings, namely that rUTI was more common among subjects who were on placebo, and NRS was more common among those with rUTI, yet there was no difference in NRS observed between AP and placebo. Why did the strong effect of AP on rUTI not extend to NRS prevention?

One of the most prominent findings in our study was that, when we defined the outcome of interest as NRS specifically associated with rUTI, subjects on AP were significantly less likely than those on placebo to have this outcome. This apparent protective effect of AP on NRS associated with rUTI is consistent with the classic proposed mechanism of RN^{8, 9} However, these findings alone do not fully explain the RIVUR results, and many questions remain. Why did so many subjects develop NRS (mostly in the AP arm) despite the absence of rUTI during the study period? Is there other mechanism, apart from UTI, that contributes to scar formation? Why was the AP protective effect not observed in overall NRS formation despite a very strong association between rUTI and higher NRS rate (with 4.1 times the odds developing NRS if patient had recurrent UTI, and AP was known to reduce rUTI by half by RIVUR report).³

One possibility is that the subjects with NRS without rUTI actually did have rUTI that was undiagnosed. Given the stringent RIVUR protocol and close follow-up this seems unlikely, but must be considered. It is also possible that at least some of the NRS outcomes represent the consequences of insults suffered prior to study entry, but which continued to evolve during the study period. In such cases, the DMSA scan appearance may worsen as progressive inflammation and resultant scarring manifests, despite the absence of any new clinical UTI episodes. While this may explain why some patients without rUTI were found to have NRS, one would expect that such patients would be equally distributed between the AP and placebo arms. Above and beyond these “sterile NRS” cases, one would still expect some NRS to be due to actual rUTI events occurring during the study period, and as these events were significantly more common in the placebo, we would expect that the NRS associated with these events (rUTI-associated NRS) would result in higher overall incidence of NRS in the placebo. Instead, the “sterile NRS” events occurred slightly more frequently in the AP arm, balancing out the increase in rUTI-associated NRS in the placebo arm, resulting in similar incidence of overall NRS. Why these “sterile NRS” events occurred disproportionately in the AP group, and their significance, remains difficult to explain.

Further complicating the picture is the role of congenital renal dysplasia, which can be difficult to distinguish from acquired renal scarring on DMSA scan. We typically think of congenital dysplasia as being associated with high-grade VUR, but is also more common in patients diagnosed prenatally with urinary tract dilation (excluded from RIVUR) and is also associated with diffuse renal scarring.^10–13 In our analysis, we did not observe more cases of global renal scarring in patients with “sterile NRS”, compared to rUTI-associated NRS. Furthermore, congenital dysplasia existing at study entry would have been noted on the baseline DMSA scan, and NRS was based on a change from baseline, diminishing the confounding effect of dysplasia. The overall distribution of NRS severity was similar in whether the NRS was rUTI-associated or “sterile” in our study.

The findings of our study should be viewed in the context of its limitations. Although RIVUR represented the most comprehensive, best-described cohort of such children available, specific characteristics of the cohort warrant consideration. RIVUR enrolled children 2 month-6 years old with heavy female predominance, and most with mild to moderate VUR. Children with congenital anatomical abnormalities were excluded.³ This may impact the generalizability, especially for patients out of the range of the cohort.

Additionally, the power of our study is limited by the cohort sample size and the overall low rate of NRS. RIVUR was not powered to study NRS as a primary endpoint, and thus it was not surprising that NRS was not found to be significantly associated with AP in the initial report. A fair portion (19%, 118/607) of the original RIVUR cohort had DMSA studies with either with inappropriate timing and/or low quality which led to even lower statistical power to differentiate the AP effect on renal scarring.

Moreover, the relatively short study period may preclude detection of new renal scar formation, as come such cases may take more than 2 years to manifest. The “healthy volunteer” phenomenon may also be playing a role; patients who are enrolled in an RCT may be less likely to develop renal scars due to parental/provider vigilance and care at the first signs of UTI. Lastly, RIVUR subjects were recruited after their first or second UTI; given that the risk of renal scarring after only 1–2 UTIs is reported to be quite low,¹⁴ it is not surprising that renal scarring events were uncommon.

CONCLUSIONS

Recurrent UTI was associated with new renal scarring in RIVUR. When limited specifically to new renal scarring associated with rUTI, AP was associated with a decrease in risk of this outcome. It remains unclear why a proportion of RIVUR subjects without rUTI developed new renal scarring. The results of this analysis should be carefully interpreted due to the inherent limitations of being secondary analysis of RIVUR trial.

Key of Definitions for Abbreviations

AP

antibiotic prophylaxis

rUTI

recurrent urinary tract infection

NRS

new renal scarring

DMSA

dimercaptosuccinic acid scan

RIVUR

Randomized Intervention for Children with Vesicoureteral Reflux

RN

reflux nephropathy

BBD

bowel bladder dysfunction