Critical Application of Adult Nephrolithometric Scoring Systems to Children Undergoing Mini-Percutaneous Nephrolithotomy

Introduction

After introduction of percutaneous nephrolithotomy (PNL) in adults, it took about a decade to be attempted in children. ¹ This delay was due to special concerns in the pediatric population, including the effect of parenchymal violation on a growing kidney and the vulnerability of children for life-threatening complications such as hypothermia, sepsis, and electrolyte imbalance. Moreover, the concurrent advent of extracorporeal shockwave lithotripsy (SWL) as a less invasive technique hindered the initial progress of PNL in children. ² Later on, experience broadening in adults, development of mini-percutaneous nephrolithotomy (MPNL), and unsatisfactory results of SWL augmented the indications of pediatric PNL. ^3,4 Nowadays, PNL has already been established as the primary treatment option for children suffering from staghorn stones, renal stones >20 mm, lower caliceal stones >10 mm, and known cystine or struvite renal stones. ⁵

In the modern era of patient-centered medical care, appropriate preoperative patient counseling mandates the development of efficient predictive tools for the outcome of any surgical intervention. Such tools are also very helpful in procedure planning, learning curve benchmarking, and data standardization for clinical research. Multiple scoring systems were introduced for prediction of PNL outcomes. ^6,7 An interesting observation is that the majority of the predictors composing these scores are different from one system to the other. ⁸ In spite of the numerous reports validating and comparing these nephrolithometric scoring systems (NSS), there is still no consensus regarding the superiority of a certain system to accomplish the target missions. ^6,7

The most popular NSS are the Guy's stone score (GSS), ⁹ the S.T.O.N.E. nephrolithometry, ¹⁰ and the Clinical Research Office of the Endourology Society (CROES) nomogram ¹¹ (Supplementary Appendix). Given that their development depended on data derived from adult populations, in addition to the anatomical and physiological peculiarities of children, doubts can be raised regarding the appropriateness of using these scores in pediatric settings. In this study, we evaluate and compare the ability of the mentioned NSS to predict the outcome of MPNL in children. We also aimed at identifying which of the predictors involved in these systems can separately affect the outcome.

Patients and Methods

During 2015, 2016, and 2017, after the permission of the local committee of medical research ethics, the data of all children younger than 14 years who were prepared for MPNL in our center were recorded prospectively. For patients who had bilateral renal stones, each renal unit had its own record. Preoperative noncontrast computed tomography was done for all patients. Any evident urinary tract infection was treated preoperatively.

Results

This study included 92 renal units in 89 children (Table 1). Stone clearance was achieved after the initial MPNL procedure in 76 (82.6%) renal units. For renal units with residual stones, five had second-look MPNL, six had SWL, and three had both. After these auxiliary procedures, nine more renal units became stone free, making the ultimate stone clearance rate 92.4%. Complications occurred with 19 (20.7%) procedures (Table 2).

Stone multiplicity, presence of a staghorn stone, prior stone treatment, S.T.O.N.E. score for the number of calices involved, and stone burden as a continuous variable and as classified in S.T.O.N.E. nephrolithometry had significant associations with stone clearance after the initial procedure (Tables 3 and 4). For complications, none of the predictors was significant except for stone attenuation as a continuous variable. Meanwhile, stone attenuation as classified in S.T.O.N.E. nephrolithometry (less than vs greater than 950 Hounsfield units) did not show a significant association with complications (Tables 3 and 4).

The three scoring systems had significant associations with stone clearance after the initial procedure; none of them was significantly associated with complications (Table 4). All of them had significant associations with both operative time and blood transfusion (Table 5).

On ROC curve analyses for the three scoring systems as predictors of stone clearance, each of them showed an AUC that was significantly more than 0.5, with the greatest AUC and Youden index for the CROES nomogram (Fig. 1). For complications, no scoring system showed an AUC significantly different from 0.5 (Fig. 2).

OPEN IN VIEWER

FIG. 1.

Receiver operating characteristic curves for the three nephrolithometric scoring systems as predictors of stone clearance after a single session of MPNL in children. Circle markers indicate the criteria associated with Youden indices. The area under the curve (binomial exact 95% CI; p-value) is 0.847 (0.757–0.914; p < 0.001) for CROES nomogram, 0.757 (0.657–0.841; p < 0.001) for Guy's stone score, and 0.694 (0.589–0.786; p = 0.011) for S.T.O.N.E. nephrolithometry. CI, confidence interval; CROES, Clinical Research Office of the Endourology Society; MPNL, mini-percutaneous nephrolithotomy.

OPEN IN VIEWER

FIG. 2.

Receiver operating characteristic curves for the three nephrolithometric scoring systems as predictors of complications of MPNL in children. Circle markers indicate the criteria associated with Youden indices. The area under the curve (binomial exact 95% CI; p-value) is 0.596 (0.489–0.697; p = 0.28) for CROES nomogram, 0.507 (0.401–0.613; p = 0.93) for Guy's stone score, and 0.632 (0.525–0.730; p = 0.10) for S.T.O.N.E. nephrolithometry.

On decision analysis curves (Fig. 3), the greatest net benefit was obtained when the treatment decision was based on the CROES nomogram, starting at a threshold probability for stone clearance as low as 40% and consistently at higher probabilities. On the contrary, S.T.O.N.E. nephrolithometry demonstrated an inconsistent curve that lies below the treat-all curve in the threshold probability interval between 60% and 70%.

OPEN IN VIEWER

FIG. 3.

Decision curves for the three nephrolithometric scoring systems comparing the net benefit of using the scores in children to make the decision of MPNL based on the probability of stone clearance.

Discussion

Despite the cautious introduction of PNL in children, it passed the test of time over the past three decades and showed high safety and efficacy with no age limit and an insignificant adverse effect on renal function. ^{13 –16} The MPNL became favored in children after achieving an efficacy similar to that of using adult-size instruments, with fewer complications. ^14,17 Although the NSS have been extensively used, studied, and compared in adult populations, ^6,7 the data about their applicability in children are limited and entirely based on retrospective studies. In agreement with our results, they demonstrated the ability of the three NSS to predict stone clearance, but they reported conflicting results regarding prediction of complications. ^{18 –21}

Only two studies were similar to the present work as regards comparing the three NSS in children. ^22,23 Contrary to our results, Aldaqadossi et al. ²² found that S.T.O.N.E. nephrolithometry had the highest accuracy for prediction of stone clearance. Also, they found that GSS was associated with complications. These differences may be due to variation in the scoring process that was not clearly delineated in their study. For example, they excluded patients with previous open stone surgery, which is one of the components of the CROES nomogram and was found to affect stone clearance in our cohort.

Recently, Çitamak et al. ²³ applied the three NSS on a large dataset and found that the CROES nomogram was the best to predict both success and complications. They also proposed a genuine simple scoring system for prediction of outcomes of PNL in children: the stone-kidney size score. It depends on the number of stones and the relationship between stone burden and renal length. Their new score achieved reasonable performance in prediction of success but questionable ability to predict complications. This score categorizes patients into just three grades that cannot accommodate the whole spectrum of renal stone complexities. Moreover, it still needs further evaluation as it was derived from a retrospective dataset extending over 20 years and involving heterogeneous imaging methods, surgical techniques, and instrument sizes.

The development of GSS was based on a literature review, expert opinions, and a plan/do/study/act process. ⁹ Oddly, it did not include a direct evaluation of stone burden in spite of its proven effect on the outcome of PNL. ^24,25 Of the predictors involved in this score, stone multiplicity and presence of a staghorn stone were found to affect stone clearance in our study. Our cohort did not include any case with a caliceal diverticulum, spina bifida, or spinal injury, so these items were not evaluated. The GSS was unique to involve the abnormal renal anatomy among its predictors. However, many studies denied the effect of renal anatomical abnormalities on the outcome of PNL. ^26,27 Moreover, the absence of concrete definitions for an abnormal renal anatomy and a partial staghorn stone threatens the reproducibility of the score, as mentioned by its creators. ⁹

S.T.O.N.E. nephrolithometry contains five predictors that were determined after a systematic review of the literature and a retrospective pilot study. ¹⁰ Two predictors were found to affect stone clearance in our study: stone burden and S.T.O.N.E. score for the number of calices involved, which also entails the presence of a staghorn stone. However, the indeterminate description of the method of counting the involved calices weakens its reliability. ²⁸ The other three predictors are exclusive for this score and essentially did not show a significant effect on stone clearance even in the initial report of the score. ¹⁰ Although there are some conflicting reports about the adverse effects of obesity and a long stone-to-skin distance on the outcome of PNL in adults, ^25,29 it is of negligible importance in children due to the constantly short stone-to-skin distance.

The CROES nomogram was derived through a multivariate logistic regression analysis of multi-institutional data of 2806 patients. ¹¹ This can be considered the most powerful and rigorous method in comparison with the other scoring systems. All of the predictors included in this score showed a significant effect on stone clearance in our study except for stone location. This can be attributed to the dealing with multiple stone locations as a single entity without discrimination according to the number and sites of involved calices. The main concern against this score is its complexity, which may hinder its clinical use. This can be solved by developing a software application that directly calculates the chance of stone clearance.

Although morbidity is not less important than efficacy as a surgical outcome, all of the three NSS share the same weakness and inconsistency in predicting complications. ⁶ In our study, stone attenuation was the only predictor that was significantly associated with complications. This can be explained by the need of harder stones for more manipulations during disintegration. In a large-scale study, Onal et al. ¹⁴ found that operative time, sheath size, a midcaliceal access, and a partial staghorn stone are the independent factors influencing complications of PNL in children. Most of these factors are intraoperative and so cannot be included in a preoperative score.

Another confusing issue in the assessment of renal stone complexity is the exact description of a staghorn stone. Attaining variable volumetric and morphological criteria, staghorn stones cannot be dealt with as a one-grade item to the extent that Mishra et al. ³⁰ developed a specific classification system to assess their complexity.

As these NSS were developed primarily for adults, they did not take age into consideration. Although we did not find significant associations between age and MPNL outcomes, sub-stratification of pediatric patients according to age may reveal a substantial difference in the performance of NSS among different pediatric age groups.

In this study, the sample size was inadequate for a multivariate analysis. A meta-analysis, future extension of the study, or multi-institutional cooperation may yield a more powerful sample size. Another limitation in our study is that we did not assess the interobserver variability for the proposed NSS.

Conclusions

The CROES nomogram is the best among the studied NSS to predict single-session stone clearance of MPNL in children. More accuracy may be achieved after modification of the stone location item and grading of the staghorn stone item. None of the studied NSS is efficient regarding prediction of complications. A thoughtful eye is needed to put the edges of this puzzle together and invent a comprehensive scoring system for prediction of the outcome of MPNL in children.