Key Points
Question
Can the McGill Interactive Pediatric OncoGenetic Guidelines (MIPOGG) differentiate between children with cancer who have a low or high likelihood of having a cancer predisposition syndrome?
Findings
In phase 1 of this predictive accuracy study, MIPOGG facilitated earlier recognition of cancer predisposition syndromes in 412 pediatric oncology patients. In phase 2, MIPOGG demonstrated a favorable diagnostic accuracy profile for pediatric-onset cancer predisposition syndromes in 658 different pediatric oncology patients (no patient overlap between phases),with a positive predictive value of 17.6% and a negative predictive value of 98.6% when compared with germline DNA sequencing through precision medicine programs.
Meaning
These results suggest that the MIPOGG electronic health tool may enhance pediatric oncology care by facilitating rapid cancer predisposition syndrome risk assessment and by standardizing the recommendations for cancer predisposition syndrome evaluation in an easy and accessible manner.
Abstract
Importance
Prompt recognition of a child with a cancer predisposition syndrome (CPS) has implications for cancer management, surveillance, genetic counseling, and cascade testing of relatives. Diagnosis of CPS requires practitioner expertise, access to genetic testing, and test result interpretation. This diagnostic process is not accessible in all institutions worldwide, leading to missed CPS diagnoses. Advances in electronic health technology can facilitate CPS risk assessment.
Objective
To evaluate the diagnostic accuracy of a CPS prediction tool (McGill Interactive Pediatric OncoGenetic Guidelines [MIPOGG]) in identifying children with cancer who have a low or high likelihood of having a CPS.
Design, Setting, and Participants
In this international, multicenter diagnostic accuracy study, 1071 pediatric (<19 years of age) oncology patients who had a confirmed CPS (12 oncology referral centers) or who underwent germline DNA sequencing through precision medicine programs (6 centers) from January 1, 2000, to July 31, 2020, were studied.
Exposures
Exposures were MIPOGG application in patients with cancer and a confirmed CPS (diagnosed through routine clinical care; n = 413) in phase 1 and MIPOGG application in patients with cancer who underwent germline DNA sequencing (n = 658) in phase 2. Study phases did not overlap. Data analysts were blinded to genetic test results.
Main Outcomes and Measures
The performance of MIPOGG in CPS recognition was compared with that of routine clinical care, including identifying a CPS earlier than practitioners. The tool’s test characteristics were calculated using next-generation germline DNA sequencing as the comparator.
Results
In phase 1, a total of 413 patients with cancer (median age, 3.0 years; range, 0-18 years) and a confirmed CPS were identified. MIPOGG correctly recognized 410 of 412 patients (99.5%) as requiring referral for CPS evaluation at the time of primary cancer diagnosis. Nine patients diagnosed with a CPS by a practitioner after their second malignant tumor were detected by MIPOGG using information available at the time of the first cancer. In phase 2, of 658 children with cancer (median age, 6.6 years; range, 0-18.8 years) who underwent comprehensive germline DNA sequencing, 636 had sufficient information for MIPOGG application. When compared with germline DNA sequencing for CPS detection, the MIPOGG test characteristics for pediatric-onset CPSs were as follows: sensitivity, 90.7%; specificity, 60.5%; positive predictive value, 17.6%; and negative predictive value, 98.6%. Tumor DNA sequencing data confirmed the MIPOGG recommendation for CPS evaluation in 20 of 22 patients with established cancer-CPS associations.
Conclusions and Relevance
In this diagnostic study, MIPOGG exhibited a favorable accuracy profile for CPS screening and reduced time to CPS recognition. These findings suggest that MIPOGG implementation could standardize and rationalize recommendations for CPS evaluation in children with cancer.
This diagnostic study of children with cancer evaluates the clinical utility of the McGill Interactive Pediatric OncoGenetic Guidelines in cancer predisposition syndrome prediction.
Introduction
Germline pathogenic and likely pathogenic variants (GPVs) in cancer predisposition genes are present in 7.6% to 14.0% of young patients with cancer1,2,3,4,5 and result in cancer predisposition syndromes (CPSs). The heterogeneous sequencing approaches, variability in the comprehensiveness of cancer gene panels used, and the diverse populations enrolled in these studies contribute to this range. Prompt CPS recognition has implications for individual patient care, cancer surveillance strategies, and consideration of genetic counseling and cascade testing for at-risk family members. Identification of CPSs relies on many factors, including recognition of phenotypic CPS features, tumor-specific and family history features, practitioner experience, and appropriately resourced health care settings. These factors combined contribute to wide variations in the likelihood that an individual child will be recognized as having a cancer secondary to an underlying CPS. Decision support tools for oncology patients based on phenotypic features and pretest probabilities can be useful to standardize practitioner recognition of appropriate patient referrals for CPS evaluation.6
Some larger centers can identify CPSs by offering forms of genomic sequencing (tumor and/or germline DNA sequencing) to every oncology patient through research or clinical initiatives, potentially revealing an array of germline variants independent of pretest probability.7,8 This strategy circumvents the need to identify patients for CPS evaluation but is not the global standard of care. Most hospitals worldwide have limited access to dedicated cancer genetics services, and decision support tools can promote rational and judicious use of resources.
The McGill Pediatric OncoGenetic Guidelines (MIPOGG) decision support tool, available in app and web-based formats, was developed after extensive review of CPS literature to assist health care professionals (HCPs) in identifying children with cancer who warrant genetic evaluation based on the likelihood of their cancer being a manifestation of an underlying CPS.9 Launched in 2019, MIPOGG contains more than 140 tumor-specific decisional algorithms that encompass all tumors listed in the International Childhood Cancer Classification and the 2016 World Health Organization Classification for Central Nervous System tumors, with embedded CPS educational modules. According to tumor type, an algorithm recommends an immediate direct referral (eTable 1 in the Supplement) or guides practitioners through sequential yes or no questions that comprise tumor-specific and universal criteria to generate a referral recommendation for or against CPS evaluation (Figure 1). A CPS evaluation is recommended if 1 or more tumor-specific or universal criteria are fulfilled. The structure and content of MIPOGG, previously published, are presented in the eMethods in the Supplement.9,10,11
Figure 1. Structure and Process of the McGill Interactive Pediatric OncoGenetic Guidelines (MIPOGG) eHealth Decision Support Tool.
Once a cancer type is selected, MIPOGG directs the user to a tumor-specific algorithm or a direct referral.
aOnce a referral for cancer predisposition syndrome (CPS) evaluation is recommended or not recommended, the MIPOGG tool offers the user access to tumor-specific educational modules that describe the known associations between each cancer type and cancer predisposition syndromes, along with hyperlinks to relevant published literature. Each cancer type has its own educational module.
bClose relative was considered a parent, sibling, aunt or uncle, grandparent, or first cousin.
cExcluding nonmelanoma skin cancer.
To evaluate the utility of MIPOGG in CPS prediction, we conducted a performance and diagnostic accuracy study in children with cancer using 2 complementary approaches. Phase 1 evaluated MIPOGG performance in identifying children with cancer and a known CPS (detected through routine clinical care); phase 2 quantified the MIPOGG test characteristics in children with cancer who underwent germline DNA sequencing through precision medicine programs.
Methods
In this international, multicenter diagnostic accuracy study, 1071 pediatric (<19 years of age) oncology patients who had a confirmed CPS (12 oncology referral centers) or who underwent germline DNA sequencing through precision medicine programs (6 centers) from January 1, 2000, to July 31, 2020, were studied. Exposures were MIPOGG application in patients with cancer and a confirmed CPS (diagnosed through routine clinical care; n = 413) in phase 1 and MIPOGG application in patients with cancer who underwent germline DNA sequencing (n = 658) in phase 2. Data analysts were blinded to genetic test results in phase 2. The performance of MIPOGG in CPS recognition was compared with that of routine clinical care, including identifying a CPS earlier than practitioners. The tool’s test characteristics were calculated using next-generation germline DNA sequencing as the comparator. The study was approved by the 2 main institutional review boards: the McGill University Health Centre Research Ethics Board and the SickKids Research Ethics Board. The institutional review boards deemed that informed consent was not required. All data were deidentified. Each collaborating site also obtained approval from their local institutional review boards. The study followed the Standards for Reporting of Diagnostic Accuracy (STARD) reporting guideline.
Phase 1: MIPOGG Performance in Patients With Cancer and a Confirmed CPS
Study Design and Participants
After research ethics approvals, a multi-institutional collaboration that involved 12 Canadian pediatric oncology centers (eTable 2 in the Supplement) was established to evaluate MIPOGG performance in oncology patients (<19 years of age) who were investigated by their clinical team for and subsequently confirmed to have a CPS (cancer diagnosis from January 1, 2000, to December 31, 2018; CPS diagnosis at any time point). A confirmed CPS was defined as a molecular diagnosis (GPV in a known cancer predisposing gene) or a clinical diagnosis by a medical geneticist that fulfilled clinical criteria for a given syndrome.12,13,14,15,16 Patients were identified via query of institutional databases and by physician recollection.
Data Collection
Deidentified patient-level data that pertained to the primary cancer diagnosis and personal and family history (available at the time of primary cancer diagnosis) were retrospectively abstracted by institutional collaborators. Three authors (C.G., N.C., and L.R.) independently applied the appropriate MIPOGG tumor algorithm, noting the MIPOGG recommendation for each patient, with no interauthor disagreement. A successful application of MIPOGG is defined in the eMethods in the Supplement.
Statistical Analysis
We evaluated the proportion of patients with cancer and a confirmed CPS who were correctly identified for CPS evaluation by MIPOGG and the ability of MIPOGG to recommend CPS evaluation at an earlier time point than routine clinical care for patients whose CPS was diagnosed subsequent to a second cancer diagnosis.
Phase 2: MIPOGG Accuracy in Patients With Cancer and Germline Sequencing Data
Study Design and Participants
After research ethics approvals, a multi-institutional collaboration that involved 6 institutions (3 in Canada, 2 in the US, and 1 in Australia) that offered comprehensive germline and tumor DNA sequencing to children (<19 years of age) with cancer (diagnosed from January 1, 2000, to July 31, 2020) was established. Enrollment indications (high-risk, relapsed, or refractory cancers [50.9%] or unselected patients with cancer [49.1%]) and genomic sequencing techniques performed are outlined in eTable 3 in the Supplement.
Data Collection
Deidentified patient-level data that pertained to the primary cancer diagnosis and personal and family history (available at the time of primary cancer diagnosis) were retrospectively abstracted by institutional collaborators. Three authors (C.G., L.R., and M.T.), blinded to sequencing results, independently applied the appropriate MIPOGG tumor algorithm, noting the MIPOGG recommendation for each case. Interauthor discrepancies were resolved by a fourth investigator (N.C.). The GPVs in CPS genes known to be associated with pediatric cancers were defined as pediatric-onset CPSs (eg, Li-Fraumeni syndrome and constitutional mismatch repair deficiency), whereas those known to predispose to adult-onset malignant tumors were deemed adult-onset CPSs (eg, heterozygous variants in BRCA1 [OMIM 113705] or ATM [OMIM 607585] or Lynch syndrome). Results of tumor sequencing were reviewed, and a second hit was defined as loss of heterozygosity (LOH) or a somatic pathogenic variant in the same gene as the GPV. Tumors were grouped into established or nonestablished associations to the identified CPS for a given patient based on expert consensus (eMethods in the Supplement). Carrier status was defined as a heterozygous state for a CPS gene that only predisposes individuals to cancer in a biallelic state. Carriers were not considered to have a CPS diagnosis. Variants in non–cancer-related genes were not captured.
Statistical Analysis
Sensitivity, specificity, positive predictive values (PPVs), negative predictive values (NPVs), and positive likelihood ratios (PLRs) and negative likelihood ratios (NLRs) of MIPOGG to identify pediatric-onset CPSs and adult-onset CPSs were calculated with 95% CIs using the Wilson score, stratified by cancer category (hematologic cancers, solid tumors, and central nervous system tumors) and subcategories. The 95% CI for the difference in 2 proportions using the Wilson score for 2 independent samples and the 95% CI for PLR and NLR based on the log method for the ratio of 2 proportions were calculated.17
In both phases, we evaluated the contribution of direct referrals, universal, and tumor-specific criteria to CPS detection. Dichotomous data were calculated using frequencies and proportions, with 95% CIs. Continuous data were calculated using medians and IQRs.
Results
Phase 1: MIPOGG Performance in Patients With Cancer and a Confirmed CPS
Across 12 Canadian sites, we identified 413 patients with cancer (median age, 3.0 years; range, 0-18 years) and a confirmed CPS, 412 of whom had sufficient information for MIPOGG application. Cancer categories and demographic data are provided in Table 1. Diagnoses of CPSs were confirmed molecularly (n = 306), clinically only (n = 102), or not specified (n = 5).
Table 1. Demographic Characteristics of the Study Populationa.
| Characteristic | Phase 1 (n = 413)b | Phase 2 (n = 658)b |
|---|---|---|
| Age, median (IQR), y | 3.0 (1.0-9.0) | 6.6 (2.6-12.0) |
| Hematologic cancers | 19 (4.6) | 182 (27.7) |
| ALL | 9 (47.4) | 123 (67.6) |
| MDS or AML | 2 (10.5) | 47 (25.8) |
| Lymphomas | 8 (42.1) | 12 (6.6) |
| Solid tumors | 205 (49.6) | 303 (46.0) |
| Bone and soft-tissue sarcomas | 31 (15.1) | 102 (33.7) |
| Neuroblastic tumors | 11 (5.4) | 66 (21.8) |
| Renal tumors | 38 (18.5) | 23 (7.6) |
| Retinoblastoma | 30 (14.6) | 30 (9.9) |
| Germ cell tumors | 2 (1.0) | 6 (2.0) |
| Liver tumors | 14 (6.8) | 15 (5.0) |
| Other solid tumorsc | 79 (38.5) | 61 (20.1) |
| CNS tumors | 189 (45.8) | 173 (26.3) |
| Gliomas | ||
| Low grade | 116 (61.4) | 28 (16.2) |
| High grade | 18 (9.5) | 56 (32.4) |
| Medulloblastoma | 11 (5.8) | 27 (15.6) |
| Embryonal tumors | 21 (11.1) | 10 (5.8) |
| Ependymomas | 3 (1.6) | 17 (9.8) |
| Choroid plexus tumors | 5 (2.6) | 3 (1.7) |
| Other CNS tumorsd | 15 (7.9) | 32 (18.5) |
| Family history available | 370 (89.6) | 615 (93.5) |
| Family history of cancer | ||
| Suggestive of a CPSe | 166 (44.9) | 115 (18.7) |
| Nonspecific | 96 (25.9) | 360 (58.5) |
| No family history of cancer | 108 (29.2) | 140 (22.8) |
| Congenital anomalies or dysmorphic features | ||
| Available | 410 (99.3) | 619 (94.1) |
| Yes | 38 (9.3) | 33 (5.3) |
| No | 372 (90.7) | 586 (94.7) |
Abbreviations: ALL, acute lymphoblastic leukemia; AML, acute myeloid leukemia; CNS, central nervous system; CPS, cancer predisposition syndrome; MDS, myelodysplastic syndrome.
Data are presented as number (percentage) of patients unless otherwise indicated.
There was no overlap of patients between study groups.
Other solid tumors include desmoid type fibromatosis; nonmedullary thyroid carcinoma; chordoma; rhabdoid tumor; germ cell tumor; melanoma; desmoplastic small round cell tumor; gastrointestinal stromal tumor; malignant peripheral nerve sheath tumor; plexiform fibrohistiocytic tumor; inflammatory myofibroblastic tumor; neuroendocrine tumor; Sertoli-Leydig cell tumor; peripheral neuroectodermal tumor; pheochromocytoma or paraganglioma; adrenocortical carcinoma; carcinoma of the ovary, salivary gland, gastrointestinal tract, or breast; squamous cell carcinoma; myoepithelioma; and spindle cell neoplasm.
Other CNS tumors include craniopharyngioma, germ cell tumor, meningioma, and pineal parenchymal tumor.
Family history of cancer suggestive of a CPS; must fit 1 or more of the following criteria: (1) cancer predisposition syndrome known in family, (2) parent or sibling with a cancer at younger than 50 years, (3) aunt, uncle, first cousin, or grandparent with a cancer at younger than 18 years, (4) same cancer type or organ affected by cancer in a close relative (parent, sibling, aunt, uncle, first cousin, or grandparent), (5) close relative with multiple primary tumors (excluding nonmelanoma skin cancer) at younger than 60 years.
MIPOGG correctly recommended 410 of 412 patients (99.5%) with a CPS for genetic evaluation at the time of primary cancer diagnosis (Figure 2A). Of 410 patients, 153 (37.3%) were identified by cancer type alone, whereas 257 (62.7%) were identified by 1 or more algorithmic criteria. MIPOGG did not trigger referral in 2 patients (0.5%): a child with rhabdomyosarcoma with a pathogenic TP53 variant (OMIM 191170) and another child with acute lymphoblastic leukemia with Weaver syndrome and subtle overgrowth features. Nine patients developed a second cancer before CPS diagnosis by HCPs; all would have been identified by MIPOGG using information available at primary cancer diagnosis (eTable 4 in the Supplement).
Figure 2. Flowchart of the Study Processes Showing the Number of Patients Recommended or Not for Cancer Predisposition Syndrome (CPS) Evaluation by the McGill Interactive Pediatric OncoGenetic Guidelines (MIPOGG) and Number of Patients With Pediatric- or Adult-Onset CPSs.
HCPs indicates health care professionals.
Phase 2: MIPOGG Accuracy in Patients With Cancer and Germline Sequencing Data
Of 658 children with cancer (median age, 6.6 years; range, 0-18.8 years) who underwent comprehensive germline DNA sequencing, 636 had sufficient information for MIPOGG application. Cancer categories and patient demographic characteristics are presented in Table 1. In 81 patients (12.3%), GPVs in CPS genes were identified, with a pediatric-onset CPS confirmed in 54 patients (8.2%) and an adult-onset CPS detected in 27 patients (4.1%).
MIPOGG recommended CPS evaluation in 279 of 636 patients (43.9%), 49 (17.6%) of whom had a GPV in a pediatric-onset CPS gene. Conversely, MIPOGG did not recommend CPS evaluation in 357 of 636 patients (56.1%), 5 (1.4%) of whom had a GPV in a pediatric-onset CPS gene (difference, 16.2%; 95% CI, 11.8%-21.1%) (Figure 2B). The proportions of pediatric- and adult-onset CPSs detected using direct referrals and tumor-specific and/or universal criteria are shown in Figure 3. Genetic diagnoses, associated cancer types, and MIPOGG referral status are further described in eFigure 1 in the Supplement.
Figure 3. Cancer Predisposition Syndrome (CPS) Detection According to the McGill Interactive Pediatric OncoGenetic Guidelines (MIPOGG).
Most patients with no identified CPSs did not meet any MIPOGG criteria. Most patients with pediatric-onset CPSs were identified from direct referrals based on their tumor type, followed by those patients who met only tumor-specific criteria or who met some combination of both tumor-specific and universal criteria. Most patients with adult-onset CPSs met no MIPOGG criteria because MIPOGG is designed for recognizing patterns in pediatric-onset CPSs.
The test characteristics of MIPOGG for pediatric-onset CPS were as follows: sensitivity, 90.7%; specificity, 60.5%; PPV, 17.6%; and NPV, 98.6% (Table 2). The PLR of 2.30 (95% CI, 2.01-2.62) and the NLR of 0.15 (95% CI, 0.07-0.35) indicate incremental value of both MIPOGG outcomes, with a higher informative value when CPS evaluation is not recommended. Differences between accuracy measurements for tumor subcategories are presented on a receiver operating characteristic plot (eFigure 2 in the Supplement) and were not statistically significant. Although not primarily designed to detect adult-onset CPSs, the test characteristics of MIPOGG in this group were as follows: sensitivity, 59.3%; specificity, 61.4%; PPV, 7.0%; and NPV, 96.9%.
Table 2. Test Characteristics of MIPOGG for Pediatric-Onset Cancer Predisposition Syndromes in Children With Cancer Who Underwent Comprehensive Genetic Testing (Study Phase 2).
| Cancer type | No. (%) of children | Sensitivity, % (95% CI) | Specificity, % (95% CI) | PPV, % (95% CI) | NPV, % (95% CI) | PLR (95% CI) | NLR (95% CI) | |||
|---|---|---|---|---|---|---|---|---|---|---|
| Total | CPS confirmed | Patients evaluable by MIPOGGa | MIPOGG Yesb | |||||||
| Total | 658 | 54 (8.2) | 636 (96.7) | 279 (43.9) | 90.7 (0.80-0.96) | 60.5 (0.56-0.64) | 17.6 (0.14-0.22) | 98.6 (0.97-0.99) | 2.30 (2.01-2.62) | 0.15 (0.07-0.35) |
| Hematologic cancers | 182 | 9 (4.9) | 176 (96.7) | 68 (38.6) | 100 (0.70-1.00) | 64.7 (0.57-0.72) | 13.2 (0.07-0.23) | 100 (0.97-1.00) | 2.83 (2.31-3.48) | 0 |
| Leukemia or MDS | 170 | 7 (4.1) | 166 (97.6) | 64 (38.6) | 100 (0.65-1.00) | 64.2 (0.56-0.71) | 10.9 (0.05-0.21) | 100 (0.96-1.00) | 2.79 (2.27-3.43) | 0 |
| Lymphoma | 12 | 2 (16.7) | 10 (83.3) | 4 (40.0) | 100 (0.34-1.00) | 75.0 (0.41-0.93) | 50.0 (0.15-0.85) | 100 (0.61-1.00) | 4.00 (1.2-13.28) | 0 |
| Solid tumors | 303 | 30 (9.9) | 298 (98.3) | 148 (49.7) | 93.3 (0.79-0.98) | 55.2 (0.49-0.61) | 18.9 (0.13-0.26) | 98.7 (0.95-0.99) | 2.08 (1.77-2.46) | 0.12 (0.03-0.46) |
| Sarcomas | 102 | 6 (5.9) | 101 (99.0) | 37 (36.6) | 66.7 (0.30-0.90) | 65.3 (0.55-0.74) | 10.8 (0.04-0.25) | 96.9 (0.89-0.99) | 1.92 (1.02-3.60) | 0.51 (0.16-1.60) |
| Neuroblastic tumors | 66 | 1 (1.5) | 64 (97.0) | 15 (23.4) | 100 (0.21-1.00) | 77.8 (0.66-0.86) | 6.7 (0.01-0.30) | 100 (0.93-1.00) | 4.50 (2.84-7.14) | 0 |
| Renal tumors | 23 | 2 (8.7) | 23 (100) | 18 (78.3) | 100 (0.34-1.00) | 23.8 (0.11-0.45) | 11.1 (0.03-0.33) | 100 (0.57-1.00) | 1.31 (1.03-1.67) | 0 |
| Liver tumors | 15 | 1 (6.7) | 15 (100) | 10 (66.7) | 100 (0.21-1.00) | 35.7 (0.16-0.61) | 10.0 (0.02-0.40) | 100 (0.57-1.00) | 1.56 (1.05-2.30) | 0 |
| Retinoblastoma | 30 | 10 (33.3) | 30 (100) | 30 (100) | 100 (0.72-1.00) | 0 | 33.3 (0.19-0.51) | NA | 1.00 (1.00-1.00) | NA |
| Other solid tumors | 67 | 10 (14.9) | 65 (97.0) | 38 (58.5) | 100 (0.72-1.00) | 49.1 (0.36-0.62) | 26.3 (0.15-0.42) | 100 (0.88-1.00) | 1.96 (1.52-2.55) | 0 |
| Brain tumors | 173 | 15 (8.7) | 162 (93.6) | 63 (38.9) | 80.0 (0.55-0.93) | 65.3 (0.57-0.73) | 19.0 (0.11-0.30) | 97.0 (0.91-0.99) | 2.31 (1.65-3.23) | 0.31 (0.11-0.85) |
| Gliomas | 84 | 11 (13.0) | 76 (90.5) | 33 (43.4) | 72.7 (0.43-0.90) | 61.5 (0.49-0.72) | 24.2 (0.13-0.41) | 93.0 (0.81-0.98) | 1.89 (1.18-3.04) | 0.44 (0.17-1.19) |
| Embryonal tumors | 37 | 2 (5.4) | 35 (94.6) | 17 (48.6) | 100 (0.34-1.00) | 54.5 (0.38-0.70) | 11.8 (0.033-0.34) | 100 (0.82-1.00) | 2.20 (1.51-3.20) | 0 |
| Ependymomas | 17 | 1 (5.9) | 17 (100) | 5 (29.4) | 100 (0.21-1.00) | 75.0 (0.51-0.90) | 20.0 (0.04-0.60) | 100 (0.76-1.00) | 4.00 (1.71-9.35) | 0 |
| Other brain tumors | 35 | 1 (2.9) | 34 (97.1) | 8 (23.6) | 100 (0.21-1.00) | 78.8 (0.62-0.89) | 12.5 (0.02-0.47) | 100 (0.87-1.00) | 4.71 (2.44-9.10) | 0 |
Abbreviations: CPS, cancer predisposition syndrome; MDS, myelodysplastic syndrome; MIPOGG, McGill Interactive Pediatric OncoGenetic Guidelines; NA, not applicable; NLR, negative likelihood ratio; NPV, negative predictive value; PLR, positive likelihood ratio; PPV, positive predictive value.
Number of patients in whom MIPOGG could be successfully applied.
A MIPOGG yes response signifies that MIPOGG recommends a CPS evaluation.
Although MIPOGG does not incorporate tumor sequencing data in its algorithms, we reviewed the presence of somatic second hits and LOH in the patients with GPVs in CPS genes. Fifty-five of 81 patients (67.9%) had available tumor DNA sequencing data (eTable 5 in the Supplement). Thirty of 55 (54.5%) had a second hit in the tumor, 22 of whom had an established pediatric cancer-CPS association. MIPOGG identified 20 of these 22 patients (90.9%) as requiring CPS evaluation (eTable 5 in the Supplement). The 2 patients who were not identified by MIPOGG had high-grade gliomas and a GPV in TP53 with somatic LOH of the wild-type allele (eTable 5 in the Supplement). Another patient with a central nervous system embryonal tumor and a monoallelic MSH2 (OMIM 609309) GPV, a cancer-CPS association that is not yet established, also had a second hit and was not recommended for CPS evaluation by MIPOGG (eTable 5 in the Supplement). A flowchart (eFigure 3 in the Supplement) details the presence or absence of a second hit for genes with and without an established association with a pediatric CPS as well as the MIPOGG recommendation for or against CPS evaluation.
Discussion
This diagnostic study assessed the accuracy and utility of the MIPOGG CPS screening tool in pediatric oncology. MIPOGG performed as well as HCPs in identifying patients at primary cancer diagnosis with a CPS and identified all patients who subsequently developed a second malignant tumor at an earlier time point than HCPs. Timely CPS recognition is critical to allow implementation of cancer surveillance and cascade testing, which can improve patient outcomes.18 MIPOGG exhibits a highly favorable CPS prediction profile in children with cancer who underwent comprehensive germline DNA sequencing, with an overall sensitivity of 90.7% and NPV of 98.6%. Because MIPOGG is a screening tool, a higher sensitivity at the expense of specificity was tolerated. The overall referral rate was 43.9%; among these children, 17.6% had a pediatric-onset CPS.
In patients who underwent large-scale germline DNA sequencing (ranging from 48 genes to whole genome sequencing), MIPOGG missed 5 pediatric-onset CPS diagnoses; 2 patients had a somatic second hit in TP53 and fit established cancer-CPS associations, highlighting an area for refinement of the MIPOGG algorithms (high-grade glioma in this case). The remaining 3 patients did not exhibit LOH or fit established cancer-CPS associations and could be considered to have incidental (secondary) findings: the first patient had a low-grade glioma and a GPV in SDHA (OMIM 600857), the second had a Ewing sarcoma and a GPV in BAP1 (OMIM 603089), and the third patient had a Ewing sarcoma and a GPV in FH (OMIM 136850) (eTable 5 in the Supplement). The presence of a second hit in the 2 patients above with TP53 GPVs (eFigure 3 and eTable 5 in the Supplement) demonstrates where tumor sequencing data were most informative. Most other patients who had a second hit with an established cancer-CPS association (n = 20) were correctly identified by MIPOGG (eFigure 3 and eTable 5 in the Supplement). Therefore, tumor sequencing data were less helpful in these patients and would not have significantly improved the performance of MIPOGG.
Strengths and Limitations
This study has strengths, including a rigorous approach leveraging data pertaining to more than 1000 pediatric oncology patients with and without CPSs to validate MIPOGG as the only electronic health CPS screening tool in pediatric oncology that has completed a series of performance assessments establishing its diagnostic accuracy measures. Additional strengths include its high discriminative qualities and its external validation through assessment in heterogeneous pediatric cancer populations.19 Similar performance results were identified by Byrjalsen et al,8 who independently demonstrated that MIPOGG detected 85.7% of pediatric-onset CPSs in a nationwide pediatric oncology sequencing initiative (n = 198).
This study also has limitations. In phase 1, most children identified by HCPs had phenotypic features, striking family histories, or tumor types suggestive of CPSs, creating a population bias and rendering patients more easily identifiable by MIPOGG. Nevertheless, these features replicate clinical scenarios, and we considered it essential to demonstrate that MIPOGG was not inferior to HCP practice. We also acknowledge limitations in the development of MIPOGG algorithms, which reflect known cancer-CPS associations at a specific time point and require continual updating.20 The incorporation of universal criteria, however, serves to potentially identify novel CPSs and previously unknown cancer-CPS associations. For example, the universal criteria identified a patient with a neuroblastoma and a pathogenic SMARCA4 variant (OMIM 603254), currently considered a secondary finding but one that may eventually be deemed a novel association.
The disclosure of secondary findings in multigene panel sequencing in clinical and research settings continues to provoke debate, especially in pediatric populations.21,22 The current goal of MIPOGG is the identification of established pediatric cancer-CPS associations in a clinical context.
Although more than 96% of patients had sufficient information to apply MIPOGG, both study phases involved retrospective data capture in which data accuracy is uncertain. Patients enrolled in phase 2 did not undergo identical DNA sequencing techniques, possibly contributing to an underestimation of MIPOGG performance in some cases. The MIPOGG referral rate of 43.9% leads to additional strain on cancer genetics services, arguing for new models of service delivery. In contrast, more than 50% of children were deemed at low risk for CPSs. Implementation of MIPOGG may therefore facilitate prioritization of genetic testing resources, with limited risk of missing CPS diagnoses (NLR, 0.15), which is particularly applicable in centers that lack the capacity to perform genetic testing in all patients or in situations where payer coverage issues need to be considered.
Conclusions
In this study, MIPOGG exhibited a highly favorable accuracy profile for CPS screening and is likely to enhance clinical oncology care by facilitating rapid CPS risk assessment and by standardizing approaches to CPS evaluation. MIPOGG will next be assessed prospectively in unselected children with cancer who are undergoing germline DNA sequencing. It will also be expanded to incorporate cancers in young adults, with cost-effectiveness evaluation. Finally, MIPOGG algorithms will be modified using artificial intelligence technologies that guide prioritization of possible CPSs through pattern recognition of phenotypes and genotypes in the setting of various tumors.
eMethods. Phases of MIPOGG Development and Tool Structure
eTable 1. Direct Referrals Included in MIPOGG (Current Edition)
eTable 2. Canadian Centers Participating in Study Phase 1
eTable 3. Genomic Sequencing Initiatives That Contributed Data to Study Phase 2
eTable 4. MIPOGG Study Phase 1: MIPOGG Detects CPSs Earlier Than Clinicians
eTable 5. Tumor DNA Sequencing and MIPOGG Referral Status in Patients With a Germline Variant in a CPS-Related Gene
eFigure 1. MIPOGG Performance in Identification of Patients With Germline Variants Submitted by Precision Medicine Programs (Study Phase 2)
eFigure 2. Receiver Operating Curve Plot
eFigure 3. Added Value of Tumor Sequencing to MIPOGG Recommendations
References
- 1.Gröbner SN, Worst BC, Weischenfeldt J, et al. ; ICGC PedBrain-Seq Project; ICGC MMML-Seq Project . The landscape of genomic alterations across childhood cancers. Nature. 2018;555(7696):321-327. doi: 10.1038/nature25480 [DOI] [PubMed] [Google Scholar]
- 2.Zhang J, Walsh MF, Wu G, et al. Germline mutations in predisposition genes in pediatric cancer. N Engl J Med. 2015;373(24):2336-2346. doi: 10.1056/NEJMoa1508054 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Wong M, Mayoh C, Lau LMS, et al. Whole genome, transcriptome and methylome profiling enhances actionable target discovery in high-risk pediatric cancer. Nat Med. 2020;26(11):1742-1753. doi: 10.1038/s41591-020-1072-4 [DOI] [PubMed] [Google Scholar]
- 4.Parsons DW, Roy A, Yang Y, et al. Diagnostic yield of clinical tumor and germline whole-exome sequencing for children with solid tumors. JAMA Oncol. 2016;2(5):616-624. doi: 10.1001/jamaoncol.2015.5699 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Fiala EM, Jayakumaran G, Mauguen A, et al. Prospective pan-cancer germline testing using MSK-IMPACT informs clinical translation in 751 patients with pediatric solid tumors. Nature Cancer. 2021;2:357-365. doi: 10.1038/s43018-021-00172-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Carver T, Hartley S, Lee A, et al. CanRisk Tool: a web interface for the prediction of breast and ovarian cancer risk and the likelihood of carrying genetic pathogenic variants. Cancer Epidemiol Biomarkers Prev. 2021;30(3):469-473. doi: 10.1158/1055-9965.EPI-20-1319 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Khater F, Vairy S, Langlois S, et al. Molecular profiling of hard-to-treat childhood and adolescent cancers. JAMA Netw Open. 2019;2(4):e192906. doi: 10.1001/jamanetworkopen.2019.2906 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Byrjalsen A, Hansen TVO, Stoltze UK, et al. Nationwide germline whole genome sequencing of 198 consecutive pediatric cancer patients reveals a high incidence of cancer prone syndromes. PLoS Genet. 2020;16(12):e1009231. doi: 10.1371/journal.pgen.1009231 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Goudie C, Coltin H, Witkowski L, Mourad S, Malkin D, Foulkes WD. The McGill Interactive Pediatric OncoGenetic Guidelines: an approach to identifying pediatric oncology patients most likely to benefit from a genetic evaluation. Pediatr Blood Cancer. 2017;64(8). doi: 10.1002/pbc.26441 [DOI] [PubMed] [Google Scholar]
- 10.Goudie C, Cullinan N, Villani A, et al. Retrospective evaluation of a decision-support algorithm (MIPOGG) for genetic referrals for children with neuroblastic tumors. Pediatr Blood Cancer. 2018;65(12):e27390. doi: 10.1002/pbc.27390 [DOI] [PubMed] [Google Scholar]
- 11.Cullinan N, Villani A, Mourad S, et al. An eHealth decision-support tool to prioritize referral practices for genetic evaluation of patients with Wilms tumor. Int J Cancer. 2020;146(4):1010-1017. doi: 10.1002/ijc.32561 [DOI] [PubMed] [Google Scholar]
- 12.Brioude F, Kalish JM, Mussa A, et al. Expert consensus document: clinical and molecular diagnosis, screening and management of Beckwith-Wiedemann syndrome: an international consensus statement. Nat Rev Endocrinol. 2018;14(4):229-249. doi: 10.1038/nrendo.2017.166 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Baser ME, Friedman JM, Aeschliman D, et al. Predictors of the risk of mortality in neurofibromatosis 2. Am J Hum Genet. 2002;71(4):715-723. doi: 10.1086/342716 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Stumpf D; National Institutes of Health Consensus Development Conference . Neurofibromatosis: conference statement. Arch Neurol. 1988;45(5):575-578. doi: 10.1001/archneur.1988.00520290115023 [DOI] [PubMed] [Google Scholar]
- 15.Evans DG, Ladusans EJ, Rimmer S, Burnell LD, Thakker N, Farndon PA. Complications of the naevoid basal cell carcinoma syndrome: results of a population based study. J Med Genet. 1993;30(6):460-464. doi: 10.1136/jmg.30.6.460 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Northrup H, Krueger DA; International Tuberous Sclerosis Complex Consensus Group . Tuberous sclerosis complex diagnostic criteria update: recommendations of the 2012 International Tuberous Sclerosis Complex Consensus Conference. Pediatr Neurol. 2013;49(4):243-254. doi: 10.1016/j.pediatrneurol.2013.08.001 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Altman DG MD, Bryant TN, Gardner MJ. Statistics With Confidence. 2nd ed. BMJ Books; 2000. [Google Scholar]
- 18.Villani A, Tabori U, Schiffman J, et al. Biochemical and imaging surveillance in germline TP53 mutation carriers with Li-Fraumeni syndrome: a prospective observational study. Lancet Oncol. 2011;12(6):559-567. doi: 10.1016/S1470-2045(11)70119-X [DOI] [PubMed] [Google Scholar]
- 19.Khalifa M, Magrabi F, Gallego B. Developing a framework for evidence-based grading and assessment of predictive tools for clinical decision support. BMC Med Inform Decis Mak. 2019;19(1):207. doi: 10.1186/s12911-019-0940-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Mirabello L, Zhu B, Koster R, et al. Frequency of pathogenic germline variants in cancer-susceptibility genes in patients with osteosarcoma. JAMA Oncol. 2020;6(5):724-734. doi: 10.1001/jamaoncol.2020.0197 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Chirita-Emandi A, Andreescu N, Zimbru CG, et al. Challenges in reporting pathogenic/potentially pathogenic variants in 94 cancer predisposing genes—in pediatric patients screened with NGS panels. Sci Rep. 2020;10(1):223. doi: 10.1038/s41598-019-57080-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Ripperger T, Evans DG, Malkin D, Kratz CP. Choose and stay on one out of two paths: distinction between clinical versus research genetic testing to identify cancer predisposition syndromes among patients with cancer. Fam Cancer. 2021. Published online February 12, 2021. doi: 10.1007/s10689-021-00228-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eMethods. Phases of MIPOGG Development and Tool Structure
eTable 1. Direct Referrals Included in MIPOGG (Current Edition)
eTable 2. Canadian Centers Participating in Study Phase 1
eTable 3. Genomic Sequencing Initiatives That Contributed Data to Study Phase 2
eTable 4. MIPOGG Study Phase 1: MIPOGG Detects CPSs Earlier Than Clinicians
eTable 5. Tumor DNA Sequencing and MIPOGG Referral Status in Patients With a Germline Variant in a CPS-Related Gene
eFigure 1. MIPOGG Performance in Identification of Patients With Germline Variants Submitted by Precision Medicine Programs (Study Phase 2)
eFigure 2. Receiver Operating Curve Plot
eFigure 3. Added Value of Tumor Sequencing to MIPOGG Recommendations