Predicting the likelihood of an isocitrate dehydrogenase 1 or 2 mutation in diagnoses of infiltrative glioma

We greatly appreciate the interest and feedback provided by Zou et al regarding our recent paper describing a statistical model that predicts the probability of IDH1/2 mutations in adult gliomas.

One concern raised by Zou et al was that detection of a mutation via either immunohistochemistry and/or sequencing depends on the sensitivity and specificity of each method. In particular, specimens in which only a few scattered cells are R132H IDH1 immunopositive may not have the same properties as those with many more positive cells. Assuming the immunohistochemistry protocol is optimized, there are 2 reasons for only scattered cells showing R132H IDH1 immunopositivity: (i) the surgical biopsy did not capture fully diagnostic material, and (ii)the IDH1/2 mutant glioma was in the process of losing mutant protein expression. At our institutions (University of Kentucky and University of Pittsburgh), the Dianova R132H IDH1 antibody underwent validation prior to routine clinical implementation. Antibody specificity relative to sequencing was 100%, consistent with published data by the original developers of the antibody.^1,2 However, those cohorts only used tissue blocks with sufficient tumor cells to make a reliable WHO-based diagnosis. Likewise, the current study only used cases with unequivocal glioma. Indeed, that is why we said in the Discussion, “This application is also not meant for cases in which the biopsy was suboptimal and a firm diagnosis of infiltrative glioma cannot be made.” In very rare instances in which a glioma loses mutant IDH1/2 (reviewed in³), the presence of immunostaining in a subset of obvious glioma cells should still be viewed as true positive, albeit only in that subset. Our web-based tool makes no judgment on whether or not such a tumor will ultimately behave like a typical IDH1/2 mutant glioma.

Zou et al also seem to be contending that the web-based application has only “limited utility” in both clinical and research settings because reflex testing should be standard for all immunonegative gliomas. We respectfully disagree for 3 reasons. First, virtually all pathology laboratories have immunohistochemical capability, but not all have the capacity to do sequencing. Or the lab may have sequencing capability but does not process enough gliomas to make in-house development of IDH1/2 sequencing worthwhile. This is especially true in underserved rural or semirural areas. Such centers must send the specimen to a referral lab for testing. In the United States, these costs are usually borne by the sendout lab because insurance reimbursement is difficult to obtain for anything not done in-house. Second, even in centers with sequencing capability, insurance companies often require preauthorization or they will not reimburse the lab. Being able to provide an evidence-based rationale may therefore be helpful. We personally agree that a 20% mutation probability still justifies sequencing, but as we said in the Discussion, “we cannot recommend a specific percent cutoff above which a tumor should always be screened for a mutation. For now, that will depend on the judgment of each user.” Third, the statement “it is highly unlikely that any researcher will base their decision to immunostain or sequence the sample on what the model predicts” may not reflect the reality of cost-conscious research in many parts of the world. For example, in the biospecimen core facility I direct for the Markey Cancer Center, each immunostained slide costs $10 (http://ukhealthcare.uky.edu/markey/biospecimen/). This is a heavily subsidized rate, yet it still adds up quickly for a junior investigator with limited pilot funds, especially if sequencing is also being considered. Furthermore, in a grant proposal that includes IDH1/2 testing of a study cohort, this application could help a scientist obtain an accurate pretest estimate of the number of cases that will be mutated and thereby perform a more robust power analysis, than would otherwise be possible without actually having tested the cohort.

Again, we thank Zou et al for their comments, as well as the Neuro-Oncology editors for the opportunity to respond. Such open and honest discussions are always essential for scientific advancement.