Noninvasive Determination of the IDH Status of Gliomas Using MRI and MRI-Based Radiomics: Impact on Diagnosis and Prognosis

Round 1

Reviewer 1 Report

I enjoyed this manuscript. A few comments:

2.2 suggested that enhancement was more common in IDH1 WT gliomas but the said median survival for CE was 465 days vs. NCET- can you please clarify this so it aligns?

2.3 last sentence talks about nonsignificant data, and that should be qualified in the sentence. Perhaps that there is a "trend" instead of an increased likelihood.

Table 1- I think it would help the reader to sort this table by variable (age, edema, contrast enhancement, etc.) instead of by study. This would make it easier to interpret results. 

3.7 can you state the percent positive and false negative rates of these studies?

 

Consider in the discussion deriving an algorithm of imaging combinations that would be most predictive of IDH1 status

 

Consider in discussion discussing how treatment of patients would change based on IDH1 status- why is this meaningful?

Author Response

Point 1 2.2 suggested that enhancement was more common in IDH1 WT gliomas but the said median survival for CE was 465 days vs. NCET- can you please clarify this so it aligns?

Response 1: I’m sorry for not expressing clearly. I will make the following explanation: Tumors were scored positive for noncontrast-enhancing tumors (nCETs) if they demonstrated any amount of nonenhancing, solid tissue, which not simply the opposite of contrast enhancement. Median survival was 780 days for GBM patients with nCET compared with 465 days without (P < .02).

Point 2 2.3 last sentence talks about nonsignificant data, and that should be qualified in the sentence. Perhaps that there is a "trend" instead of an increased likelihood.

Response 2: We are grateful for the suggestion and re-wrote the sentence in the revised manuscript as the following: Patients with wild-type IDH1 had a hemorrhage tendency compared to patients with mutant IDH1 (15 vs. 5; P = 0.286).

Point 3 Table 1- I think it would help the reader to sort this table by variable (age, edema, contrast enhancement, etc.) instead of by study. This would make it easier to interpret results.

Response 3: Thanks for your suggestion, we have made changes and added some aMRI in the manuscript.

Point 4 3.7 can you state the percent positive and false negative rates of these studies?

Response 4: We have added in our manuscript: Stadlbauer et al. analyzed imaging biomarkers for glioma patients and found that WHO grade II glioma showed areas with increased OEF (+18%, P < .001, n = 20), whereas WHO grade III and IV gliomas showed regions with decreased OEF (-54% [P<0.001, n = 21], -49% [P <0.001, n = 41]), while LGGs showed increased OEF (+18%, P <0.001, n = 20) compared with normal tissues.[78] This allowed clear differentiation between low- and high-grade glioma (AUC, 1) ,with a sensitivity of 1 and a specificity of 1 for the patient cohort. MTI had the highest diagnostic performance (AUC, 0.782; sensitivity,0.854; specificity,0.714) for differentiation between gliomas of grades III and IV among all biomarkers. CMRO2 was decreased (P = .037) in low-grade glioma with a mutated IDH gene, and MTI was significantly increased in glioma grade III with IDH mutation (P = .013) when compared with the IDH wild-type counterparts. CMRO2 showed the highest diagnostic performance for IDH gene mutation detection in low-grade glioma (AUC, 0.818; sensitivity,0.727; specificity,0.838) and MTI in high-grade glioma (AUC, 0.854; sensitivity,1; specificity,0.667) and for all WHO grades (AUC, 0.899; sensitivity,0.96; specificity,0.818) among all biomarkers.

Point 5 Consider in the discussion deriving an algorithm of imaging combinations that would be most predictive of IDH1 status.

Response 5: We have added in our conclusion: Results from the current study suggest that imaging features could be used to predict IDH1.although this needs to be confirmed in a large prospective trial, these results suggest that imaging features might be able to serve as a useful biomarker of IDH1 status. For example, in cMRI，most IDH1 mutant tumors were nCET, and the frontal lobe predilection for IDH1 tumors is notable. Edema, necrosis, and hemorrhage are associated with poor prognosis. In aMRI，the rCBVmax(PWI) is significantly correlated with IDH-mutated tumors，which significantly lower than that of IDH-wildtype. When the rADCmin(DWI) of a GBM was >0.98 , it was suggestive of an IDH-mutated GBM. Gliomas with IDH mutations tended to show a higher rADC, ADCmin and a lower maximum FA (DTI). In addition, DKI parameters (axial Kurtosis [Ka], radial Kurtosis [Kr] ,mean Kurtosis [Mk]) were significantly lower in IDH-mutated gliomas than in IDH-wild-type groups. The mechanism by which IDH mutations promote tumorigenesis has shown that 2-HG is the main signature metabolite.2-HG is normally below the sensitivity threshold of MRS (1 mM) and thus can hardly be detected, but it may become measurable due to accumulation when IDH is mutated. IDH-wildtype showed elevated maximumAPT and minimum APT signal intensities compared with IDH mutated gliomas. CMRO2 was decreased in low-grade glioma with a mutated IDH gene, and MTI was significantly increased in glioma grade III with IDH mutation when compared with the IDH wild-type counterparts. Radiomics, characterized by an entire three-dimensional tumor landscape, is based on the extraction of quantitative features from medical images. Current deep learning approaches are typically convolutional neural networks (CNNs)to identify IDH status. For IDH1 mutations, the most predictive features are as follows: well-defined tumor borders, central areas of cysts with low T1 and FLAIR suppression and minimal or absent enhancement.

Point 6 Consider in discussion discussing how treatment of patients would change based on IDH1 status- why is this meaningful?

Response 6: We have added in conclusion: A non-invasive and accurate method to predict IDH mutation may have great potential in routine clinical practice and could help with the implementation of appropriate management procedures in patients with glioma. Preoperative prediction of IDH status may potentially help in appropriate management procedures in patients, such as planning for treatments (including surgy). For example, neoplastic areas often extend beyond enhanced regions, and the application of DTI enables better definition of tumor borders, showing the areas of tumor invasion. Thus, DTI parameters can be used to guide surgeries with improved patient outcomes. Also, more aggressive and experimental treatments may be justified in patients with poor prognosis. In addition, some patients who cannot undergo surgery (such as brainstem glioma patients or patients with poor physical condition) can intervene treatment by predicting IDH status through preoperative imaging.

Reviewer 2 Report

Lower-grade gliomas (LGG), World Health Organization (WHO) grades II/III, are diffusely infiltrative tumors of the CNS. With time, these tumors typically progress to glioblastoma (WHO grade IV), which has a median survival of only 12–18 months despite treatment.

Isocitrate dehydrogenase (IDH) is a common molecular marker in glioma and is frequently used for predicting prognosis. Prior studies have shown that the prognosis of IDH-mutant gliomas is better than IDH-wild type gliomas. Development of method to detect IDH status in gliomas is important for diagnose and treatment.

 

Comments:

1) According to most countries' glioma treatment guideline, IDH mutation status should be detected by IHC or sequencing. Non-invasive methods to detect IDH mutation are under development. The most commonly studied non-invasive method is detection of IDH1 mutation in the plasma by circulating tumor cells or circulating tumor DNA. 

   The MRI or other radio/image based non-invasive IDH detection methods are still under study. However, in my option, the only reasonable mechanism for MRI based IDH status check is that detection of abnormal accumulation of 2-hydroxyglutarate within the tumor. Other tumor profiles found by MRI, such as the location of the tumor, the edema of the tumor, the tumor blood perfusion et al, have no Statistical Significance correlation to IDH status. 

  In this manuscript, the authors focus a lot on the correlation between IDH status and tumor location, tumor edema, tumor blood perfusion and diffusion... Please give reasonable mechanism to explain why these profiles can reflect IDH status.

 

2)Due to immature method and no standard protocol, determination of the IDH status by noninvasive MRI (instead by IHC or sequencing) may cause a lot of false negative and false positive during the noninvasive diagnose. This false diagnose may lead to mistake or delay treatment. Please explain more about the clinical benefit of noninvasive IDH status detection. 

   In real world medical, did noninvasive IDH detection by MRI used for patient diagnose and treatment?  

 

3)A lot of paradoxical in this manuscript. For example, in page 3, section 2.1, reference 26 indicated that IDH mutated gliomas were mainly located in a single lobe (P<0.001). But reference 32 indicated no difference between IDH mutated and IDH wildtype tumors in the frontal lobe (P=1). Which reference should be cited?  More paradoxical can be found in the following manuscript. Please correct them and keep the context consistent.

 

4) In table 1, can the authors indicated which kind of MRI or MRI based radionics was used in these studies individually?

  Moreover, Not all study cases were included in table 1, especial those about aMRI. The table is more readable than text. Can the authors summarize all articles cited in the manuscript in table 1?

 

5) No clear conclusions. According to the whole article, can the authors summarize in a world that the crucial MRI  profiles that correlation to IDH status with P value < 0.05? 

 

6)Are there difference between IDH1 and IDH2 mutation in MRI based radiomics? 

 

Minor revise:

7)Lots of semicolon (";") appear in abstract. 

8) The page numbers are incorrect.

Author Response

Point 1 Acc根据大多数国家的胶质瘤治疗指南，IDH突变状态应通过IHC或测序检测。检测 IDH 突变的非侵入性方法正在开发中。最常研究的非侵入性方法是通过循环肿瘤细胞或循环肿瘤DNA检测血浆中的IDH1突变。MRI 或其他基于无线电/图像的非侵入性 IDH 检测方法仍在研究中。然而，在我的选择中，基于 MRI 的 IDH 状态检查的唯一合理机制是检测肿瘤内 2-羟基戊二酸的异常积累。MRI发现的其他肿瘤特征，如肿瘤位置、肿瘤水肿、肿瘤血液灌注等，与IDH状态没有统计学意义相关性。在这份手稿中，

回应1：很抱歉，相关临床文献仅提供有力证据，但目前尚无明确机制解释IDH状态与肿瘤位置、肿瘤水肿、肿瘤血液灌注和扩散有关。我们总结了大量临床文献中P<0.05的结论。以下是对这些机制的推测，可能需要进一步研究。

位置：IDH1突变作为胶质瘤形成过程中的早期事件发生，我们发现具有这种突变的肿瘤特别位于额叶，特别是在侧脑室延侧延伸的周围区域，提示该区域可能富含最终导致 IDH1 突变的神经胶质瘤的细胞。人们普遍认为，大多数神经胶质瘤起源于一组神经干/祖细胞，这些细胞与脑室下区 (SVZ)、侧脑室衬里和海马体分离。SVZ 已被确定为大脑中最大的生发区域，并且经常发现脑室周围或与 SVZ 相邻的神经胶质瘤。尽管神经干细胞保留了其产生星形胶质细胞、少突胶质细胞和神经元的潜力，它们的区分能力由空间代码决定；不同类型的神经胶质瘤可能来自不同的前体细胞，这些前体细胞在开始或大脑发育过程中具有相对区域特异性。有充分证据表明，IDH1 突变的神经胶质瘤来自一个独特的“起源细胞”。研究表明，IDH1 突变的神经胶质瘤的“起源细胞”可能作为神经前体群体的一部分存在，其分化潜能有限，主要局限于延髓侧脑室的延伸，为这一假设提供了支持的放射学证据。（DOI：有充分证据表明，IDH1 突变的神经胶质瘤来自一个独特的“起源细胞”。研究表明，IDH1 突变的神经胶质瘤的“起源细胞”可能作为神经前体群体的一部分存在，其分化潜能有限，主要局限于延髓侧脑室的延伸，为这一假设提供了支持的放射学证据。（DOI：有充分证据表明，IDH1 突变的神经胶质瘤来自一个独特的“起源细胞”。研究表明，IDH1 突变的神经胶质瘤的“起源细胞”可能作为神经前体群体的一部分存在，其分化潜能有限，主要局限于延髓侧脑室的延伸，为这一假设提供了支持的放射学证据。（DOI： 10.1038/nn.3610，DOI：10.1111/ene.12578）

水肿：GBM 中的瘤周水肿被认为是血管源性的，由血管通透性增加引起，因为缺氧诱导的毛细血管形成缺乏功能性紧密连接，包含开窗和不规则的基底膜内皮细胞。破坏的血脑屏障 (BBB) 导致血浆外渗到病变周围的脑实质中。我们还没有发现描述水肿和 IDH-1 之间关联的文献，可能需要进一步研究。

肿瘤血液灌注和扩散：研究表明，IDH 突变肿瘤的特征是 EGLN 脯氨酰 4-羟化酶的表达增加，这些酶标记 HIF1A 进行多泛素化和蛋白酶体降解，导致 HIF1A 活化减少和缺氧、血管和血管生成相关信号传导的下游抑制. 这些信号通路是侵袭性肿瘤行为的先决条件，携带 IDH 突变的低级别和间变性神经胶质瘤的相对惰性过程强调了这一点，而与其野生型对应物相比，它们的预后通常与胶质母细胞瘤相似。最重要的是，IDH 突变体和野生型肿瘤之间的这些差异分子特征转化为不同的表型，在临床环境中可通过 rCBV 成像进行非侵入性预测。同样地，IDH-mut GBMs 较高的 ADC 值可以解释为 IDH-mut 肿瘤的较低血管生成潜力和细胞结构。(DOI: 10.1038/srep16238)

Point 2 由于方法不成熟，没有标准协议，通过无创MRI（而不是通过IHC或测序）确定IDH状态可能会在无创诊断过程中导致大量假阴性和假阳性。这种错误诊断可能导致错误或延误治疗。请详细解释无创 IDH 状态检测的临床益处。在现实世界的医疗中，MRI 的无创 IDH 检测是否用于患者诊断和治疗？

回应 2：我们在结论中补充说：免疫组织化学和基因组序列分析被认为是检测胶质瘤患者 IDH 突变的“金标准”方法。但是这两种方法都没有提供 IDH1 基因状态的术前检测。因此，一种无创且准确的预测 IDH 突变的方法可能在常规临床实践中具有巨大潜力，并有助于在胶质瘤患者中实施适当的管理程序。最近的研究表明，与其他分子亚型相比，大体全切除对 IDH 突变型胶质瘤更有益。尽管无论 IDH 状态如何，最大限度地切除肿瘤是标准治疗，但术前预测 IDH 状态可能有助于患者的适当管理程序，例如治疗计划（包括手术）。例如，肿瘤区域通常超出增强区域，DTI 的应用可以更好地定义肿瘤边界，显示肿瘤侵袭区域。因此，DTI 参数可用于指导手术以改善患者结果。此外，对于预后不良的患者，更积极和实验性的治疗可能是合理的。此外，一些不能进行手术的患者（如脑干胶质瘤患者或身体状况不佳的患者）可以通过术前影像学预测IDH状态来干​​预治疗。各种使用先进技术和方法的研究试图从术前图像中预测 IDH 基因突变状态；然而，在现实世界的医学中，临床上广泛使用的传统 MRI 预测更为重要，但仍然是一项具有挑战性的任务。

第 3 点 这份手稿有很多自相矛盾的地方。例如，在第 3 页第 2.1 节中，参考文献 26 表明 IDH 突变的神经胶质瘤主要位于单叶（P<0.001）。但参考文献 32 表明额叶中 IDH 突变和 IDH 野生型肿瘤之间没有差异（P=1）。应该引用哪个参考文献？在以下手稿中可以找到更自相矛盾的地方。请更正它们并保持上下文一致。

回应3：我为我们的错误道歉，我们已经修改了你提到的问题并仔细检查了全文。

第 4 点 在表 1 中，作者能否指出在这些研究中分别使用了哪种 MRI 或基于 MRI 的放射学？此外，并非所有研究案例都包含在表 1 中，尤其是关于 aMRI 的案例。该表比文本更具可读性。作者能否总结表 1 中手稿中引用的所有文章？

回应4 ：感谢您的建议，我们已经在原稿中进行了修改并添加了一些aMRI。

第 5 点 没有明确的结论。根据整篇文章，作者能否总结出与 IDH 状态相关且 P 值 < 0.05 的关键 MRI 剖面？ 

回应 5：我们在结论中添加了：当前研究的结果表明影像特征可用于预测 IDH1。尽管这需要在大型前瞻性试验中得到证实，但这些结果表明影像特征可能能够作为有用的生物标志物IDH1 状态。例如，在 cMRI 中，大多数 IDH1 突变肿瘤是 nCET，并且 IDH1 肿瘤的额叶好发性是显着的。水肿、坏死和出血与预后不良有关。在aMRI中，rCBVmax(PWI)与IDH突变肿瘤显着相关，显着低于IDH野生型。当 GBM 的 rADCmin(DWI) > 0.98 时，提示 IDH 突变的 GBM。具有 IDH 突变的胶质瘤倾向于显示更高的 rADC、ADCmin 和更低的最大 FA (DTI)。此外，DKI 参数（轴向峰度 [Ka]，IDH突变神经胶质瘤的径向峰度[Kr]，平均峰度[Mk]显着低于IDH野生型组。IDH突变促进肿瘤发生的机制表明2-HG是主要的特征代谢物。2-HG通常低于MRS（1 mM）的灵敏度阈值，因此几乎无法检测到，但由于积累可能变得可测量当 IDH 发生突变时。与 IDH 突变的神经胶质瘤相比，IDH 野生型显示出升高的最大 APT 和最小 APT 信号强度。与 IDH 野生型对应物相比，具有 IDH 基因突变的低级别胶质瘤中 CMRO2 降低，而 IDH 突变 III 级胶质瘤中 MTI 显着增加。以整个三维肿瘤景观为特征的放射组学基于从医学图像中提取定量特征。当前的深度学习方法通​​常是卷积神经网络 (CNN) 来识别 IDH 状态。对于 IDH1 突变，最具预测性的特征如下：明确的肿瘤边界、低 T1 和 FLAIR 抑制的囊肿中心区域以及最小或没有增强。

第 6 点 基于 MRI 的放射组学中 IDH1 和 IDH2 突变之间是否存在差异？

回应 6：没有 IDH1 突变的肿瘤通常具有影响 IDH2 基因类似氨基酸的突变。具有 IDH1 或 IDH2 突变的肿瘤具有独特的遗传和临床特征，具有此类肿瘤的患者比具有野生型 IDH 基因的患者具有更好的结果。IDH2（R172）基因编码唯一与IDH1（R132）同源的以NADP+作为电子受体的人类蛋白质。IDH1 突变在几乎所有神经胶质瘤亚型中都很常见，范围从 50% 到 80%，除了原发性多形性胶质母细胞瘤。IDH1 是在 II 级和 III 级胶质瘤中通常发生突变的少数基因之一。而 IDH2 突变比 IDH1 突变少得多；两者很少同时发生，但似乎都是胶质瘤特有的。这可以解释为什么很少有单独针对 IDH2 突变的临床研究。所以，目前还没有大量研究支持基于 MRI 的放射组学中 IDH1 和 IDH2 突变之间的差异。(DOI: 10.1056/NEJMoa0808710)

小修改：

 

Point 7 摘要中出现大量分号（“;”）。 

回复 7：很抱歉，我们仔细检查了摘要，没有找到分号，如果您有其他问题，请随时与我们联系。

第8点页码不正确。

回复 8：我们已经更正了您提到的格式问题。

 

Author Response File: Author Response.docx