A scoring system for predicting downgrading and resolution of high-grade infant vesicoureteral reflux

1 INTRODUCTION

Vesicoureteral reflux (VUR) is a common finding in a paediatric population and occurs in about 1%-2%, with prevalence inversely proportional to age, as many resolve spontaneously. The natural course of VUR depends on grade, age, sex, concurrent bladder dysfunction and many other factors that have been shown to be negative predictors of the spontaneous resolution of VUR.^1-3 From these known risk factors, a number of studies have suggested models for predicting VUR resolution.^4-7 Most of these studies include children with all grades of VUR and with high-grade VUR in a minority. High-grade VUR showed poor prognosis of resolution in these mixed populations.

Since the overall aim of reducing the risk of recurrent febrile urinary tract infection (UTI) and new renal damage is driving the clinical treatment of VUR patients, risk factors for recurrent febrile infections have also been studied. Algorithms focusing on the risk of UTI for use in the clinical setting have been created. Not surprisingly, the risk factors that were found were much the same as for the non-resolution of VUR: high-grade VUR, age, female sex, reflux at low bladder volume, bladder bowel dysfunction and UTI history.^{8, 9}

Children younger than two years have a higher resolution rate than older children, if all VUR grades are included.^{5, 10} An index for predicting resolution in this group has been suggested, including sex, ureteral anomalies, high-grade VUR and the timing of VUR occurrence as variables.^{6, 7} These studies also included children with high-grade VUR, but it appears that they were censored at the end of follow-up, as they did not resolve or were surgically treated, that is time to event was missing. However, previous studies of infants with high-grade VUR have shown downgrading and spontaneous resolution of reflux, even in this group, but to a lesser extent than for lower grades.¹¹ The downgrading or resolution has mainly been seen in grade 4 and unilateral grade 5 but seldom in bilateral grade 5 in a one-year perspective.¹²

Infants with high-grade VUR have higher morbidity than infants with low-grade VUR, due to the complexity of the disease. In these patients, VUR is often combined with massive dilation of the upper urinary tract, generalised renal damage, concomitant bladder dysfunction and a high risk of UTI. Despite this, there is a subgroup with early resolution, seen already during the first years of life,¹¹ and therefore predicting the probability of VUR resolution in the individual child would be beneficial before deciding on treatments.

In this study, the objective was to develop a pragmatic tool for predicting VUR resolution and downgrading (to grade ≤2) at a given time, in high-grade reflux presenting during infancy. The study aimed to construct a scoring system from variables shown to have significant independent effects on VUR resolution in infants.³ From these variables, we aimed to provide a sum score with cut-off values, selecting patients with a high, intermediate and low probability of spontaneous resolution of reflux and thus to select patients in need of more active VUR treatment.

2 MATERIALS AND METHODS

3 RESULTS

3.2 Prediction of VUR resolution/downgrading to grade ≤2

Applying univariable logistic regression, the variables included in Table 2 were assessed for the prediction of VUR ≤ 2 at the latest follow-up for each child in order to identify parameters appropriate for inclusion in the multivariable models.

TABLE 2. Variables identified in univariate logistic regression analyses as predictors to VUR ≤ 2, and thus appropriate for inclusion in the multivariable models. The best multivariable model selected the variables shown in the table

Variable	Value	n (%) of event	Univariable			Multivariable
			OR (95% CI) VUR = 0,1,2	P-value	Area under ROC-Curve (95% CI)	OR (95% CI) VUR = 0,1,2	P-value
Sex	Boy (1)	24 (37%)
	Girl (2)	3 (13%)	0.24 (0.07-0.90)	.035	0.61 (0.53-0.70)	0.13 (0.02-0.68)	.016
Presentation of VUR	Prenatal US	5 (21%)
	UTI	21 (33%)	1.86 (0.61-5.66)	.28	0.56 (0.46-0.65)
Grade of VUR at inclusion (IV-V)	IV	17 (36%)
	V	10 (24%)	0.55 (0.22-1.39)	.21	0.57 (0.46-0.68)
Uni/Bilateral reflux	Unilateral	11 (44%)
	Bilateral	16 (25%)	0.42 (0.16-1.12)	.084	0.59 (0.48-0.70)
Breakthrough UTI	No (0)	22 (42%)
	Yes (1)	5 (14%)	0.21 (0.07-0.63)	.0055	0.67 (0.57-0.76)	0.21 (0.06-0.71)	.012
No of breakthrough UTIs	0-<1	22 (42%)
	1-<2	2 (10%)
	2-<3	1 (17%)
	3-<4	2 (33%)	0.57 (0.32-1.00)	.049	0.66 (0.55-0.76)
Renal abnormality	No	5 (63%)
	Yes	22 (27%)	0.22 (0.05-1.02)	.052	0.57 (0.49-0.65)
Renal abnormality	None (0)	5 (63%)
	Focal (1)	7 (32%)
	Generalised (2)	15 (25%)	0.52 (0.27-1.02)	.059	0.59 (0.48-0.71)	0.32 (0.13-0.82)	.017
Uni/Bilateralrenal abnormality	None	5 (63%)
	Unilateral	18 (29%)
	Bilateral	4 (21%)	0.43 (0.17-1.07)	.069	0.60 (0.49-0.70)
Bladder dysfunction	No	12 (50%)
	Uncertain	10 (27%)
	Yes	4 (17%)	0.44 (0.22-0.85)	.015	0.66 (0.54-0.78)
Bladder capacity (% of expected/10)	2.9-<10.4	9 (39%)
	10.4-<20	8 (35%)
	20-<30.8	6 (23%)
	30.8-46.7	3 (20%)	0.95 (0.91-1.00)	.063	0.62 (0.49-0.75)
Bladder capacity (%)	<=150	15 (42%)
	>150	11 (22%)	0.39 (0.15-0.99)	.047	0.62 (0.50-0.73)
Post-void Residual urine (% of bladder capacity/10)	0-<1	8 (57%)
	1-<2	8 (33%)
	2-<3	6 (38%)
	3-<4	1 (7%)
	4-9	3 (18%)	0.69 (0.49-0.97)	.031	0.67 (0.54-0.79)
GFR categorized	Normal (0)	20 (42%)
	Subnormal (1)	7 (18%)	0.30 (0.11-0.81)	.017	0.64 (0.54-0.75)	0.22 (0.07-0.71)	.011

The following variables were identified as predictors of VUR ≤ 2: male sex, absence of breakthrough UTI, absence of bladder dysfunction, low bladder capacity (≤150%), low post-void residual urine and normal GFR (≥80% of expected for age). Additionally, unilateral VUR and absence of, or focal renal abnormality showed borderline predictive ability, fulfilling the requirements for entering the stepwise logistic regression (Table 2). From these items, the best multivariable model selected sex, breakthrough UTI, renal damage categorised and GFR categorised (Table 2). In this model, the area under the ROC curve (AUC) was 0.82 (95% CI 0.72-0.92), that is indicating a strong model (Figure 1).

3.3 Score for VUR resolution/downgrading to grade ≤2

Based on the estimates obtained from the selected multivariable logistic model the scoring was developed for each of the four variables resulting in a sum score ranging between 0 and 14 (Table 3). Figure 2 presents the probability curve with confidence intervals for the constructed sum score and describes the algorithm. The AUC for the model including only the score was 0.82 (95% CI 0.72-0.91) (Figure 1).

TABLE 3. Scoring points for each predictor in forecasting the probability of resolution/downgrading to grade ≤2 of infantile high-grade VUR, estimated from data at age one year. Points obtained from estimates (weight of variable) in multivariable logistic regression analyses

Variable
Gender	Male = 0	Female = 4
Breakthrough UTI	No = 0	Yes = 3
Renal damage	None = 0	Focal = 2	Generalised = 4
GFR	Normal = 0	Subnormal = 3

• Abbreviations: GFR, glomerular filtration rate; UTI, urinary tract infection.

The estimated probability for VUR resolution to grade ≤2 ranged between 91.2% (95% CI 69.5%-97.9%) for score 0% to 0.7% (95% CI 0.1%-6.2%) for score 14 (Figure 2).

The sum of scores in those with spontaneous resolution/downgrading to grade ≤2 was a mean of 4.5 (SD 2.67) compared with 7.9 (SD2.73) in the group with persistent VUR (P < .0001), (Table 4). A score of eight points or more resulted in low probability of VUR resolution (≤14% for each score level, mean 6.7% [95% CI 0.8%-22.1%]), (Table 5). Using this cut-off, the sensitivity for correctly identify VUR grade ≥3 was 45.9%, the specificity was 92.6% and positive predictive value (PPV) 93.3%. A high probability of VUR resolution/downgrading to grade ≤2 was suggested for scores of 1-4, with a probability of resolution of 91%-57% for each score level (Table 5). A sum score of 5-7 was thus interpreted as having an intermediate probability of a spontaneous improvement in VUR. The scoring model was internally validated, by using the bootstrapping method, with 5000 sampling studies. The suggested model was selected as the second most common model in 414 (9.0%) of all samples, with median AUC of 0.87 (IQR 0.85-0.90). The most commonly selected model included only suboptimal/optimal GFR and breakthrough infection. It was selected in 495 (10.7%) cases and presented a median AUC of 0.79 (IQR 0.76-0.82).

TABLE 4. Number of patients in the No event (VUR > 2) and Event (VUR ≤ 2) groups with regard to each scoring value (0-14) and with comparisons between scores/patient between the groups

Variable	No event (n = 62)	Event (n = 27)	P-value
Score
0	0 (0%)	3 (11%)
2	0 (0%)	3 (11%)
3	1 (2%)	0 (0%)
4	8 (13%)	10 (37%)
5	3 (5%)	3 (11%)
6	3 (5%)	2 (7%)
7	18 (30%)	4 (15%)
8	6 (10%)	0 (0%)
9	4 (7%)	0 (0%)
10	7 (12%)	1 (4%)
11	7 (12%)	1 (4%)
14	4 (7%)	0 (0%)
Score	7.9 (2.8) 7 (3-14)	4.5 (2.7) 4 (0-11)	<.0001

Note

• For categorical variables, n (%) is presented. For continuous variables, the mean (SD)/median (min-max) is presented.
• Scoring variable = 4*gender [0,1] + 3*breakthrough UTI [0,1] + 2*type of renal abnormality [0,1,2] + 3*GFR categorised [0,1].

TABLE 5. Probability of the spontaneous resolution/downgrading of VUR (grade ≤2) in infantile high-grade VUR, estimated from data at age one year. High probability in scores of ≤4, intermediate probability scores 5-7 and low probability scores ≥8

	Score	P
High	0	.91
	1	.86
	2	.79
	3	.69
	4	.57
Intermediate	5	.44
	6	.32
	7	.22
Low	8	.14
	9	.09
	10	.06
	11	.03
	12	.02
	13	.01
	14	.01

4 DISCUSSION

In this study, we have constructed a scoring system for prediction of spontaneous reflux resolution/downgrading to grade ≤2 in children with high-grade VUR (grades 4 and 5) presenting during the first year of life. To our knowledge, this has not been published for this highly selective group of VUR patients before. There have been several studies suggesting scoring systems for the prediction of reflux resolution with both manual calculations^{6, 7} and algorithms available as online applications,^{4, 5, 22} but most of them include all ages and all grades of VUR.

In the present study, the independent significant predictors included in the scoring system were sex, renal damage (none, focal or generalised), GFR (subnormal or normal) and recurrent UTI before the age of one year. Earlier publications have also suggested these variables to be important for the resolution of high-grade reflux.^1-3 In other scoring systems published, including all grades of VUR; sex and recurrent UTI are variables of importance in most, whereas information on renal damage and function was rarely available for inclusion in analyses.^{4-7, 22}

Renal damage was an important factor in the scoring predicting resolution and downgrading of high-grade VUR in this study. There is general consensus that infants with high-grade VUR, especially if presenting with febrile UTI, should be referred for renal scan investigations, due to the high prevalence of renal damage (in the present study, 91%). In lower VUR grades and in older children, on the other hand, this is not always necessary. The inclusion of renal damage in our scoring system is therefore justified, especially since the absence of renal damage in high-grade VUR children also previously has been shown to predict the resolution of VUR.^1-3 Furthermore, in a study of a cohort of 161 children of different ages and all grades of VUR (only 12% high-grade VUR included), a computational model showed that, when adding renal scan data, the accuracy of the model improved.^{4, 22}

Regarding estimates of renal function as compared to measured GFR, the latter investigations may not be routine at all institutions. If so, estimated GFR calculated from the formula published by Schwartz (see methods) from information on serum creatinine and body length/height can be used.¹³ This estimation is not as exact as assessments with for example 51Cr EDTA, but it is most often sufficient to identify patients with severely impaired renal function. For the proposed score, only information on normal and subnormal (<80% of normal) function is required. Another obstacle is that GFR changes during the first two years and reach adult levels (110 ml/min/1.73 m²) at approximately two years of age. Winberg's logarithm regarding GFR in healthy infants is therefore recommended for the calculation of age-dependent reference values, enclosed in a Table S1 to make calculations easier.¹⁹

It is well known that VUR in general is more common in girls but high-grade infant reflux has a preponderance among males. Sex was also an important factor in predicting reflux resolution in this study, with poorer prognosis in girls with high-grade VUR in this age group. Girls have been shown to have slower rates of resolution, especially in cases of bilateral VUR, both in studies including all ages⁵ and in studies specifically assessing infants with high-grade VUR.¹¹ It can be hypothesised that the slower resolution in girls in this age group is associated with the higher risk of recurrent UTIs. This was supported by the results in the Swedish Reflux Trial.23 In this study, a higher risk of recurrent UTIs and new renal damage was seen in girls, but not in boys, in the surveillance group in children with grades 3-4 VUR at age one to three years. Interestingly, this was mainly seen in girls with non-resolution of reflux during follow-up.²³

Infants with high-grade VUR have a high risk of recurrent UTI despite continuous antibiotic prophylaxis. The frequency of breakthrough UTI in this study was as high as 42% but comparable to other cohorts with a high proportion of uncircumcised infant males. Several studies have reported breakthrough UTI in 46%-59% of children with high-grade VUR^{8, 24, 25} and with higher frequency in girls than boys, especially after the first year.^{25, 26} Infants and children with high-grade VUR also have a higher risk of renal scarring after a febrile UTI.²⁷ In view of the importance of minimising febrile recurrent UTIs in order to reduce the risk of progressive renal damage, two computerised models for predicting breakthrough UTI have been published.^{8, 9} In both studies, high-grade VUR, a history of febrile UTI, bladder bowel dysfunction and female sex were predictors of recurrent febrile infections, which are more or less the same as for the prediction of the non-resolution of VUR. These studies do not, however, specifically deal with high-grade infant VUR but include children with all grades of VUR presented at all ages.

There are several arguments in favour of the rationale for using variables from a cross-section of about one year of age instead of data available at the time of VUR diagnosis. The first is to have complete, comparable investigations at a clear time point, in all the study patients, in order to make the data applicable and helpful for clinicians deciding on further treatment in infants with high-grade VUR.

Secondly, many of the variables with a known association with VUR resolution are not known for the individual patient at the time of VUR diagnosis. Renal scans performed in the early postnatal period are known to have poor quality, due to the physiological low renal function at that time and especially when seen in conjunction with dilated upper urinary tracts.¹⁷ The same holds true for estimated or investigated glomerular filtration rate measurements.

Finally, recurrent infections, despite prophylaxis, are most frequent during the first year^{26, 28} but are only known for a small fraction of patients at the time of VUR diagnosis.

Furthermore, the timing for decision regarding intervention or surveillance at the age of one year can be supported by current urological guidelines^{29, 30} and the management in many institutions. During the first year, antibiotic prophylaxis is the recommended first-line treatment with a conservative attitude towards early anti-reflux procedures, even though some bilateral grade 5 with heavily dilated upper urinary tracts and UTI problems might require early temporary bladder drainage. If breakthrough UTI and/or pre-existing or new renal cortical damage and/or persistent high-grade VUR at the age of 12-24 months that is unlikely to resolve; surgical intervention should be considered according to the recent urological guidelines^{29, 30} and also supported by the treatment algorithm in the frequently used UpToDate® webpage.

Our scoring has the advantage of being easy to use in the individual consultant situation. It has three clear cut-off levels for predicting the chance of spontaneous resolution and downgrading of the reflux: high, intermediate and low probability, related to the values on the 14-point scoring scale. It could be argued that reflux is only one of many factors in the disease, where the focus of all VUR treatment should be preserved renal function. Nevertheless, active treatment of reflux, in addition to the treatment of bladder function, is one of the things that can be done to prevent recurrent UTIs and thus, in the long run, prevent new renal scars. The challenge is to choose those patients who actually benefit from surgical intervention from this perspective. A scoring system predicting VUR resolution in infants could be of use to assist the decision of surgical treatment in symptomatic high-grade VUR even if high scores should not alone be seen as an indication for such decision.

The small number of patients in the study cohort is a limitation. Although it is expected and recommended within the prediction modelling research, no available access to an external cohort to compare with the current research group made the external validation of the model to be missing in this paper. It remains to be performed in the future studies. In addition, a scoring system predicting complete resolution of VUR was not possible to construct since grade 0 was a rare endpoint in this population. Still, we consider the endpoint VUR grade ≤2 as relevant, since these children will not be considered for any further follow-up or treatment if non-symptomatic.

5 CONCLUSION

This model provides a practical tool in the management of infants born with high-grade reflux. Scores ≥8 at one year of age indicate a high risk of persistent reflux and the need for close surveillance or surgical intervention.

CONFLICT OF INTEREST

None of the authors has any conflict of interest.