Impact of Frontline Ivosidenib on Volumetric Growth Patterns in Isocitrate Dehydrogenase (IDH) mutant Astrocytic and Oligodendroglial Tumors

Abstract

Purpose.

Isocitrate dehydrogenase (IDH) mutant gliomas are usually treated with radiotherapy and chemotherapy, which increases the risk for neurocognitive sequelae during patients’ most productive years. We report our experience using off-label first-in-class IDH1mut-inhibitor ivosidenib and its impact on tumor volume in IDHmut gliomas.

Experimental Design.

We retrospectively analyzed patients aged ≥18 years with radiation/chemotherapy-naïve, IDH1mut, non-enhancing, radiographically active, grade 2/3 gliomas, and ≥2 pretreatment and ≥2 on-ivosidenib MRIs. T2/FLAIR-based tumor volumes, growth rates and PFS were analyzed. Log-linear mixed-effect modeling of growth curves adjusted for grade, histology and age was performed.

Results.

We analyzed 116 MRIs of 12 patients (median age 46years [range:26–60]) 10 males, 8 astrocytomas (50% grade 3), 4 grade 2 oligodendrogliomas. Median on-drug follow-up was 13.2 months (interquartile range[IQR]:9.7–22.2). Tolerability was 100%. 50% of patients experienced ≥20% tumor volume reduction on treatment and absolute growth-rate was lower during treatment (−1.2±10.6cc/year) than before treatment (8.0±7.7cc/year; p≤0.05). Log-linear models in the Stable group (n=9) showed significant growth before treatment (53%/year; p=0.013), and volume reduction (−34%/year; p=0.037) after 5-months on treatment. After-treatment volume curves were significantly lower than before treatment (after/before treatment ratio 0.5; p<0.01). Median-time-to-Best Response was 11.2(IQR:1.7–33.4) months, and 16.8(IQR:2.6–33.5) months in patients on drug for ≥1 year. PFS-9mo was 75%.

Conclusions.

Ivosidenib was well-tolerated and induced a high volumetric response rate. Responders had significant reduction in tumor growth rates and volume reductions observed after a 5-month delay. Thus, ivosidenib appears useful to control tumor growth and delay more toxic therapies in IDH-mutant non-enhancing indolently growing gliomas.

Introduction

Isocitrate dehydrogenase (IDH) mutated gliomas include WHO Grade 2 and 3 tumors that are infiltrative, slow-growing, astrocytic (AS) or oligodendroglial (OD) and occur primarily in the third or fourth decade of life.(1) IDH is an enzyme involved in the citric acid cycle pathway that catalyzes the oxidative decarboxylation of isocitrate. IDH has two isoforms, IDH1 and IDH2. A mutation in either of the isoforms leads to accumulation of the onconeometabolite D-2-hydroxyglutarate (2HG) that is thought to drive tumorigenesis through DNA hypermethylation and contribute to epileptogenesis through direct glutamatergic activity.(2,3) Unlike in cholangiocarcinoma or leukemia where IDH mutation is associated with similar or perhaps even worse prognosis than wildtype,(4,5) in gliomas mutant IDH is a robustly favorable prognostic and predictive marker with many folds longer median survival than IDH wildtype diffuse gliomas of any grade and is now integral part of the WHO classification of CNS tumors.(6,7) Mutant IDH is a desirable druggable target due to its persistence throughout the disease course in glioma. Inhibitors of mutant IDH are approved for systemic cancers such as leukemia,(8–11) and are currently being investigated in glioma. Ivosidenib (AG-120) is a first-in-class mutant IDH1 inhibitor with a favorable pharmacokinetic profile that includes extremely high potency (nanomolar scale) and adequate penetration past the intact blood-brain barrier in animal model.(12,13) Ivosidenib also has promising safety and efficacy data in patients with IDH1 mutant gliomas according to preliminary findings published from ongoing clinical trials (NCT02073994).(14–16)

Due to the indolent clinical course and longer survival times in IDH mutant glioma, high level evidence for efficacy based on overall survival endpoints from ongoing clinical trials are expected to take more than a decade.(17,18) In the interim, rigorous radiographic assessments of treatment responses has the potential to inform clinical decisions. One- or two-dimensional lesion measurements using RECIST or RANO criteria have highly limited intra- and interrater reliability even when applied to contrast enhancing lesions due to variations in tumor size, shape and growth patterns.(17–19) Emerging evidence corroborate that volumetric MRI data provide higher interrater agreement and less intermeasurement error for response assessment leading to clinically more meaningful information compared to 2D RANO analyses in non-contrast enhancing gliomas.(16,21) Increased access to lesion segmentation tools has enabled a new framework that integrates volumetric growth trajectory changes as a method to improve sensitivity and shorten the time needed to establish anticancer drug efficacy.(16,22–25) These approaches are now supported by recent CNS clinical trial data as well.(21,26)

In clinical practice, IDH mutated gliomas come with unique challenges. Following maximum safe tumor resection, the ASCO-SNO guidelines recommend neuro-oncologists to offer radio- and chemotherapy to most patients with IDH mutated gliomas.(27) However, the optimal timing of treatment initiation is less clearly established by evidence and is thus subject to clinical interpretation that may vary on a case-by-case basis. Decision-making includes weighing expected survival benefit provided by the treatment, versus the present evidence and expected disease activity and longitudinal toxicities of both radiotherapy and chemotherapy. As most people diagnosed with IDH mutated gliomas are working age and expected to survive for many years, often decades, treatment-induced neuro-cognitive side-effects impact employment,(28,29) and general quality-of-life.(30) Therefore, in a select set of individuals, active surveillance, when possible, is preferred to avoid the delayed neurotoxicities of radiation and cytotoxic chemotherapy. Well-tolerated alternatives to radiation-based therapies would be preferential to continuing active surveillance for patients with asymptomatic, indolently growing IDH mutated gliomas.

Since 2018, we have utilized ivosidenib as off-label therapy in patients harboring asymptomatically enlarging non-contrast enhancing IDH mutated grade 2/3 gliomas. In this single institution retrospective study, we share our clinical experience with ivosidenib tolerability and efficacy as assessed by 3D volumetrics. We highlight the real-world efficacy of ivosidenib therapy for non-enhancing grade 2/3 IDH1 mutated gliomas, response timeframes and possible pitfalls and its potential role as a non-toxic first-line therapy for indolently active IDH1 mutant glioma.

Patients and Methods

This is a retrospective study of patients who have received ivosidenib therapy (500 mg once daily) under the care of physicians within the Brain Cancer Program at Johns Hopkins Hospital 2018–2022. Inclusion criteria were: i) Age > 18 years; ii) histomolecular diagnosis of a diffuse glioma (astrocytoma [1p/19q intact] or oligodendroglioma [1p/19q co-deleted]); iii) documented IDH1 gene mutation on sequencing or IHC analysis; iv) evidence of radiographically active tumors; v) additional tumor resection not advised; vi) T2/FLAIR hyperintensity lesion of ≥1 cm); vii) ≥2 pretreatment and ≥2 on-treatment MRIs. Patients with prior history of anti-tumor therapy with radiation or chemotherapy or T1w Gd enhancement were excluded. Patients were not treated with glucocorticoids during the MRI scanning intervals shown with the exception of brief perioperative dexamethasone courses for patients who underwent surgery ≥3 months prior to starting ivosidenib. CBC with differential, comprehensive metabolic panel and EKG were available for all patients approximately every 3–4 months during ivosidenib therapy.

Image analysis

The segmentation process occurred in three-steps to minimize bias. First, FLAIR images were collected, anonymized and loaded into 3D slicer 4.11 (Slicer.org, (22) for preprocessing. Preprocessing included randomization of image acquisition order, bias correction (31), and co-registration (DOK and JS). Second, a physician (SP) blinded to patient identity and image acquisition date performed semiautomatic tumor segmentation. This was performed by first setting a signal intensity that encompassed the abnormal-appearing voxels but excluded normal-appearing brain, followed by slice-wise segmentation using the level-tracing tool of 3D Slicer to select the FLAIR hyperintense voxels deemed to be consistent with tumor. These segmentations were manually adjusted at lesion edges and at hypointense tumor voxels as needed. Third, a second physician with >10 years of experience in lesion segmentation (DOK) reviewed the segmentations and performed minor corrections in the segmented volumes for quality control. Volumetric data then was exported into a spreadsheet template that reidentified the dates of images and graphed the volumetric information over time relative to ivosidenib dosing. Annualized absolute tumor growth rates (AGR; CC/year) were estimated for: (1) ≥1 year before the start of ivosidenib (baseline); (2) the first year; (3) the entire time on ivosidenib.

Statistical analysis

Using the volumetric data, we generated the following outcome variables: i) initial response (IR), defined as the first time point with lower growth rate than the previous 2 intervals; ii) best response (BR), defined as timepoint with lowest volume at least 3 months after the initiation of ivosidenib; iii) growth arrest (GA), defined as the first timepoint preceded by no interval growth (i.e., annualized growth rate ≤0 CC/yr and ≤0%/yr) that is confirmed on a subsequent time point. IR, BR and GA rates and median times to achieve these were described. Pair-wise comparison of before- vs. on-treatment tumor volume was performed. Volumes and growth rates were compared between subgroups including grade (2 vs. 3), histopathology (Astrocytoma vs. Oligodendroglioma), and between groups dichotomized based on median age (≥45 vs. <45 years). Normality was evaluated via the Kolmogorov-Smirnoff test. Paired T-tests and 2-tailed T-tests were performed for normally distributed variables, and Wilcoxon’ ranked and Mann-Whitney U tests for non-normally distributed variables.

Growth curve analysis was performed using log-linear mixed effects models weighted by follow-up time by patient. Tumor growth rate was separated before or after Ivosidenib therapy. To assess for a delayed treatment effect, the cut-off point was tested every month from the start of Ivosidenib therapy to 6 months after Ivosidenib therapy. Both log-tumor volume and tumor growth rates were modeled as random-effects. Age, WHO grade 2 vs. 3, resection and astrocytoma vs. oligodendroglioma were modeled as fixed effect. Models were tested in all 12 patients, and in the Stable/Responder group of 9 patients and in the Progressor group of 3 patients.

Progression free survival (PFS) was established radiographically where progressive disease was defined as ≥40% increase in T2/FLAIR volume, stable disease as <40% increase or >65% decrease in volume, and partial response as <100% but >65% decrease in volume as previously described, recommended and implemented.(32,33,21) As none of the patients reached the above-defined threshold of partial response, we refer to patients with 2 or more consecutive scans with reduced volume after treatment initiation as “responders” in the manuscript. Kaplan-Meier curves were generated for PFS and subgroups were compared using the log-rank test. Threshold of statistical significance was defined as p≤0.05. Statistical analysis was carried out using SPSS 28.0 (IBM, Armonk, NY). Written consent was waived for this retrospective analysis, which has been approved by the Johns Hopkins School of Medicine Institutional Review Board (IRB00304806).

Data availability statement

The data supporting this study are available within the paper and its Supplementary Data files. All other data are available from the authors upon reasonable request.

Results

A total of 116 scans across 12 patients (10 male, median age 46 years [range 26–60], grade 2: 4 astrocytomas, 4 oligodendrogliomas; grade 3: 4 astrocytomas) were studied covering a median time of 13.2 months (inter quartile range 9.7–22.2) on ivosidenib therapy. Demographic characteristics are summarized in Table 1 (additional data in Suppl Table 1.). Ivosidenib was 100% tolerable, with only one patient experiencing fluctuating diarrhea (CTCAE Grade 1). No cardiac or hematologic side effects were observed. There were 3 discontinuations (25%), all after being on drug for at least 1 year. These were for clinical progression (i.e. increased seizure frequency, N=1) or qualitative radiology interpretation of progression (N=2).

Table 1.

Summary of demographics.

Demographics	N (%)
Patients	12
Age (median [range])	46 (26–60) years
Male / Female	10 (83%) / 2 (17%)
Histopathological type and WHO Grade
Astrocytoma Grade 2, IDH1mut	4 (33%)
Astrocytoma Grade 3, IDH1mut	4 (33%)
Oligodendroglioma Grade 2, 1p/19q co-del, IDH1mut	4 (33%)
Total Scans (pretreatment, on-treatment)	116 (71, 45)
Follow up on ivosidenib (median [IQR])	13.2 (9.7–22.2) months
Patients with ≥1 year of treatment	8 (67%)
Discontinuation of ivosidenib	3 (25%)
for clinical progression	1 (8%)
for radiographic progression	2 (17%)
Adverse events
Grade 1 diarrhea	1 (8%)

IQR: interquartile range

Volumetric results and findings at best response

Mean FLAIR hyperintense tumor volumes were 29.5±32.1cc before treatment, 25.3±22.0cc at the last on-treatment follow-up, and 23.8±22.5cc at BR (Wilcoxon’s p≥0.2; Table 2). Median time to BR was 11.2 months (range 1.7–33.4). Among patients who received more than one year of ivosidenib, median time to BR was 16.8 months (range 2.6–33.5). BR was stable disease in all cases with mean volume changes of −5.7±10.8cc or −0.02±37.9% relative to baseline. Eight patients (66.7%) had lower tumor volume at BR than before the treatment started, with 50% of patients having 20% or higher volume reduction at BR (Figure 1A and 1B).

Table 2.

Whole study cohort as well as subgroup volumetric and growth rate results.

Group		Volumes (cc)			Absolute Growth Rate (AGR; cc/yr)
		Baseline	Last on Tx	Best Response	Before treatment	First year	Entire duration
Whole Group (n=12)	Mean	29.5	25.3	23.8	8.0	−0.2	−1.2
	±SD	32.1	22.0	22.5	7.7	7.9	10.6
	p-value	—	ns	0.09	—	0.05	0.05
Grade 2 (N=8)	Mean	36.4	28.5	27.6	7.9	−1.9	−5.6
	±SD	37.2	24.8	25.6	8.4	8.3	7.4
	p-value	—	ns	0.07	—	0.09	0.04
Grade 3 (N=4)	Mean	15.8	18.8	16.2	8.0	3.2	7.5
	±SD	13.3	16.1	14.2	7.3	6.9	11.5
	p-value	—	ns	ns	—	ns	ns
Small Tumors (n=6)	Mean	7.4	10.2	8.0	2.1	3.2	2.1
	±SD	5.1	6.5	5.1	1.6	4.9	5.0
	p-value	—	ns	ns	—	ns	ns
Large Tumors (n=6)	Mean	51.7	40.4	39.7	13.8	−3.7	−4.5
	±SD	32.7	21.9	21.9	6.9	9.3	14.1
	p-value	—	ns	0.07	—	0.01	0.03

Large tumor defined as ≥14 cc at baseline based on the studied populations median.

Figure 1.

Waterfall plot of volume change from baseline to best response for each patient in percentage (A), and cubic centimeters (B). Patient IDs are indicated below or above the bars and the color code assigned to each patient is maintained across Figures 1A, 1B and 2A.

Growth dynamics and trajectory analysis

AGR curves before and on therapy are shown in Figure 2A (supporting data in Suppl. Table 2). Three patients (25%) were Progressors given their lack of growth rate reduction and eventual >40% increase in volume. Nine patients (75%) achieved growth arrest after a median of 11.8 months (3.4–17.4) with first growth rate reduction being noted 4.9 months (range 2.6–15.2) after treatment initiation. AGR was significantly lower compared to the pretreatment baseline (8.0±7.7 cc/yr) at the end of first year on treatment by −0.2±7.9 cc/yr (p<0.05) and was lowest at the last on-treatment scan (−1.2±10.6 cc/yr p=0.05; Table 2).

Figure 2.

Absolute growth rate curves including pretreatment and on-treatment time points for each individual patients (A). Dashed line represents the model fit of all datapoints from between 3 years before and 2 years on treatment. Growth rates model fit curves (B) for the entire population (black line), the Stable/Responder group (blue line), and Progressor group (red line). P-values represent the significance level of the difference between the before-treatment and on-treatment growth rate

Log-linear mixed effects modeling of growth rates in the entire study group revealed non-significant trends with tumor growth (29%/year) before, and tumor reduction (−6%/year) after ivosidenib with a non-significant treatment effect (i.e., on-treatment / before-treatment ratio of 0.73; Figure 2B and Suppl Table 2.). In the Stable group (n=9, 75%), there was significant tumor growth before ivosidenib (53%/year, p=0.013) and significant tumor reduction beginning at 5-month of therapy (−34%/year; p=0.037). In this group, the relative growth rates (on-treatment / before-treatment) was highly significant at every cut-off time point tested (after/before growth rate ratios 0.54 to 0.44; p< 0.01).

Subgroup analysis

Comparison between patient subgroups showed that oligodendrogliomas took nearly 3-fold less time to achieve growth arrest than astrocytomas (4.2±0.8 vs. 12.4±4.2 months; one-sided independent T-test p<0.01). No statistically significant volumetric or growth dynamic differences were found in the comparisons of all Grade 2 versus Grade 3 tumors; Grade 2 versus Grade 3 Astrocytomas; groups dichotomized at the median age (45 years).

PFS Analyses

PFS at 6, 9 and 12 months were 83%, 75% (n=12) and 70% (n=10), respectively. Median PFS was not reached, mean PFS was 26.4 months (CI95% 19.7–33.0). Similar PFS6 and mean PFS (82% and 28.0 [CI95% 21.6–34.5] months, respectively; n=11) resulted when using an additional 3-month follow-up MRI to confirm progression. Log-rank test revealed no difference between age, grade, or histopathologic groups (Figure 3).

Figure 3.

Kaplan-Meier curves for PFS comparing WHO grade 2 and grade 3 tumors

Discussion

In this volumetric study of ivosidenib therapy for radiographically active, radiation- and chemotherapy-naïve IDH1-mutated lower-grade gliomas, we found high response rates most evident after significant delay. Most tumors had a persistent net decrease in volume as best response which required nearly a year to realize for most and continued tumor volume reductions were observed as far as 33 months on treatment (last time evaluated). Log-linear growth curve modeling showed significant blunting of growth rates in stable/responder patients after treatment initiation and volume reductions became significant after 5 months on treatment. Oligodendrogliomas required less time to respond than astrocytomas. PFS at 9 months was 75% and larger tumors had superior responses demonstrated by higher growth rate reductions.

To our knowledge this is the first report applying quantitative volumetric imaging to evaluate a real-world experience with ivosidenib in glioma patients. Our findings lead to several important conclusions regarding ivosidenib as a promising new clinical treatment option for patients with IDH1mut glioma. Firstly, we show that ivosidenib can arrest tumor growth and/or induce tumor regression in patients with grade 2/3 IDH1 mutated glioma with the potential to delay the transition to radiation-based therapies beyond that achieved by active surveillance alone. As these patients tend to be working age with longer life-expectancies compared to patients with IDH-wildtype glioma, they are at higher risk of the long-term cognitive side-effects of combination chemotherapy and radiotherapy that often manifest as cognitive impairment causing difficulties at the workplace.(28,29) Thus, extending the years of uncompromised cognitive performance has substantial quality of life and economic benefits for these patients. Secondly, there was an average of 5 months to exhibit consistent growth rate reduction in response to ivosidenib and in one responder tumor stabilization was not realized until nearly a year after treatment initiation. Thus, premature discontinuation may be an inherent pitfall of using IDH inhibitor therapy in gliomas, and conventional 2D RANO measurements might not capture responses adequately as also described recently by Ellingson at al.(21) Lastly, patients with larger tumors and thus at higher risk for neurocognitive decline from larger field radiation are shown to also benefit from this treatment. Our volumetric findings extend those from a Phase I trial expansion cohort of 66 patients reporting a 2.9% objective response rate based on RANO LGG analysis of on-treatment MRIs(15). Additionally, the average growth rate reductions in 75% of ivosidenib-treated patients reported here are 2-fold higher than that observed in untreated lower-grade non-enhancing gliomas and comparable to those who underwent chemotherapy, radiation or both reported by Huang et al (after/before treatment rate 0.54 vs. 0.67, respectively). (16) Furthermore, spontaneous tumor regressions were not observed by Huang et al, in contrast to the 50% rate of on-treatment tumor regression in our present study. Our conclusions are strengthened by the use of a second MRI scan to reduce the impact of noise on measurement variability and growth trend analyses.

Our present observations raise some potentially clinically relevant hypotheses. We observed focal changes of FLAIR hyperintensity patterns in the core and leading edge of some of the tumors after the initiation of ivosidenib. A possible explanation for these changes could be heterogeneous intratumor responses to ivosidenib within a single patient’s tumor. This implies that careful analyses of voxel-wise tumoral MR signal intensity changes using radiomic techniques may be a potential approach to predict responses or development of resistance to ivosidenib earlier than possible using FLAIR signal volumetric assessment alone. Our observation that larger and more rapidly growing tumors appear to respond at a much higher rate appears counterintuitive; however, a possible explanation is higher tumor production of and reliance on 2-hydroxyglutarate and thus higher sensitivity to IDH inhibition. This hypothesis is currently being tested by our group, utilizing MR spectroscopy before and on ivosidenib to assess the change in 2-hydroxyglutarate and or metabolites as a marker of treatment response.

Limitations

As discussed above, this study is not without limitations. Firstly, the data was retrospectively collected and included FLAIR imaging acquired using various scanners and resolutions. Additionally, the sample size is relatively low and did not include interindividual controls. However, this limitation is compensated by the long-term pre- and on-therapy follow up for each patient and the large number of datapoints that were utilized in the analysis. This revealed information regarding growth trajectory effects in the context of IDH inhibition that other studies had not yet assessed to our knowledge. Additional molecular data, such as CDKN2A/B alterations were not readily available for all patients. While ivosidenib therapy appeared to be very well tolerated, due to the retrospective design, low-grade toxicity may be underestimated since all side effects may not have been captured in the patient record. Similarly, compliance that impacts efficacy and side effect assessments were not formally recorded in the patient records. However, the limited tolerability data presented here is congruent with what was found in the Phase 1 trial for ivosidenib in glioma, where grade ≥3 adverse events were only seen in 3% of patients (15). Lastly, this study did not include a neurocognitive component, which will require further follow-up and study in the future.

Conclusions

Ivosidenib was safe and 100% tolerable in this cohort of patients with promising efficacy in grade 2/3 non-enhancing gliomas. Responses required 5 months to be volumetrically detectable with responses becoming most evident after nearly a year of treatment. The response kinetics indicate that IDH targeting therapy is best reserved for patients considering further surveillance with low risk for developing neurologic deficits should further indolent tumor growth occur during ~6–12 months required to fully assess response. These results inform clinicians and glioma patients about the expectations and time commitments on therapy with ivosidenib or possibly other IDH inhibitors currently under development based on analyses that would have been difficult to capture using current conventional clinical trial designs. This study further supports including ivosidenib as a treatment option for non-enhancing grade 2/3 IDH1 mutated gliomas and the integration of volumetric growth trajectories in both clinical trial design and clinical practice in lower grade gliomas.

Translational Relevance.

This manuscript reports for the first time the real-world efficacy of first-line ivosidenib therapy in patients with IDH1 mutant gliomas. These findings support a new treatment option that allows patients with Grade 2/3 IDH-mutated gliomas, including patients with large tumors that would require larger radiation field, to postpone (or potentially avoid) radiation-based therapy and its well-established neuro-cognitive toxicities. We also show the power of using serial quantitative 3-dimensional tumor growth kinetics, pretreatment and during treating, to clinically evaluate non-enhancing glioma responses to a novel therapeutic.