Abstract
Background
Due to the decreasing prevalence of IDH1 mutations in older patients, the 2016 World Health Organization (WHO) classification of brain tumors proposed not to perform sequencing for isocitrate dehydrogenase (IDH) in glioblastoma patients ≥55 years old. We present a cost-effectiveness analysis to estimate the financial impact of these guidelines.
Methods
From 2010 to 2015 we performed 1023 IDH tests in gliomas, amounting to ~$1.09 million in direct laboratory test costs. Samples were tested using R132H-specific immunohistochemistry, DNA sequencing validated for detection of noncanonical IDH1/2 mutations, or both methods.
Results
In cases tested by DNA sequencing, the fraction of non-R132H mutations was 5.4%, which included only 2 high-grade gliomas in patients ≥55 years (0.9%). When remodeling the optimal age cutoff in our patient population using 5-year age-binning, we found a 10-times higher pretest probability for the presence of a noncanonical IDH1 mutation in the setting of a negative IDH1-R132H immunohistochemistry result in patients <55 years. Applying the independently confirmed age cutoff of 55 years to glioblastoma patients (64%) would result in $403200 saved (43%). By not performing sequencing in patients ≥55 years, the turn-around time to final integrated neuropathological diagnosis is reduced by 53%, allowing these patients to gain earlier benefits from personalized genomic medicine.
Conclusion
The negligible prevalence of noncanonical IDH mutations in glioblastoma patients ≥55 years argues against universal IDH sequencing in this population. We predict that adoption of this age-based sequencing cutoff recommendation from the 2016 WHO guidelines will result in significant cost and time savings throughout the global health care system.
Keywords: biomarker, cost-effectiveness, IDH testing, reimbursement, turn-around time
Importance of the study
The new WHO classification for brain tumors explicitly demands molecular testing; however, the decreasing prevalence of IDH1 mutations in older patients resulted in a proposal not to perform sequencing for IDH in glioblastoma patients ≥55 years. The proposal has several important practical as well as financial implications for molecular testing of brain tumors. We used data from >1000 molecular IDH tests, amounting to >$1 million in direct laboratory test costs, to determine the financial implications of the proposed age-based testing cutoff. Our results indicate that adoption of the 2016 WHO guidelines will result in significant cost and time savings throughout the global health care system.
Mutations in the isocitrate dehydrogenase genes (IDH1/IDH2) are frequent in diffuse gliomas.1–3 The clinical utility of detecting these variants has direct diagnostic,4 prognostic,5,6 and potentially therapeutic implications.7,8 Specifically, IDH mutation detection distinguishes lower-grade glioma from astrocytosis,4 predicts improved survival in gliomas,6 predicts chemosensitivity to alkylating chemotherapies,9 and identifies tumors that may benefit from novel targeted or vaccination therapies.7,10 Based on these findings, IDH mutation screening has become an integral part of the diagnostic workup of brain tumors.
The IDH1 p.R132H imparts a gain-of-function phenotype to the IDH1 enzyme, leading to the overproduction of the oncometabolite 2-hydroxyglutarate.11 This so-called canonical mutation is the most frequent alteration and has triggered the development of reliable screening by mutation-specific antibodies.12 However, not all substitutions at the IDH1 p.R132 locus can be detected by immunohistochemistry (non-R132H variants), and analogous mutations at the R172 locus in the IDH2 gene have similar oncogenic effects. The current consensus is that all gliomas should be screened for IDH1 p.R132H mutations by immunohistochemistry,13,14 and sequencing (eg, Sanger or next-generation technology–based) is recommended as a second-step test.13 While these recommendations imply that equivocal or immunohistochemistry (R132H)-negative cases should be screened, the exact clinical workflow has not been finalized because the prevalence of IDH mutations in diffuse gliomas differs by patient age and tumor grade.15,16 Specifically, the prevalence of IDH1/IDH2 variants in glioma patients decreases with age and is much more common in lower-grade lesions (World Health Organization [WHO] grades II–III) than in WHO grade IV glioblastoma.1 Previous studies have attempted to model the probability of an IDH mutation given a number of different variables (grade, age, R132H-IDH1 immunohistochemistry result, etc), and based on these data, the 2016 WHO has suggested an age cutoff of 55 years for sequencing in glioblastoma following a negative R132H-IDH1 immunohistochemistry result.13,17,18 While clinically useful, this approach has, to our knowledge, not been directly linked to test cost and the associated financial implications.
Thus, despite the availability of several valid methods for variant screening, the most cost-effective screening algorithm taking age and WHO grade into account has not been formally examined. Here we compared the detection rates at the protein versus DNA level to derive the most cost-effective, age-based screening model.
Materials and Methods
Setting
The project site was Massachusetts General Hospital (MGH). The project, undertaken as a prospective quality improvement initiative, does not require formal review and approval by the institutional review board for research activities per its policies (institutional checklist version May 25, 2012).
Project Design
The project design is a retrospective cohort analysis. We investigated diagnoses of diffuse gliomas in patients between January 2010 and December 2015. Outside consultation cases were excluded. These data were then used to model the utility of different screening modalities (immunohistochemistry and DNA sequencing). The primary endpoint of the study was the optimal age cutoff in years for the most effective screening method.
Archival Searches
We searched pathology and hospital information systems using diagnostic terms and codes from the International Classification of Diseases for Oncology. Keywords “immunohistochemistry” and “IDH1” were used to identify cases that had immunohistochemistry testing. Outside consultation and duplicate cases were removed. When multiple specimens existed from the same patient, we separated date of original diagnosis (first sample) from date of testing (typically the most recent sample). Reviewing all primary brain tumor cases sent for sequencing between January 2010 and December 2015 identified cases tested by DNA sequencing. Again, outside consultation and duplicate cases were removed. At least one board-certified pathologist confirmed each primary diagnosis by review of original sections, immunohistochemical stains, and/or sequencing results. All study cases were originally classified according to the 2007 WHO guidelines.19 We converted these diagnoses into the 2016 WHO CNS terminology via consideration of IDH mutation as well as 1p/19q copy number status.13,18 We provide reclassification details (Fig. 1); however, for our cost-effectiveness analyses, we combined all glioblastomas (WHO grade IV) and all lower-grade diffuse gliomas (WHO grades II and III).
Fig. 1.
Detection rates. (A) Overview of all study cases with pertinent clinicopathological features showing original (WHO 2007) and reclassified diagnoses (WHO 2016). Arrows indicate cases with WHO grade changes. (B) Reclassification according to WHO 2016 guidelines resulted in upgrading from WHO grade III to IV in N = 12 cases. These tumors had oligodendroglioma-like histology but failed to meet the molecular requirements for a diagnosis of anaplastic oligodendroglioma of the 2016 WHO classification. (C) Blue boxes outline cases tested by sequencing, while red boxes outline cases tested by immunohistochemistry (IHC). Overlapping boxes represent cases tested by both methods. In all, 482 cases received DNA testing, with 331 cases testing negative and 151 testing positive. Of 541 total cases tested by IHC, 406 were negative and 135 were positive. (D) Cases tested by both methods were binned by 5-year age intervals, except at the extremes. Detection rate is the percentage of cases testing positive for an IDH variant by each method. Black boxes designate IHC detection rate and gray boxes designate sequencing detection rate. Sequencing outperforms IHC detection at younger ages (<55 y), suggesting higher rates of noncanonical IDH1 and IDH2 variants in younger patients. (E) IDH-variant detection rate by age binned at 55 years. Black bars represent IHC and gray bars represent sequencing. In patients <55 years, DNA testing reveals a significantly higher detection rate compared with IHC (*P < 0.001, t-test), whereas there was no difference between both methods in patients ≥55 years. Abbreviations: GBM, glioblastoma; WT, wild-type; MUT, mutant; NOS, not otherwise specified (WHO 2016; here IHC only, no DNA testing performed); AA, anaplastic astrocytoma; AO, anaplastic oligodendroglioma, IDH-mutant, 1p/19q codeleted; AO NOS, anaplastic oligodendroglioma, not otherwise specified; A, diffuse astrocytoma; O, oligodendroglioma, IDH-mutant, 1p/19q codeleted; O NOS, oligodendroglioma, not otherwise specified; AOA, anaplastic astrocytoma (WHO 2007). For details of cases, see Table 1.
Definitions
Test groups refer to the retrospective assembly of cases into different groups based on the diagnostic screening modality. We distinguish 3 test groups: cases diagnosed by immunohistochemistry alone, cases diagnosed by immunohistochemistry and sequencing (combined), and those diagnosed by sequencing alone (DNA). In this study, sequencing included both “hotspot” sequencing (PCR sequencing targeted to commonly mutated areas of common genes implicated in cancer development, including IDH1 and IDH2), and next-generation sequencing (anchored multiplex PCR). “Detection rate” refers to the number of detected variants divided by the total number of cases tested by the respective methods. Here, “canonical IDH1 variant” refers to the most common IDH1 variant (p.R132H), whereas “noncanonical variant” refers to all other IDH1 and IDH2 variants.
Cost Modeling
The cost model developed in this study uses current billable cost for each screening modality: $135 per p.R132H-specific IDH1 immunohistochemistry, $420 single gene sequencing, and $1800 for next-generation sequencing. Turn-around times employed average times tracked during routine operations at MGH (ie, 2 days for immunohistochemistry, and 14 days for next-generation sequencing). We assumed that tumors harboring the canonical IDH1 mutation (ie, detectable by p.R132H-IDH1 specific immunohistochemistry) would not require sequencing. For the cost and turn-around time model, the following formulas were applied:
No age-based cutoff: M no = nyc IHC + [(nywt + nync) NGS] + nmc IHC + [(nmwt + nmnc) NGS]
With ≥55 year NGS cutoff:Mage = nyc IHC + [(nywt + nync) NGS] + nmc IHC,
where M = modeled result; ny = patients <55 years; nm= patients ≥55 years; c = canonical mutations; wt = wild-type; nc = noncanonical mutations; IHC = immunohistochemistry; and NGS = next-generation sequencing. For the cost model we set IHC to $135 and NGS to $1800 and for the turn-around time model we set IHC to 2 days and NGS to 14 days; by dividing the total number of days by the number of tests, we estimate the average turn-around time.
Statistics
Comparison of demographics between test groups was performed by chi-squared analysis for tumor type and WHO grade, and unpaired t-test for patient age, as appropriate. Significant differences in age and detection rates were determined using an unpaired t-test; P < 0.05 was considered indicative of statistically significant differences.
Results
Study Cohort
All cases were diagnosed according to both the prior (WHO 2007)19 as well as the updated (WHO 2016) classifications.18 The updated classification takes IDH mutation status as well as 1p/19q copy number status into account and incorporates them into the name (“integrated neuropathological diagnosis”).13,15,18 Accordingly the official diagnostic tumor type changed in essentially all tumors. An overview of the study cohort is provided in Table 1 and Fig. 1A. As expected, all 72 oligoastrocytomas (by WHO 2007) were reclassified into oligodendrogliomas (if 1p/19q codeleted) or astrocytomas/glioblastomas (if not codeleted). With respect to WHO grade, in our series the reclassification resulted in upgrading to WHO grade IV in 12 cases (1.7%; N = 12/680; Fig. 1B). Specifically, these 12 cases exhibited oligodendroglioma-like histology but showed no codeletion of 1p/19q and thereby failed to meet the molecular requirements for a diagnosis of anaplastic oligodendroglioma by the 2016 WHO classification.18 For our cost-effectiveness analyses, we combined all glioblastomas and all lower-grade diffuse gliomas (Table 1).
Table 1.
Study cohort
| Features | All Cases N = 680 | DNA or IHC N = 337 |
Combined N = 343 ( = 686 tests) |
P-value |
|---|---|---|---|---|
| Tumor type (WHO 2016 classification) | <0.01a | |||
| Glioblastoma | 431 (64%) | 211 (63%) | 220 (64%) | <0.01a |
| Glioblastoma, IDH wild-type | 366 (54%) | 168 (50%) | 198 (58%) | 0.53b |
| Glioblastoma, IDH-mutant | 45 (7%) | 23 (7%) | 22 (6%) | |
| Glioblastoma, NOS | 20 (3%) | 20 (6%) | – | |
| Lower-grade glioma (WHO grade II/III) | 249 (36%) | 126 (37%) | 123 (36%) | <0.01a |
| Anaplastic astrocytoma, IDH wild-type | 38 (6%) | 6 (2%) | 32 (9%) | 0.03b |
| Anaplastic astrocytoma, IDH-mutant | 67 (10%) | 25 (7%) | 42 (12%) | |
| Anaplastic astrocytoma, NOS | 29 (4%) | 29 (9%) | – | |
|
Anaplastic oligodendroglioma, IDH-mutant, 1p/19q codeleted |
31 (5%) | 20 (6%) | 11 (3%) | |
| Anaplastic oligodendroglioma, NOS | – | – | – | |
| Diffuse astrocytoma, IDH wild-type | 12 (2%) | 3 (1%) | 9 (3%) | 0.55b |
| Diffuse astrocytoma, IDH-mutant | 32 (5%) | 11 (3%) | 21 (6%) | |
| Diffuse astrocytoma, NOS | 12 (2%) | 12 (4%) | – | |
| Oligodendroglioma, IDH-mutant, 1p/19q codeleted | 26 (4%) | 18 (5%) | 8 (2%) | |
| Oligodendroglioma, NOS | 2 (0.3%) | 2 (1%) | – | |
| WHO grade | 0.59 | |||
| IV | 431 (63%) | 211 (63%) | 220 (64%) | |
| III | 165 (24%) | 80 (24%) | 85 (25%) | |
| II | 84 (12%) | 46 (14%) | 38 (11%) | |
| Age at diagnosis, y | ||||
| Median | 57 | 58 | 56 | 0.05* |
| Number of patients <55 | 304 (45%) | 149 (44%) | 155 (45%) | 0.82 |
| Number of patients ≥55 | 376 (55%) | 188 (56%) | 188 (55%) |
Abbreviations: DNA, tumors tested by sequencing assay; IHC, tumors tested by IDH-R132H specific immunohistochemistry; NOS, not otherwise specified (specific diagnosis in the WHO 2016 classification). P-values compare DNA or IHC vs combined testing and are derived from chi-square testing (tumor type and WHO grade), Student’s t-test (age), and Fisher’s exact test (number of patients). aSignificant differences in the distribution of tumor types is related to the absence of an NOS category in the combined testing group. bResults from Fisher’s exact test comparing the IDH wild-type and IDH-mutant subgroups. *The nearly significant difference in median age is caused by a significantly higher age in the IHC alone group (63 y vs 53 y in the DNA alone; P = 0.001, t-test); the age difference between DNA alone and combined testing was not significant (53 vs 56 y; P = 0.60, t-test).
Study Cohort by Test Method
To determine the optimal screening method for IDH variant detection, we separated the 680 primary brain tumor cases by test method (Table 1). Briefly, cases were screened for IDH variants by either immunohistochemistry (198 cases), sequencing (139 cases), or both (343 cases, or 686 tests), for a total of 1023 tests (Fig. 1C). Out of 198 cases screened by immunohistochemistry alone, 51 were positive (26% IDH variant detection rate; Fig. 1C), while 47 of 139 cases tested by sequencing alone showed IDH gene variants (34% detection rate; Fig. 1C). The difference in variant detection rate is likely at least in part explained by the inability of immunohistochemistry to detect noncanonical IDH1 or IDH2 variants, as the IDH1 p.R132H antibody is specifically detecting the canonical R132H-IDH1 variant protein. In our testing practice, ~29% (N = 198/680) of cases were tested by R132H-IDH1 immunohistochemistry alone, with ~22% showing immunonegativity (N = 147/680). Review of the clinicopathological features by screening methods shows that there were no significant differences in the tumor types (WHO 2007: P = 0.2, chi-squared, not shown); however, the absence of sequencing data in 147 cases (Fig. 1C) resulted in 43 diagnoses in one of the “not otherwise specified” (NOS) categories18 because we were unable to definitively exclude a noncanonical IDH mutation (Table 1). There were no significant differences in the WHO grade (P = 0.59, chi-squared) or the distribution of cases of patients ≥55 years or younger (P = 0.82, Fisher’s exact test). A nearly significant difference in median age (P = 0.05, t-test) was present due to the higher average age of patients tested by immunohistochemistry alone (63 y); however, there was no significant difference in average age of patients tested by DNA alone versus those tested by both methods (P = 0.6). Therefore, the observed order practice showed no distinct preference to screen by one modality versus both based upon age, WHO grade, or tumor type. As would be expected, the majority of the 680 study cases were WHO grade IV glioblastoma (64%), whereas 36% were diagnosed as WHO grade II or III, lower-grade gliomas (see Table 1). Based on the clinicopathological features and distribution and fraction of glioblastomas and lower-grade gliomas, we consider our study cohort representative of primary brain tumors that would qualify for IDH testing in clinical practice.
IDH1 Immunohistochemistry Is a Perfect Screening Tool for p.R132H Variants
Given that we sought to determine the most cost-effective method for optimal IDH variant detection, further analysis focused on the subgroup of samples that had been tested by both immunohistochemistry and DNA sequencing. Of these 343 cases, an IDH variant was detected in 104 tested by sequencing (30%), while immunohistochemical screening detected 84 variant cases (25%) (Fig. 1C). Review of the 20 discrepant cases that tested positive for an IDH variant by sequencing, but not by immunohistochemistry, revealed that all 20 cases were either noncanonical IDH1 variants or variants in IDH2 (Table 2). Therefore, IDH1 p.R132H-specific immunohistochemistry had a 100% sensitivity and specificity to detect canonical IDH1 variants.
Table 2.
Demographics in patients assessed by combined testing
| Features | DNA + IHC+ (Canonical IDH1 p.R132H variants) N = 84 |
DNA + IHC– (Noncanonical IDH1 and IDH2 variants) N = 20 |
DNA– IHC– (No IDH1/2 variants) N = 239 |
|---|---|---|---|
| Tumor type (WHO 2016 classification) | |||
| Glioblastoma | 19 (23%) | 3 (15%) | 198 (83%) |
| Glioblastoma, IDH wild-type | – | – | 198 (83%) |
| Glioblastoma, IDH-mutant | 19 (23%) | 3 (15%) | – |
| Glioblastoma, NOS | – | – | – |
| Lower-grade glioma (WHO grade II/III) | 65 (77%) | 17 (85%) | 41 (17%) |
| Anaplastic astrocytoma, IDH wild-type | – | – | 32 (13%) |
| Anaplastic astrocytoma, IDH-mutant | 31 (37%) | 11 (55%) | – |
| Anaplastic astrocytoma, NOS | – | – | – |
| Anaplastic oligodendroglioma, IDH-mutant, 1p/19q codeleted | 10 (12%) | 1 (5%) | – |
| Anaplastic oligodendroglioma, NOS | – | – | – |
| Diffuse astrocytoma, IDH wild-type | – | – | 9 (4%) |
| Diffuse astrocytoma, IDH-mutant | 16 (19%) | 5 (25%) | – |
| Diffuse astrocytoma, NOS | – | – | – |
| Oligodendroglioma, IDH-mutant, 1p/19q codeleted |
8 (10%) | – | – |
| Oligodendroglioma, NOS | – | – | – |
| WHO grade | |||
| IV | 19 (23%) | 3 (15%) | 198 (83%) |
| III | 41 (49%) | 12 (60%) | 32 (13%) |
| II | 24 (29%) | 5 (25%) | 9 (4%) |
| Age at diagnosis, y | |||
| Median; average (range) | 46; 45* (23–87) | 35; 37* (22–54) | 60; 59 (9–86) |
| <55 | 62 (74%) | 20 (100%) | 73 (31%) |
| ≥55 | 22 (26%) | – | 166 (69%) |
| IDH1-variant p.Arg132 | Canonical | Noncanonical | WT |
| His | 84 | – | – |
| Cys | – | 10 | – |
| Gly | – | 4 | – |
| Ser | – | 2 | – |
| Leu | – | 1 | – |
| IDH2-variant p.Arg172 | |||
| Ser | – | 1 | – |
| Lys | – | 2 | – |
Abbreviations: DNA, tumors tested by sequencing assay; IHC, tumors tested by IDH-R132H specific immunohistochemistry; NOS, not otherwise specified (specific diagnosis in the WHO 2016 classification); +, positive/variant detected; −, negative/variant not detected. *Patients with noncanonical IDH mutations are significantly younger compared with patients with canonical IDH1 p.R132H mutations (P = 0.003, t-test); distribution of grades between the 2 groups showed no significant differences (P = 0.633, chi-square).
Clinicopathological Features by IDH Genotype
Examination of the 239 tumors that tested negative by both immunohistochemistry and sequencing (Table 2) showed a significantly higher proportion of WHO grade IV tumors (83%) compared with those cases positive by one (15%) or both methods (23%; P < 0.0001, chi-square). These results are in concordance with the known occurrence of IDH variants in lower-grade gliomas relative to their rare occurrence in glioblastomas.1 Patients testing positive for any IDH variant (combined immunohistochemistry positive and sequencing positive, or immunohistochemistry negative and sequencing positive) had a lower average age (45 and 37 y, respectively) than those patients testing negative for an IDH variant by both detection methods (59 y; Table 2). These findings are in agreement with the known predilection for IDH variants in gliomas from younger patients.5 Interestingly, the age of patients testing positive by both methods (ie, canonical R132H-IDH1 variants) and those testing positive by sequencing only (ie, noncanonical IDH1 and IDH2 variants) was significantly different in our cohort. Specifically, patients with noncanonical variants were on average 8 years younger (P < 0.01, t-test; Table 2).
Detection Rate by Age
To further investigate these age differences of patients with canonical versus noncanonical variants, we plotted variant detection rate by each method (immunohistochemistry and sequencing) over time at 5-year age intervals (Fig. 1D). Consistent with the significantly younger average age of patients with noncanonical IDH1 and IDH2 variants, sequencing significantly outperformed immunohistochemistry in patients <55 years of age (P < 0.05, t-test; Fig. 1C). Interestingly, at 55 years and above, immunohistochemistry and sequencing had nearly equivalent detection rates, indicating essentially that all variants in these later age groups were canonical R132H-IDH1 variants, or variants that could be detected by immunohistochemistry alone. In this cohort, therefore, sequencing in patients ≥55 years in search of noncanonical IDH1 and IDH2 variants was neither necessary nor cost-effective.
Detection Rate of Noncanonical Variants
To investigate the age distribution of noncanonical IDH1 and IDH2 variants in a larger study set, we expanded the study population to include all cases (N = 482) that had been sent for sequencing, regardless of whether immunohistochemistry testing was done. In the subset of cases tested by sequencing, we identified 47 additional IDH variants (n = 41 canonical and n = 6 noncanonical). Specifically, we found 10 canonical and 4 noncanonical variants, adding up to a total of 29 and 7, respectively, in the glioblastoma group (Fig. 2A), and we found 31 canonical and 2 noncanonical variants adding up to 96 and 19, respectively, in the grade II/III glioma group (Fig. 2B). Consistent with the results in the combined testing group, noncanonical IDH1 and IDH2 variants were significantly rarer in patients ≥55 years (P < 0.0001, Fisher’s exact test), with only 2 noncanonical variants present in a total of 253 cases (0.8%) compared with 26 in 229 patients (11.4%) under the age of 55 (P < 0.0001; Fig. 2C). Additionally, when incorporating WHO grade, noncanonical IDH1 and IDH2 variants were rare exceptions in older patients with glioblastoma (2 detected in 206 glioblastomas, 0.9%; Fig. 2A, 2D) and in our series were absent in lower-grade lesions in patients ≥55 years (0 detected in 47; Fig. 2B, 2D). Thus, our data on the prevalence of noncanonical IDH1 and IDH2 variants confirm the feasibility of the proposed ≥55 years age cutoff.
Fig. 2.
Detection rates by age and IDH genotype. Comparison of IDH mutations in (A) glioblastoma vs (B) WHO grade II/III glioma by age in the 482 patients assessed by next-generation sequencing; noncanonical IDH mutations are depicted in red on stacked bar-graphs (insets show breakdown of number per IDH mutation). (C) Overall comparison in the 343 patients assessed by both methods shows significantly higher prevalence of non–R132H-IDH1 and -IDH2 variants in the younger age group (P < 0.001; Fisher’s exact test). (D) Comparison of the age-based prevalence of noncanonical IDH mutations by histological diagnosis.
Modeling Cost-effectiveness
In addition to this prevalence-based, age-appropriate cutoff for IDH variant screening by sequencing, we sought to independently model the financial implications of this age cutoff. Specifically, we aimed to estimate potential cost savings, improvements in turn-around time, as well as the number of patients put at risk if applying an age-based screening algorithm to our cohort of 482 cases that had been sent for sequencing from 2010 to 2015. To construct this model we used the current in-house cost of the immunohistochemistry and sequencing screening modalities currently used at MGH for detection of IDH variants, immunohistochemistry (~$135 per case), and our NGS platform, SnapSHOT (~$1800 per case). Turn-around time was modeled at 2 days for immunohistochemistry and 14 days for the NGS assay.
Applying this model (see Materials and Methods), if all cases are screened by both immunohistochemistry and sequencing—except those cases harboring R132H-IDH1 mutations detectable by immunohistochemistry alone—the estimated direct test cost of screening all 482 cases amounts to $707670 with a total (theoretical) turn-around time of 5962 days (Table 3). When applying the age-based cost model for patients ≥55 years, a total of $403200 in NGS cost and a total (theoretical) turn-around time of 3642 days are saved. When plotted by age, the screening cost to detect one noncanonical IDH1 or IDH2 variant by immunohistochemistry and sequencing varies between $10000 and $50,000 per positive case for patients under age 55 but sharply increases to approximately $250000 per positive case detected for patients 55 years of age and older. Immunohistochemistry-based screening costs, however, are far lower, rising to a maximum of ~$50000 per positive case for patients ≥55 years (Fig. 3). For single gene sequencing (with a direct test cost estimate of ~$420), we estimated a maximum of ~$105000 per positive case for patients ≥55 years. If an age cutoff of 55 years of sequencing is applied to the model, cost and turn-around time (74% reduction) are greatly reduced in patients ≥55 years. Specifically, using the 482 cases in our cohort, the age cutoff leads to a 43% overall reduction in screening costs and a 53% reduction in overall turn-around time, while only reducing the total IDH variant detection rate by 0.4% (N = 2 of 482; Fig. 3).
Table 3.
Cost-model for optimized detection of noncanonical IDH variants*
| <55 y N = 229 |
≥55 y N = 253 |
Total N = 482 |
|
|---|---|---|---|
| IDH Status | N = | N = | N = |
| IDH wild-type | 109 | 222 | 331 |
| Noncanonical mutation | 24 | 2 | 26 |
| Subtotal | 133 | 224 | 357 |
| Canonical IDH mutation | 96 | 29 | 125 |
| Cost model (no age-based cutoff) | $ | $ | $ |
| NGS cost (~$1800) | 239 400 | 403 200 | 642 600 |
| IHC cost (~$135) | 30 915 | 34 155 | 65 070 |
| Total cost | 270 315 | 437 355 | 707 670 |
| Age-based cost model for NGS | $ | $ | $ |
| NGS cost | 239 400 | N/A (0) | 239 400 |
| IHC cost | 30 915 | 34 155 | 65 070 |
| Total cost | 270 315 | 34 155 | 304 470 |
| Savings | 403 200 (43%) | ||
| Turn-around time (TAT) | Days (avg.) | Days (avg.) | Days (avg.) |
| TAT | 2320 (10) | 3642 (14) | 5962 (12) |
| TAT ≥55 y cutoff for NGS | 2320 (10) | 506 (2) | 2826 (6) |
| Reduction | – | (74%) | (53%) |
*Cost model is based on N = 482 cases tested by next-generation sequencing and assuming that canonical IDH1 mutations can be detected using p.R132H immunohistochemistry, and when present do not require additional workup.Abbreviations: avg., average; N/A, not applicable.
Fig. 3.
Test cost per positive. Cost is modeled using the same population as in Fig. 2. Cases are binned into 5-year age groups except at the extremes. Cost represents the total cost of the cases screened ($1800 per case sequenced by NGS, $420 × 2 per case sequenced by single gene sequencing [IDH1 and IDH2], and $135 per case screened by immunohistochemistry) divided by the number of non–R132H-IDH1 and -IDH2 variants detected in that age group. The cost to detect variants that would be missed by R132H-IDH1 immunohistochemistry increases dramatically in the older age group, in which noncanonical variants are rare.
Discussion
Here we report clinical practice data from 1023 IDH tests—amounting to ~$1.09 million in direct laboratory test cost—that establish the rarity of non–R132H-IDH1 mutations in patients ≥55 years of age. Instituting an age-based cutoff for sequencing-based testing of patients ≥55 years could reduce screening costs by 43% and reduce turn-around time by 53% in glioma patients. These data should drive test method selection, and our age-based screening model reduces test cost and turn-around time to final integrated neuropathological diagnosis.
One key result of our quality improvement initiative is the level of additional IDH-variant detection when adding sequencing-based testing. In our practice, 147 patients had a negative immunohistochemistry result and were not tested by sequencing (ie, may harbor a noncanonical IDH mutation); however, this subgroup was composed primarily of older patients (median age: 65 ± 18 standard deviation) and only 36 patients were <55 years. With the most pessimistic assumption that genotyping can at most add 16% non–R132H-IDH1 mutations (derived from the 25- to 30-year-old patient subgroup; Fig. 1D), we would theoretically put maximally 25 patients at risk (2% of all tests). However, based on the 482 patients tested by both methods, we can estimate that the actual risk is less than 0.9%. Specifically, when stringently applying this age cutoff in our cohort, we would have missed 2 patients who harbored noncanonical IDH mutations and were ≥55 years (Fig. 2A). The first patient was a 61-year-old man with glioblastoma (Fig. 2A) who received a tumor resection followed by adjuvant chemoradiation and whose tumor demonstrated an aggressive nature despite the genotype (overall survival, 1 y). The second was a 63-year-old man (Fig. 2A) with an original diagnosis of anaplastic oligoastrocytoma at age 55 (reclassified in our study as glioblastoma, IDH-mutant given the lack of 1p/19q codeletion and presence of focal microvascular proliferation) who had received adjuvant chemoradiation followed by re-resection 8 years later. He survived one additional year after re-resection (overall survival, 9 y). In the second case, the 8-year survival after history of a prior, lower-grade diagnosis would have raised questions about the negative IDH1 p.R132H immunohistochemistry result and would have resulted in further sequencing in today’s practice. On the other hand, the first case would have likely been missed by the proposed age cutoff. Thus, of the 2 identified exceptions, only one case reflects the small fraction of exceptional cases that are being put at risk (~1 in 206 patients or 0.48%). Based on the rarity of these exceptions, in conjunction with prior studies focusing on detection rates,7 our cost modeling strongly suggests the implementation of the proposed age-based testing cutoff not only for financial but also for efficiency reasons. As a result of our analysis, the utilization proposition is not to attempt testing for noncanonical mutations in patients ≥55 years unless there is strong clinical suspicion, because the economic impact highly outweighs the added clinical utility. Furthermore, the second case illustrates that clinical context continues to play a vital role in interpreting genetic test results and that separation of diagnostic groups cannot fully account for the biological spectrum of the disease. While the 2 exceptions may have formally resulted in misdiagnosis, the potential clinical consequences regarding treatment would only extend to patients with lower-grade gliomas, where clinicians may choose a less aggressive therapy for IDH-mutant tumors. The absence of noncanonical IDH mutations in lower-grade glioma patients ≥55 years in our series (Fig. 2B) suggests relatively limited value in screening for noncanonical variants in these tumors. In glioblastoma, we regard our cost model and the resulting proposal as an example of leveraging meaningful test utilization in the context of efficient use of health care resources.
Based on our data, we would not recommend an age-based cutoff for IDH screening by IDH1 p.R132H-specific immunohistochemistry.14 Our findings are at first glance somewhat discrepant from a recent study by Robinson et al.14 Briefly, in their 578 tested cases, Robinson et al report 11 IDH-mutant tumors in patients ≥55 years old (1.9%).14 In our cohort (combined testing) we found 22 IDH-mutant tumors in patients ≥55 years (N = 22/343 = 6.4%; Table 2). However, the cohort by Robinson et al had a significantly higher fraction of glioblastomas (N = 420/578; 72.6%) compared with 64% glioblastomas in our series (N = 431/680; P = 0.0001). Additionally, without knowing the age distribution of their cohort, percentages are hard to compare. When reviewing the breakdown of their IDH-mutant tumors in patients ≥55, all 11 cases were the canonical p.R132H-IDH1 variant and identified by R132H-IDH1 immunohistochemistry. By specifying the age-based prevalence of canonical and noncanonical IDH mutations, we confirm that p.R132H-IDH1 immunohistochemistry screening should be performed regardless of age and that the proposed age cutoff in glioblastoma places maximally 0.4%–0.9% of patients at risk of being not identified as having tumors harboring a noncanonical IDH mutation.
Panel testing can add additional molecular-genetic information,20 and we do not argue against its use. It is important to emphasize the added value of NGS panels because panel testing allows detection of variants beyond IDH1 in a single test, including IDH2, which is crucial for determination of the IDH mutation status for integrated neuropathological diagnosis. In our cohort we found 11.5% (3/26) noncanonical IDH variants in IDH2. However, from large-scale integrated genomic analyses, we know that other variants may occur in gliomas, and the true clinical utility and added value of panel testing is beyond the scope of this discussion.
In the US health care system, 6 new glioblastoma cases are diagnosed per 100000 people each year,18 with two-thirds of these cases occurring in patients ≥55 years of age and older (http://www.abta.org/ last accessioned 3/6/2017). Extrapolating our data over this population would result in ~$23.7 million of laboratory cost savings. Furthermore, implementation of this screening paradigm will greatly reduce turn-around time for patients, in that the 99.1% of glioblastoma patients aged ≥55 years would not benefit from waiting 2 additional weeks for sequencing results over the 48-hour turn-around of the IDH1 p.R132H immunohistochemistry result. In other words, when making the decision to implement an age-based testing cutoff, one has to consider not only the small risk of the non-tested population (ie, 0.4%–0.9%), but also the unnecessary delay for those 99.1% of patients ≥55 years waiting for appropriate treatment stratification or enrollment in clinical trials within days of their surgery.
Our study is limited to one academic medical center with a high number of secondary referrals and preselection of cases. However, our hospital also serves the local population, and while primary care settings may slightly differ, when reviewing the epidemiologic data of our test cohort (Table 1) we regard this study as representative of national trends. Another factor potentially limiting our results is the potential for differing workup among the neuropathologists in our practice. An additional test that can be helpful in directing the need for sequencing in diffuse astrocytic gliomas is alpha thalassemia/mental retardation syndrome X-linked (ATRX) immunohistochemistry, as ATRX loss has been associated with IDH mutations in lower-grade astrocytomas and in glioblastomas.16,21 For example, in the setting of negative IDH1 p.R132H immunohistochemistry, the absence of ATRX nuclear expression would more strongly suggest the possibility of a noncanonical IDH1 or IDH2 mutation. Thus, we would not recommend against DNA testing in patients ≥55 years in all circumstances and we regard the age-based cutoff and our cost model as supportive of preventing overutilization in cases with low diagnostic yield. The utility of this model will ultimately require validation in prospective studies, ideally by independent groups.
In summary, these clinical practice data from 1023 assays in 680 patients have confirmed the rarity of non–R132H-IDH1 and -IDH2 mutations in patients ≥55 years of age. This rarity reinforces the 2016 WHO recommendations of not sequencing IDH in glioblastoma patients ≥55 years. These data should furthermore drive selection of the test method, and our age-based screening model reduces test cost and turn-around time and is thus one approach to balance test utilization with cost optimization in clinical practice.
Supplementary Material
Supplementary material is available at Neuro-Oncology online.
Funding
None.
Conflict of interest statement. A.J.I. shares equity in and acts as a consultant for Archer Dx.
Acknowledgments
We thank the entire clinical team of the Center for Integrated Diagnostics as well as the Neuropathology and Immunopathology services at MGH. In particular we thank Julie Batten, Hayley Robinson, Yi Cao, Caitlin E. Finn, Aymen Baig, and Amelia N. Raymond for expert technical assistance.
References
- 1. Balss J, Meyer J, Mueller W, Korshunov A, Hartmann C, von Deimling A. Analysis of the IDH1 codon 132 mutation in brain tumors. Acta Neuropathol. 2008;116(6):597–602. [DOI] [PubMed] [Google Scholar]
- 2. Hartmann C, Hentschel B, Tatagiba M et al. ; German Glioma Network. Molecular markers in low-grade gliomas: predictive or prognostic? Clin Cancer Res. 2011;17(13):4588–4599. [DOI] [PubMed] [Google Scholar]
- 3. Yan H, Parsons DW, Jin G et al. IDH1 and IDH2 mutations in gliomas. N Engl J Med. 2009;360(8):765–773. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Camelo-Piragua S, Jansen M, Ganguly A et al. A sensitive and specific diagnostic panel to distinguish diffuse astrocytoma from astrocytosis: chromosome 7 gain with mutant isocitrate dehydrogenase 1 and p53. J Neuropathol Exp Neurol. 2011;70(2):110–115. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Hartmann C, Meyer J, Balss J et al. Type and frequency of IDH1 and IDH2 mutations are related to astrocytic and oligodendroglial differentiation and age: a study of 1,010 diffuse gliomas. Acta Neuropathol. 2009;118(4):469–474. [DOI] [PubMed] [Google Scholar]
- 6. Reuss DE, Mamatjan Y, Schrimpf D et al. IDH mutant diffuse and anaplastic astrocytomas have similar age at presentation and little difference in survival: a grading problem for WHO. Acta Neuropathol. 2015;129(6):867–873. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Schumacher T, Bunse L, Pusch S et al. A vaccine targeting mutant IDH1 induces antitumour immunity. Nature. 2014;512(7514):324–327. [DOI] [PubMed] [Google Scholar]
- 8. Wang F, Travins J, DeLaBarre B et al. Targeted inhibition of mutant IDH2 in leukemia cells induces cellular differentiation. Science. 2013;340(6132):622–626. [DOI] [PubMed] [Google Scholar]
- 9. Juratli TA, Cahill DP, McCutcheon IE. Determining optimal treatment strategy for diffuse glioma: the emerging role of IDH mutations. Expert Rev Anticancer Ther. 2015;15(6):603–606. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Dimitrov L, Hong CS, Yang C, Zhuang Z, Heiss JD. New developments in the pathogenesis and therapeutic targeting of the IDH1 mutation in glioma. Int J Med Sci. 2015;12(3):201–213. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Dang L, White DW, Gross S et al. Cancer-associated IDH1 mutations produce 2-hydroxyglutarate. Nature. 2009;462(7274):739–744. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Capper D, Zentgraf H, Balss J, Hartmann C, von Deimling A. Monoclonal antibody specific for IDH1 R132H mutation. Acta Neuropathol. 2009;118(5):599–601. [DOI] [PubMed] [Google Scholar]
- 13. Louis DN, Perry A, Reifenberger G et al. The 2016 World Health Organization classification of tumors of the central nervous system: a summary. Acta Neuropathol. 2016;131(6):803–820. [DOI] [PubMed] [Google Scholar]
- 14. Robinson C, Kleinschmidt-DeMasters BK. IDH1-mutation in diffuse gliomas in persons age 55 years and over. J Neuropathol Exp Neurol. 2017;76(2):151–154. [DOI] [PubMed] [Google Scholar]
- 15. Louis DN, Perry A, Burger P et al. ; International Society Of Neuropathology–Haarlem. International Society of Neuropathology–Haarlem consensus guidelines for nervous system tumor classification and grading. Brain Pathol. 2014;24(5):429–435. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Reuss DE, Sahm F, Schrimpf D et al. ATRX and IDH1-R132H immunohistochemistry with subsequent copy number analysis and IDH sequencing as a basis for an “integrated” diagnostic approach for adult astrocytoma, oligodendroglioma and glioblastoma. Acta Neuropathol. 2015;129(1):133–146. [DOI] [PubMed] [Google Scholar]
- 17. Chen L, Voronovich Z, Clark K et al. Predicting the likelihood of an isocitrate dehydrogenase 1 or 2 mutation in diagnoses of infiltrative glioma. Neuro Oncol. 2014;16(11):1478–1483. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Louis DN, Ohgaki H, Wiestler OD, Cavenee WK. World Health Organization Classification of Tumours of the Central Nervous System.Lyon: International Agency for Research on Cancer; 2016. [Google Scholar]
- 19. Louis DN, Ohgaki H, Wiestler OD, Cavenee WK. World Health Organization Classification of Tumours of the Central Nervous System.Lyon: International Agency for Research on Cancer; 2007. [Google Scholar]
- 20. Dubbink HJ, Atmodimedjo PN, Kros JM et al. Molecular classification of anaplastic oligodendroglioma using next-generation sequencing: a report of the prospective randomized EORTC Brain Tumor Group 26951 phase III trial. Neuro Oncol. 2016;18(3):388–400. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Jiao Y, Killela PJ, Reitman ZJ et al. Frequent ATRX, CIC, FUBP1 and IDH1 mutations refine the classification of malignant gliomas. Oncotarget. 2012;3(7):709–722. [DOI] [PMC free article] [PubMed] [Google Scholar]