The diagnostic accuracy and clinical utility of pediatric renal tumor biopsy: Report of the UK experience in the SIOP UK WT 2001 trial

The International Society of Paediatric Oncology (SIOP) protocols recommend preoperative chemotherapy appropriate for Wilms tumors (WTs) in children with renal tumors aged ≥6 months, reserving biopsy for “atypical” cases. The Children's Cancer and Leukaemia Group (CCLG) joined the SIOP‐WT‐2001 study but continued the national practice of biopsy at presentation.


INTRODUCTION
Renal tumors represent 7% of pediatric cancers, with Wilms tumor (WT) the most common, accounting for about 90% of renal cancers in European Cancer registries. 1,2 WT has a median age of diagnosis of 3 years and generally has a good prognosis with 5-year overall age, MN represents fewer than 10% of renal tumors, 4

and in total 75%
all MNs in the UK occur under the age of 6 months. 5 CCSK has a similar age distribution to WT but requires more intensive chemotherapy to achieve similar 5 year outcomes compared with WT. 6,7 RTK typically presents in the first 1-2 years and has a considerably poorer prognosis, despite more intensive chemotherapy, with median survival less than 1 year. 8 RCC typically presents in adolescence, is chemotherapy resistant, and requires effective surgical clearance, with molecular targeted therapy used in appropriate cases.
There are two philosophies for the initial management of pediatric renal tumors. The Children's Oncology Group (COG) protocols advocate upfront nephrectomy to achieve an accurate tissue diagnosis prior to postoperative chemotherapy. The International Society of Paediatric Oncology (SIOP) Renal Tumour Study Group (SIOP-RTSG) protocols use preoperative chemotherapy to minimize the risk of preoperative tumor rupture, give a more favorable stage distribution, and to allow histological response to chemotherapy to be included in risk stratification of postoperative treatment. 9,10 Upfront nephrectomy is advocated only for children under 6 months of age where there is a much higher incidence of tumors that do not require cytotoxic treatment. There are no significant differences in overall survival between concurrent COG and SIOP renal tumor trials. 3 The UK and Irish Children's Cancer and Leukaemia Group (CCLG) changed from immediate nephrectomy to a preoperative chemotherapy approach when they joined the SIOP WT 2001 trial after the UKW3 trial compared the two approaches. 11 However, we continued using percutaneous biopsy (PCNB) prior to commencing chemotherapy in all cases to reduce the risk of giving inappropriate chemotherapy to children with non-WTs. For other countries participating in the SIOP WT 2001 trial, biopsy was not recommended routinely but only permitted when there was clinical diagnostic uncertainty.
Biopsy carries risks that may not outweigh the clinical benefits of identifying non-WTs, especially in the age range where WT is the most likely diagnosis. The largest study of pediatric biopsy associated morbidity in 182 cases enrolled in UKW3 found 19% of patients suffered pain, 7% infection, and 5% postprocedure bleeding. 12 A further concern is that biopsy poses a risk of local tract seeding, but only a few cases are reported in the literature. 13,14 Large adult series (>10,000 patients) suggest needle track recurrence occurs in <0.01% cases 15 compared with 0.5% in the above pediatric series. 12 In the UKW3 trial, biopsy was not significantly associated with risk of local relapse in multivariate analysis. 16 In this study, we analyzed the diagnostic accuracy of locally reported biopsies compared to the central pathology review (CPR) nephrectomy diagnoses, and assessed the implications of any discordancy for the clinical effectiveness of a policy of universal renal biopsy for pediatric renal tumors by age group.

METHODS
A retrospective study of children enrolled through the CCLG in the SIOP WT 2001 trial. 10  was recommended using a 12-14G cutting needle under ultrasound guidance and with consideration of using a coaxial technique.
The pathology reports of renal tumor biopsies were obtained retrospectively from participating CCLG centers by requesting to provide all biopsy reports for the children they had enrolled, and trial case files of G. Vujanic (Chair of the CCLG Pathology Panel) for biopsies that were sent for CPR (note that biopsies were not required to be sent routinely for CPR). Biopsies were reported locally prior to nephrectomy.
Each local pathology report was coded by a nonpathologist (TJJ) into one of 15 categories (Supporting Information File S1) and the type of the biopsy (fine needle aspirate, PCNB or open) was recorded based on the description in the report. Five of the categories were considered as types of "nondiagnostic": "normal renal tissue," "necrotic sample," "insufficient sample," "indeterminate diagnosis," and "indeterminate diagnosis-not a WT." A case was classed as "indeterminate" if the pathologist offered a differential diagnosis or stated they were not confident of a single diagnosis. The full local pathology report of all cases classified as "indeterminate diagnosis" was reviewed by an expert pediatric pathologist (GMV). Groups A-C cases with a biopsy that was nondiagnostic or indeterminate and included WT in the differential as false negatives and the worst case specificity as treating nondiagnostic/indeterminate biopsies in Group D cases as false positives: the analysis used a modified version of the "intention to diagnose" approach 18 based on the intention to discriminate between WT and non-WT. Overall concordance between first biopsy and subsequent nephrectomy diagnosis was assessed. A more detailed assessment within each clinically relevant group was then performed as some discordant results would not result in incorrect management.
Subgroup analyses were performed for three specific age groups: the first were children under the age of 6 months, for whom upfront nephrectomy is already advised, the second between 6 months and 9 years and 11 months, and the third group children aged 10 years or over, where the relative incidence of non-WTs starts to increase.
Analyses were performed in the R statistical environment 19 with graphical output produced using ggplot2. 20 Exact binomial 95% confidence intervals (CI) were calculated for sensitivity, specificity, nondiagnostic, and accuracy estimates.

Inclusion
There were 849 cases enrolled by 25 centers ( showing a similar distribution (Supporting Information Figure 2).
There was no significant difference in the proportion of non-WTs for cases where a biopsy report was obtained compared to where it was not (49/548 vs 12/163, Fisher's exact test, P = 0.633)

Type of biopsy
Of the 552 initial biopsies, 541 were PCNB, 10 open, and in one case the type of biopsy could not be determined from the pathology report and was missing from the registration case report form.
F I G U R E 1 Stacked histogram for the age at presentation for all unilateral cases. Bin width = 2 months

Sensitivity analyses
The overall sensitivity of biopsy to identify tumors that needed alter-  (Figure 2).

Concordance
The concordance between biopsy and nephrectomy is detailed in Sup- One of the Groups A-C cases with a discordant biopsy was an anaplastic sarcoma of the kidney, which at the time of presentation was not a The concordance between biopsy diagnosis and nephrectomy based on clinically relevant groups is summarized in Table 2

Subgroup analyses
Cases were subgrouped based on the different age distributions of tumors. In the age range at which WT is commonest (

Clinical implications
Overall, 37 of 548 (6.7%) children had a non-WT and the biopsy identified a diagnosis that required a change from the standard SIOP preoperative chemotherapy management ("clinical effectiveness"). Furthermore, six cases of non-WT were not detected by biopsy and 28 cases of WT/NR had nondiagnostic biopsies ( F I G U R E 3 3 × 2 contingency table for the sensitivity and specificity of biopsy to distinguish "non-Group D tumors" in children aged greater than or equal to 6 months but less than 10 years. Format is the same as shown in Figure 2

DISCUSSION
This is the largest study of the diagnostic accuracy of pediatric renal tumor biopsy. In addition to calculating estimates of concordance and nondiagnostic rates, we also report on how effective biopsy (per-formed by PCNB in 98% of cases) is in identifying tumors that require a change in management from the standard SIOP approach.
The overall concordance between biopsy and final nephrectomy (91.7%) is slightly lower than the 94% figure found in Vujanic et al's study from UKW3. 12 However, Vujanic et al's study only included cases where both the biopsy and nephrectomy were sent for CPR and so may represent a biased subgroup analysis. This suggests that where biopsy is performed, CPR may help improve diagnostic accuracy, as is the case for nephrectomy specimens. 21 We found that 6.5% of all biopsies were nondiagnostic, which is slightly higher than the 4% figure found in the UKW3 study, 12 but not dissimilar to a pooled estimate of 6% for biopsy inadequacy from a systematic review of PCNB of pediatric tumors. 22 Importantly, we report the rates of different types of nondiagnosis. "Invalid" nondiagnostic results, whereby the test procedure fails to provide material required for diagnosis ("insufficient sample," "wholly necrotic tissue," "normal renal tissue"), are different from valid nondiagnostic results (where the procedure was adequate but there was diagnostic uncertainty-an "indeterminate result") as the latter may provide some clinically useful information (e.g., this is a malignant neoplasm that is not a WT, but it is not clear which non-WT). 23 While PCNB alone may lead to unrepresentative material (an invalid non-diagnostic result), this risk is minimized by using an image-guided coaxial technique to sample material from the most suspicious areas of the tumor. 24 Note that 36% of the nondiagnostic results in this study were indeterminate results. It is unclear whether this was because pathologists were less certain of the diagnosis or more prepared to admit diagnostic uncertainty in SIOP UK 2001 than UKW3 or whether the different researchers in each paper classified pathology reports differently.
For example, the degree of diagnostic certainty associated with identical phraseology can be interpreted differently between surgeons and pathologists. 25 The SIOP approach is to treat all pediatric renal tumors (unless they are completely cystic on imaging studies or occur in children under 6 months of age or clinical features are atypical) with empirical preoperative chemotherapy appropriate for the most likely diagnosis of WT.
Hence, the potential utility of biopsy is to correctly identify non-WTs Groups are as described in Table 1, according to how a correct biopsy diagnosis could have changed management.
to permit an early change in management. However, the sensitivity of biopsy to identify these tumors is considerably poorer than the specificity to avoid misdiagnosis of WT. There were also higher levels of diagnostic uncertainty with non-WT cases. Identification of high-risk tumors that require chemotherapy other than CCSK was particularly poor (although these cases, such as RTK and DSRCT, are a minority of all high-risk tumors).
Our data suggest that overall biopsy would be expected to correctly change management in 6.7% cases. If we confine the analysis to the 6-119 months age group and include CCSK in the group that are considered to be adequately treated by vincristine and actinomycin-D preoperative chemotherapy, this drops to six of 518 (1.2%) cases.
This is a reasonable analysis as although there are no trials comparing the different regimens for CCSK, there are no survival differences between the UK (where the CCLG advocates AVD preoperatively if a CSSK is identified) and Germany, where biopsy is rarely employed and thus an AV regimen is de facto used or between SIOP and NWTS/COG approaches. 26 Ultimately, the percentage of cases in which biopsy will change management is lower than the incidence of procedural complications previously reported. 12 Our data suggest that biopsy in children over 6 months is not effective in avoiding unnecessary chemotherapy in children with benign tumors but it is clearly effective in identifying RCCs (an important diagnosis to make to avoid unnecessary chemotherapy and allow timely surgical resection). Based on European population based registry data, RCC accounts for 34% of all renal cancers in 10-14 years old compared with 4.3% in 5-9 years old but WT is still the most common diagnosis in both age groups. 2 We find using a cut-off of 10 years and above increases the clinical effectiveness of biopsy to 26%, with all cases of RCC being correctly identified. Comparing the distribution of histological diagnoses of cases registered in the trial with the national cancer registry of England confirms the suitability of this age threshold for our population (Supporting Information Figure 3). Furthermore, recent reviews of the use of PCNB in adult RCC suggest that this technique is safe and provides adequate diagnostic information to guide subsequent surgical management. 27 There are some limitations of this study. First, we achieved partial capture of the intended cohort as some centers did not respond to requests to provide biopsy reports. Some patients from nonresponding centers were included as they also had biopsies sent for CPR. If pathologists were more likely to send diagnostically challenging cases for CPR, this may have increased the nondiagnostic rate. Second, this study was confined to cases recruited to a trial focused on WT, hence benign tumors may be underrepresented and some tumor types that also occur in nonrenal sites were registered in other trials. As such, the estimates of concordance are not fully generalizable to all pediatric renal tumors. Third, changes in pathological practice over the last 15 years, such as the use of specific immunohistochemistry and molecular biology techniques to distinguish between different non-WTs, may mean that the nondiagnostic and discordance rates may be overestimated compared to current practice. As the use of CPR is associated with increased diagnostic accuracy, 21 our estimates cannot be applied to situations where biopsies are subject to CPR. We used the local biopsy reports as it was not part of the trial protocol for biopsies to be sent routinely for CPR (unlike all nephrectomies). Some biopsies were sent for CPR concurrently with the nephrectomy specimens and so were not blinded. Fourth, our estimates for clinical effectiveness are inferred from the biopsy data rather than the actual treatment decisions made locally. In practice, a biopsy result is always taken in the context of imaging results and the child's clinical picture. Finally, our results may not be generalizable to contexts where biopsy is used more selectively.
In summary, this study provides estimates for the diagnostic accu-