Risk stratification has become increasingly complex since the first NWTSG study, NWTS-1 (Table 2).

Evolution of risk stratification

In NWTS-1 and NWTS-2, patients were only stratified by stage17,18. In the subsequent two studies, NWTS-3 and NWTS-4, tumour histology (favourable or unfavourable) was incorporated into stratified treatment19,20. Finally, in NWTS-5, two new variables, patient age and tumour nephrectomy weight (TNW), were added to stratification5,21. The results of these five NWTSG studies established the clinical and biological features employed for risk stratification in the initial series of COG ‘AREN’ therapeutic studies16 (Table 1). These features include stage, histology (including post-chemotherapy histology for patients with bilateral or predisposed Wilms tumour), age, TNW, loss of heterozygosity (LOH) of 1p and 16q, response of pulmonary metastases to chemotherapy and the presence of extrapulmonary metastases. Chemotherapy regimens studied and currently used in risk-adapted, first-line treatment of FHWT include EE-4A, DD-4A, VAD, regimen I and regimen M (Table 3).

Table 3 Children’s Oncology Group chemotherapy regimens

Evidence from previously conducted studies has informed the newly developed recommendations regarding the prognostic factors included in risk stratification, including prospectively studied factors, retrospectively studied factors and factors of less certain importance.

Prospectively studied prognostic factors

The prognostic features of stage, histology, age, TNW, LOH of 1p and 16q, response of pulmonary metastases to chemotherapy and the presence of extrapulmonary metastases have all been previously studied in prospective NWTSG or COG therapeutic studies, which has informed their use in the updated model.

Stage

In NWTSG and COG, FHWT is staged using a combination of radiological and surgicopathological staging (Table 4). Cross-sectional imaging of the chest, abdomen and pelvis is required for all patients, to determine the resectability of the tumour and for detection of metastatic and bilateral disease. Local (abdominal) stage is determined using surgicopathological staging from an initial diagnostic procedure. In the first NWTSG studies, patients with metastatic disease clearly had worse outcomes than those with localized disease treated with the same chemotherapy, although the study was not designed to demonstrate this difference17. Observations from subsequent studies have enabled narrowing of the outcome gap between patients with stages I and IV disease by use of intensified therapy strategies for increased stages8,9,11,22,23,24. Notably, either chest X-ray or CT scans were accepted for diagnosis of pulmonary metastases in the NWTS studies22. Results of the SIOP 2001 study showed that patients with CT-only lung nodules (nodules visible on CT but not visible on chest X-ray) had better outcomes than those with lung nodules visible on chest X-ray and worse outcomes than those with localized disease25. In NWTS study results, patients with CT-only lung nodules had improved event-free survival (EFS) when treated with intensified chemotherapy, thereby supporting prognostic value and enhancing risk stratification; therefore, CT scans are now required for staging22.

Table 4 Children’s Oncology Group staging of Wilms tumour

Patients with stage V (bilateral) FHWT have historically experienced poor outcomes26. In AREN0534, the first prospective study involving these patients, an intensified neoadjuvant chemotherapy regimen was adopted and delayed nephrectomy histology was incorporated to determine the subsequent treatment regimen27. Markedly improved outcomes (4-year EFS 84.2%) were achieved using this approach compared with patients treated using NWTSG study approaches (4-year EFS 65%)27. Except when explicitly stated, throughout this manuscript the discussion of features incorporated into risk stratification pertain to unilateral FHWT. The influence of features such as LOH of 1p and 16q and lung metastatic response to chemotherapy on bilateral FHWT is currently uncertain but is an area of active investigation.

Subtle changes to staging definitions have occurred over time; therefore, stage shifting needs to be considered. For example, in NWTSG studies, intraoperative local tumour spill that was confined to the flank, which also applied to the rare patients who had a biopsy and then subsequently underwent complete nephrectomy before starting chemotherapy, was not considered an indication for stage III designation11. However, in COG studies, any biopsy or local tumour spill was designated as stage III, after recognizing that patients who had local tumour spill but otherwise met stage II criteria experienced an increased risk of local recurrence28,29,30 (Table 4). The terms ‘spill’ and ‘rupture’ have historically been inconsistently defined and used, and at times conflated in previous studies and analyses; thus, ‘spill’ will be clearly defined in future COG studies to be an intraoperative event involving tumour capsule disruption at the time of surgery (including biopsy), whereas ‘rupture’ will be defined as a preoperative event leading to tumour capsule disruption, determined either intraoperatively by the surgeon or identified by a pathologist. Because ‘rupture’ (regardless of symptoms or imaging findings) is an indication for whole-abdomen irradiation (WAI), this distinction is important for accurate risk stratification and therapy decisions. This distinction will be further emphasized in future COG studies to improve data collection and treatment decisions. Additionally, with widespread availability of CT scans and improved outcomes with detection of CT-only lung nodules, COG requires chest CTs for accurate staging.

Lymph-node sampling (LNS) is important for accurate staging. The presence of tumour within abdominal lymph nodes confers a local stage III designation and is a predictor of EFS and overall survival (OS)11,24. In NWTSG and COG staging, the finding of a non-viable tumour within a lymph node is considered lymph-node involvement. Enlarged abdominal lymph nodes on imaging are well established to be frequently reactive rather than involved with tumour; thus, surgical sampling is required for accurate staging31. In a Surveillance, Epidemiology and End Results and Florida Cancer Data System study, survival was observed to be lower for patients who did not undergo LNS than for those who did (5-year OS 87%) versus 1–5 (91%); 6–10 (93%); or >10 (95%) lymph nodes sampled (P = 0.005). A survival advantage for patients having 1–5 lymph nodes (hazard ratio (HR) 0.6, P = 0.016), 6–10 lymph nodes (HR 0.521, P = 0.048), and >10 lymph nodes (HR 0.403, P = 0.039) sampled compared with patients with zero lymph nodes examined was shown on multivariate analysis32. In NWTS-5, failure to sample lymph nodes was associated with an increased risk of relapse in patients with stages I or II disease, suggesting that some patients had undetected lymph node involvement (that is, really had stage III disease) owing to the absence of LNS33. Similarly, in AREN0532, a non-statistically significantly reduced EFS was observed in patients who did not have LNS, with 4-year EFS of 84% among patients with stage III disease without LNS (n = 148) versus 89% (n = 387, P = 0.067) in those with LNS11. A combined analysis of AREN03B2 and AREN0532 showed improved outcomes for patients with stage III disease who had LNS (4-year EFS 90.3%) relative to those without LNS (EFS 80.0%, P = 0.0037)34.

A review of the National Cancer Database showed that lymph nodes are not sampled in 10–15% of patients35, and in AREN03B2, failure to sample lymph nodes was the most common surgical protocol violation36. The likelihood of finding a positive lymph node increases with increasing number of lymph nodes sampled37,38, en bloc sampling increases the number of lymph nodes sampled39, and sampling between 6 and 10 lymph nodes decreases the false-negative rate to <10%35. Formal lymph-node dissection (such as retroperitoneal lymph-node dissection) is not necessary; however LNS is required for accurate staging. In previous NWTSG and COG Wilms tumour studies, patients were staged and enrolled without LNS as if their lymph nodes were negative, meaning that a patient could be assigned stages I or II without examination of an uninvolved lymph node40. For future COG unilateral FHWT trials, patients will be ineligible if LNS does not occur. Any patient either ineligible for enrolment (upfront nephrectomy without LNS) or removed (enrolled, delayed nephrectomy without LNS) will receive whatever treatment their treating institution or team recommends. The COG is a research organization and does not provide guidance regarding individual patient management outside the confines of research protocols.

Histology

The presence of anaplasia was first recognized as an adverse prognostic factor owing to the results of NWTS-1 (ref. 41). It was subsequently shown to indicate chemotherapy resistance and tumour aggressiveness42, and incorporated into risk stratification. Tumours with anaplasia were classified as unfavourable histology in NWTS-3 and subsequent studies19. Identification of anaplasia, and its classification as focal or diffuse43, can be a challenge. Tumour heterogeneity means that anaplasia is often not discovered on biopsy but is identified subsequently, when the entire tumour is removed42,44, which is one reason among others why the COG advocates for upfront nephrectomy. Additionally, use of central pathology review has identified that many patients with diffuse anaplasia are not recognized as having anaplasia on institutional pathology review42,45, which highlights the value of pathological expertise in making this diagnosis. This observation is a key factor that prompted the requirement for centralized pathology review for eligibility in COG Wilms tumour trials.

The COG integrates post-chemotherapy histology into risk stratification for patients with bilateral FHWT, treating these patients with neoadjuvant chemotherapy and resection, followed by adjuvant treatment based on stage and histological classification, which has been adapted from the SIOP classification but is not identical to it2,46,47,48. In the upcoming unilateral FHWT trial, histological types that affect potential de-escalation or escalation of therapy will include upfront epithelial Wilms tumour, and post-chemotherapy completely necrotic or blastemal-predominant Wilms tumour.

Age

In NWTS-1, age <2 years was identified as a favourable prognostic factor in a subset of patients with stage I disease17, although the results of NWTS-2 suggested that the inferior outcomes observed among patients of increased age were attributable to advanced stage or anaplastic histology18. In one United Kingdom Childhood Cancer Study Group (UKCCSG) analysis, increased age was associated with poor survival in those with stage I disease when treated with vincristine only49, with results of other UK studies showing age >4 years to be an independent risk factor49,50. Age has not been shown to be significant in other multivariate analyses30,51,52. In SIOP–RTSG studies, age cohorts with cut-off points of 2, 4 or 10 years of age were associated with inferior EFS compared with age 6 months to 2 years in multivariate analyses; however, OS was not significantly different except in patients >4 years50. In NWTS-4, age >4 years lost significance as an adverse prognostic factor when adjusted for histology and lymph-node involvement53. For patients with stage I FHWT enrolled in AREN03B2, no association with age and EFS was demonstrated54.

Age has not been used in risk stratification for patients with high-stage FHWT. Some data indicate that substantially older (≥15 years) patients with Wilms tumour (‘adult Wilms tumour’ or ‘adolescent and adult Wilms tumour’) have poor outcomes55,56; however, the reasons for this observation are unclear. Hypotheses for differential outcomes include possible differences in tumour genetics or treatment tolerance55,57,58,59.

Tumour nephrectomy weight

The combination of age <2 years and TNW (the weight of the nephrectomy specimen including tumour and kidney) <550 g was identified as a favourable characteristic of FHWT in the 1970s60,61. Increasing intensity of treatment did not improve the outcome of patients with these characteristics; therefore, chemotherapy might not be necessary61. An initial study including eight patients with these characteristics treated without chemotherapy resulted in only one recurrence, which occurred as a metachronous tumour in the contralateral kidney in a child with a genitourinary anomaly, raising the possibility of a genetic predisposition (new primary tumour rather than relapse)62. Analyses of NWTSG studies suggested that the risks of adjuvant chemotherapy might outweigh the benefits for this subset of patients63,64, subsequently termed having very-low-risk (VLR) Wilms tumour.

Validation of a very-low-risk subgroup

In NWTS-5 patients with VLR Wilms tumour were hypothesized to maintain excellent outcomes without chemotherapy21. Results showed that 5-year EFS was 84% and 5-year OS was 98% among 77 patients with VLR disease treated initially with surgery alone65. The study was closed when the EFS fell below 85% meeting prespecified study closure parameters; however, because OS remained high, the strategy of surgery only was further studied in AREN0532, in which 116 patients with VLR disease (requiring real-time central review confirming negative lymph nodes, and lack of a predisposition syndrome or radiological contralateral nephrogenic rests) were enrolled7. The results of this study demonstrated excellent results, with 4-year EFS of 89.7% and 4-year OS of 100%7.

Loss of heterozygosity of 1p and 16q

Analysis of NWTS-3 and NWTS-4 studies showed that LOH of 1p or 16q, present in 12% and 17% of patients with FHWT, respectively, was associated with reduced relapse-free survival and OS66. In NWTS-5, the hypothesis that LOH at these loci was associated with a poor prognosis was prospectively tested5. Risk of relapse and death was increased with either, and the worst outcomes occurred in patients with combined LOH of both 1p and 16q (henceforth referred to as ‘combined LOH’)5. In AREN0532 and AREN0533, intensified therapy for patients with combined LOH was prospectively studied, improving survival to a 4-year EFS of 87.3% for stages I or II (versus 68.8% in NWTS-5, P = 0.042), and 90.2% for stages III or IV (versus 61.3% in NWTS-5, P = 0.001)9.

Lung metastatic response to chemotherapy

In NWTSG studies, all patients with pulmonary metastases identified on chest X-ray were treated with whole-lung irradiation (WLI), whereas radiotherapy for CT-only lung metastases was left to the discretion of the treating institution22. Differential outcomes for patients with complete versus incomplete response of lung nodules to an initial 6 weeks of chemotherapy were first identified in SIOP studies67. In AREN0533, de-intensification of therapy (continued DD-4A with omission of WLI) for patients with lung-only metastases with rapid complete response (RCR) of pulmonary nodules to two cycles of chemotherapy, and intensification (regimen M with WLI) for those whose pulmonary disease had a slow incomplete response (SIR) after two cycles, were prospectively studied. Patients with RCR had a 4-year EFS of 79.5% (versus an expected 85% based on NWTS-5) and an OS of 96%; and those with SIR had an EFS of 90% (versus an expected 75%) and OS of 96%8. 1q status had a substantial effect on the EFS of patients with RCR, but not those with SIR.

Extrapulmonary metastases

Most patients with stage IV FHWT present with pulmonary metastases alone, but others present with extrapulmonary metastases with or without lung involvement, with liver being the most common extrapulmonary metastatic site68. In NWTS-4 and NWTS-5, there was no significant difference in EFS or OS between patients with stage IV FHWT with liver (with or without lung metastases) (n = 96) and those with stage IV FHWT with lung-only metastases (n = 513)68. In AREN0533, patients with extrapulmonary metastases were assigned to chemotherapy with regimen M, intensified from DD-4A received in NWTS-5. In a COG analysis in which patients with extrapulmonary metastases from NWTS-5, AREN0533 and AREN03B2 were pooled, outcomes were inferior for patients with extrapulmonary metastases (observed 4-year EFS of 77.3%)69, compared with those with lung-only stage IV (EFS 85.4%) treated using the AREN0533 treatment strategy8. No statistical differences in EFS or OS were found between patients with extrapulmonary metastases treated in AREN0533 compared with those in NWTS-5, but the small cohort sizes, heterogeneous patient characteristics and metastatic sites, flawed data collection, and lack of consistent local control management confounded assessment of the role of regimen M; thus, the optimal chemotherapy for patients with extrapulmonary metastases remains uncertain69. The role and quality of evidence supporting local treatments of sites of extrapulmonary metastases vary by metastatic site. Radiotherapy is recommended for nearly all metastatic sites, but the role of surgery for extrapulmonary metastases is not certain, and has not been strictly prescribed or reported in past studies.

Apart from metastatic sites of the liver, brain and bones, the specifics of what defines extrapulmonary metastasis have not been clearly established. For the purposes of AREN03B2 and AREN0533, certain findings (such as pulmonary tumour emboli, malignant pleural effusions and extra-abdominal lymph nodes (cervical and intrathoracic or mediastinal)), when identified at institutional review, were consistently designated as extrapulmonary metastases, a decision made by consensus opinion of the central reviewers, and not based on data. Notably, peritoneal implants or pelvic tumours identified at diagnosis, presumed to be a result of local ‘drop mets’ from tumour rupture rather than haematogenous spread, are not considered sites of extrapulmonary metastases.

Retrospectively studied prognostic factors

Some prognostic features of FHWT have been identified retrospectively, and have yet to be prospectively studied or integrated into risk stratification. These include epithelial histology; 1q gain; LOH 11p15; combination of LN involvement with isolated LOH of 1p or 16q; and post-chemotherapy blastemal-predominant histology in the COG treatment context.

Features relevant to very-low-risk disease

Features relevant to VLR disease include molecular characteristics, epithelial histology and TNW.

Molecular features of VLR disease can be used to predict risk of relapse. The results of AREN0532 validated findings from NWTS-5 that LOH of 11p15 is associated with relapse in patients with VLR Wilms tumour treated without chemotherapy5,7. LOH of 11p15 was present in 37% of patients with VLR Wilms tumour (40 of 108), with 20% (8 of 40) experiencing disease relapse, accounting for 67% of all VLR relapses (8 of 12)7. Loss of imprinting of 11p15 was present in 7.4% (8 of 108 patients), and was also associated with relapse in 25% (2 of 8 patients). Combined LOH of 1p and 16q was rare in patients with VLR (3 of 108); 33% (1 of 3 patients) relapsed, but small numbers preclude statistical conclusions from being drawn7. Last, 1q gain, a poor prognostic factor in stage I (non-VLR) disease in NWTS-5 (ref. 53) and in higher stage disease14, was found in 5.5% (6 of 108) of patients with VLR. Of these 6 patients, 1 experienced relapse, but small numbers limit conclusions about its prognostic influence in patients with VLR di