Received: 28 November 2019 | Revised: 6 March 2020 | Accepted: 8 March 2020

DOI: 10.1111/cas.14387

ORIGINAL ARTICLE

Utility of isocitrate dehydrogenase 1 as a serum protein

biomarker for the early detection of non-small-cell lung cancer:
A multicenter in vitro diagnostic clinical trial

Nan Sun1 | Shouguo Sun1 | Yibo Gao1 | Yuan Li1 | Zhiliang Lu1 | Zuyang Yuan1 |
Yun Che1 | Jianbing Huang1 | Shuangshuang Mao1 | Yuanyuan Lei1 |
Ruochuan Zang1 | Ning Li1 | Wei Cui2 | Jun Qi2 | Feng Chen2 | Jia Gao2 |
Jinling Wang3 | Rong Min3 | Yan Chen4 | Guangli Shi4 | Fengwei Tan1 | Jie He1
1
Department of Thoracic Surgery, National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical
Sciences and Peking Union Medical College, Beijing, China
2
Department of Clinical Laboratory, National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical
Sciences and Peking Union Medical College, Beijing, China
3
Department of Clinical Laboratory, Xuanwu Hospital, Capital Medical University, Beijing, China
4
Department of Clinical Laboratory, Beijing Chest Hospital, Capital Medical University, Beijing, China

Correspondence
Fengwei Tan and Jie He, Department of Abstract
Thoracic Surgery, National Cancer Center/ We aimed to verify the expression status and diagnostic significance of isocitrate de-
National Clinical Research Center for
Cancer/Cancer Hospital, Chinese Academy hydrogenase 1 (IDH1) in non-small-cell lung cancer (NSCLC), especially during early
of Medical Sciences and Peking Union stages. Serum IDH1 levels were measured by ELISA. A total of 1223 participants (660
Medical College, 17 Panjiayuan Nanli,
Chaoyang District, Beijing 100021, China. patients with NSCLC, 276 healthy controls [HCs], 95 patients with benign pulmonary
Email: [email protected] (FT); prof. conditions [BPCs], 135 patients with other cancers [OCs], and 57 samples with inter-
[email protected] (JH)
fering factors) were divided into a training cohort and a validation cohort according
Funding information to 3 testing centers. The IDH1 concentrations in the NSCLC group were obviously
National Key R&D Program of China,
Grant/Award Number: 2018YFC1315000, higher than those in the control groups (P < .001). Area under the receiver operating
2018YFC1315003 and 2016YFC0905400; characteristic curves (AUCs) for discriminating NSCLC patients from controls (HC,
National Natural Science Foundation of
China, Grant/Award Number: 81871885; BPC, and OC) were 0.870 and 0.745 (sensitivity, 63.3% and 55.0%; specificity, 86.8%
PUMC Youth Fund, Grant/Award Number: and 86.3%) in the training cohort and validation cohort, respectively. The AUCs for
2017320013; Non-profit Central Research
Institute Fund of Chinese Academy of discriminating stage 0-IA lung cancer patients from HCs were 0.907 and 0.788 (sen-
Medical Sciences, Grant/Award Number: sitivity, 58.6% and 59.1%; specificity, 92.9% and 89.3%) in 2 cohorts, respectively.
2018RC320010; National Key Basic
Research Development Plan, Grant/ Isocitrate dehydrogenase 1 showed specificity for NSCLC and had no diagnostic
Award Number: 2016YFC0905400; CAMS value for other common cancers. Furthermore, IDH1 was significantly reduced in
Innovation Fund for Medical Sciences,
Grant/Award Number: 2017-, I2M, -1-005,

Abbreviations: AAb, autoantibody; ADC, adenocarcinoma; AUC, area under the curve; BPC, benign pulmonary condition; CI, confidence interval; CT, computed tomography; ECLS,
EarlyCDT-Lung Test; HC, healthy control; IDH1, isocitrate dehydrogenase 1; IQR, interquartile range; LDCT, low-dose computed tomography; miRNA, microRNA; MSC, miRNA
signature classifier; NLST, National Lung Screening Trial; NSCLC, non-small-cell lung cancer; OC, other cancer; OD, optical density; ROC, receiver operating characteristic; RT, room
temperature; SCC, squamous cell carcinoma; SCLC, small-cell lung cancer; αKG, alpha ketoglutarate.
Nan Sun, Shouguo Sun, and Yibo Gao contributed equally to the work.

This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in
any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.
© 2020 The Authors. Cancer Science published by John Wiley & Sons Australia, Ltd on behalf of Japanese Cancer Association.

Cancer Science. 2020;111:1739–1749.  wileyonlinelibrary.com/journal/cas | 1739

1740 | SUN et al.

2016-, I2M and -1-001; Special Foundation

for Central Committee Health Care, Grant/ postoperative serum. Isocitrate dehydrogenase 1 shows clinical utility as a serum pro-
Award Number: W2017BJ39 tein biomarker for the early diagnosis of NSCLC.

KEYWORDS

blood biomarker, diagnosis, early detection, IDH1, NSCLC

1 | I NTRO D U C TI O N and bronchoscopy, and blood markers are thus ideal aids. In addi-
tion, blood markers can be used as auxiliary indicators for differ-
Lung cancer is the most common incident cancer and the leading entiating between benign and malignant tumors, assisting staging,
cause of cancer-related death in China and worldwide.1 Non-small- and monitoring therapeutic effectiveness and prognosis. Traditional
cell lung cancer is the most frequent histologic subtype of lung blood biomarkers for lung cancer, such as CEA, CA125, SSCA, NSE,
cancer. Despite the use of many effective therapeutic approaches, Cyfra21-1, and proGRP, are used for diagnostic classification and
including surgery, chemotherapy, radiotherapy, targeted therapy, prognostic evaluation, and the significance for early diagnosis or
and immunotherapy, the 5-year survival rate of lung cancer remains screening is not clear. However, the currently available traditional
rather low. 2 Lung tumors at advanced stages with lymph node or markers show low sensitivity and insufficient diagnostic effective-
distant metastasis are the most commonly observed by clinicians. ness.10 It is difficult to make an accurate diagnosis based on a single
Due to the lack of early symptoms, approximately 15% of patients marker. Studies have investigated the combined detection of exist-
are diagnosed with lung cancer at a localized stage, when treatment ing biomarkers to improve sensitivity, but this can result in a corre-
is usually less extensive and more successful.3 The 5-year survival sponding reduction in specificity without any obvious increased in
rate is 53.5% for cases detected when the disease is still localized.4 diagnostic accuracy.11,12
Thus, the key to improve survival rates among lung cancer patients The development of molecular diagnostics and various omics
is early detection. technologies has led to the discovery of potential markers that can
The mortality associated with lung cancer is expected to decrease play an important role in the screening and early diagnosis of lung
in the wake of LDCT screening for early detection. Several large- cancer. The ECLS involves a continuous optimization of combina-
scale randomized trials have been used for evaluating the efficiency tions of 6 AAbs.13,14 The ECLS has been used to aid in the early de-
of LDCT screening worldwide. In 2011, the NLST showed that com- tection of lung cancer in high-risk populations, and a 7-AAb panel
pared with chest X-ray, LDCT screening could reduce lung cancer showed a sensitivity of 47% (9/10) and a specificity of 90% (739/817)
mortality in high-risk groups by 20%.5 The ongoing Dutch-Belgian in a prospective consecutive cohort that was deemed by clinicians to
Randomized Lung Cancer Screening Trial (NELSON) was designed be at an increased risk of developing lung cancer.15 Boeri et al16 col-
to explore whether LDCT screening can reduce lung cancer-associ- lected plasma of lung cancer patients before and after diagnosis by
ated mortality by 25% in high-risk groups over a 10-year follow-up CT and found that miRNAs in the blood samples collected 1-2 years
period.6 However, despite promising results, LDCT screening also before diagnosis predicted lung cancer. The research group further
has many limitations. First, a question remains regarding whether, showed that a miRNA panel, referred to as the MSC, had a marked
in addition to age and smoking status, factors such as sex and sec- 99% negative predictive value. When the MSC was combined with
ondhand smoking should be considered to optimize the classifica- LDCT, the false-positive detection rate of LDCT could be reduced
tion of high-risk groups. Furthermore, the high false-­positive rates 5-fold from 19.4% to 3.7%. In addition, the MSC was associated
of LDCT screening are an important issue. Approximately 96.4% with the prognosis of lung cancer.17,18 Cancer is a systematic disease
of the positive screening results for the LDCT group in the NLST that leads to the generation of numerous molecules, resulting from
were false positive. The high rate of false-positive results can lead genetic alterations in cancer cells, the tumor microenvironment, or
to overdiagnosis and overtreatment, which can aggravate medical organic reactions, that can be found in blood. Thus, it is urgently
7
and psychological burden in patients. Additionally, the optimization necessary to explore new blood markers with increased sensitivity
of screening programs and nodular management strategies are also and specificity for the detection of early-stage lung cancer. These
aspects worth discussing. Some researchers have proposed the con- markers could facilitate the early diagnosis and screening of NSCLC.
cept of volume doubling time for the management of lung nodules.8 Isocitrate dehydrogenase 1 is a catalytic enzyme in the TCA
Finally, some other important factors, such as radiation dosage and cycle that is mainly found in the cytoplasm and peroxisomes. It cat-
cost-effectiveness, restrict the use of LDCT screening.9 alyzes the oxidative decarboxylation of isocitrate to αKG, which
Tumor markers in blood are important for the development of is an important enzyme that produces NADPH in the cytoplasm.
techniques involving in vitro diagnosis and are easy to obtain in a Isocitrate dehydrogenase 1 has attracted widespread attention
noninvasive, inexpensive, and rapid manner. The individualized di- due to the Arg132 mutation in the active center of the enzyme.
agnosis of lung cancer is a hot research topic. An accurate diagnosis Arg132-mutant IDH1 shows novel enzymatic activity and catalyzes
of early-stage lung cancer is difficult to achieve using existing non- the generation of the carcinogenic metabolite 2-hydroxyglutarate,
invasive diagnostic methods, including imaging, sputum cytology resulting in a decrease in the concentration of its original product
SUN et al. | 1741

αKG. 2-Hydroxyglutarate suppresses the activity of prolyl hydroxy-

lase and reduces the degradation of hypoxia-inducible factor.19 This
cancer-associated mutation is found in approximately 70% of grade
II and III gliomas, 8.5% of acute myeloid leukemia, and a few other
types of tumors, but it has not been observed in lung cancer. 20,21
Using a proteomics discovery assay, we previously discovered that
WT IDH1 is overexpressed in NSCLC tissues and that IDH1 can pro-
mote the proliferation of NSCLC cell lines and the growth of xeno-
graft tumors in vivo. 22 In a retrospective study, by comparing 976
NSCLC patients and 479 healthy individuals, we revealed the diag-
nostic value of plasma IDH1 levels and showed that the diagnostic
power of IDH1 for NSCLC is superior to that of traditional blood
markers, such as CEA, CA125, and Cyfra21-1. 23 Thus, IDH1 can be
used as a plasma biomarker for the diagnosis of NSCLC. However,
the previous study was undertaken in a single center in the absence
of confounding diseases and factors included as controls and was
thus not sufficient to verify the diagnostic value of the markers.
Therefore, we investigated whether IDH1 shows clinical sig-
F I G U R E 1 Study profile to determine the utility of isocitrate
nificance for the early detection of NSCLC. In this study, we thus dehydrogenase 1 for the early detection of non-small-cell lung
aimed to verify the expression status and diagnostic significance of cancer. †Postoperative serum samples were obtained from 105
IDH1 in NSCLC, especially in early stages and to investigate IDH1 enrolled non-small-cell lung cancer (NSCLC) patients. BPC, benign
levels in patients with benign pulmonary lesions or OC, and samples pulmonary condition; HC, healthy control; OC, other cancer.

with interfering factors or posttreatment lung cancer from multiple

centers. before any treatment. Healthy controls were included after near-
term general cancer screening without any sign of tumor and with
a CT imaging diagnosis corresponding to “No apparent abnormality
2 | M ATE R I A L S A N D M E TH O DS was found in imaging examination”. Samples with interfering factors
were enrolled based on the criteria in HCs, and represented some of
2.1 | study population the conditions, such as hyperlipidemia, hemolysis, and jaundice, that
could interfere with IDH1 detection. None of the participants had
In total, 1380 participants were consecutively recruited between a history of any malignancy or benign tumor and had not received
August 2016 and April 2017 from 3 hospitals in Beijing, namely, the any antitumor therapy before diagnosis or surgery. Female subjects
National Cancer Center/Cancer Hospital of the Chinese Academy of were required not to be pregnant. We also obtained postoperative
Medical Sciences and Xuanwu Hospital and Beijing Chest Hospital, serum of 105 enrolled NSCLC patients within 1-3 days after surgery
Capital Medical University. The clinical trial scheme was compiled (Figure 1).
according to governmental requirements and submitted to the We excluded 157 subjects because of missing information, his-
National Medical Products Administration. tory of tumors, or lack of diagnosis. The remaining subjects were
We collected serum samples 1 week before the surgical resec- divided into a training cohort and a validation cohort. The training
tion of NSCLC tumors, which were diagnosed based on X-ray, CT, cohort comprised patients with NSCLC, BPCs, or OCs and HCs re-
and biopsy and confirmed by pathology according to the WHO’s cruited from National Cancer Center/Cancer Hospital of the Chinese
Classification of Tumors of the Lung. 24 Tumor stage was defined Academy of Medical Sciences. The validation cohort was recruited
according to the 7th IASLC/AJCC TNM staging system; for the pur- from the other 2 centers with the same grouping criteria. Detection
poses of this study, we classified stage 0-IA tumors as early-stage of IDH1 expression and data collection were carried out by indepen-
NSCLC. The BPC group included 21 patients with benign lung tumors dent researchers at the 3 centers. The detection kit operators were
(pulmonary hamartoma, sclerosing hemangioma, hemangioendothe- blinded to diagnostic information. Patient information was kept con-
lioma, pulmonary fibroma, and atypical adenomatous hyperplasia) fidential. The study was approved by medical ethics committees at
and 74 patients with benign pulmonary disease (chronic obstructive the 3 centers.
pulmonary disease, pneumonia, tuberculosis, pneumatocele, and
chronic bronchitis). Other cancers comprised common human malig-
nancies, such as liver, breast, esophageal, gastric, colorectal, cervi- 2.2 | Sample collection and storage
cal, renal, and other lung cancer including SCLC. Patients with BPCs
and OCs were identified based on pathological evidence obtained Peripheral blood samples were extracted into anticoagulant-free
by surgery, endoscopy, or sputum culture, and serum was collected tubes during routine biochemical examination for patients and
1742 | SUN et al.

centrifuged at RT for 10 minutes according to standard protocols. with 95% CIs. The correlation between serum IDH1 concentrations
We collected the serum samples immediately after inspecting them and clinicopathologic characteristics was analyzed with Pearson’s χ2
at the clinical laboratory. Serum samples were centrifuged, and the test and Spearman’s rank correlation test. We compared IDH1 levels
supernatants were divided into 500-μL aliquots and immediately in serum before and after surgical resection in NSCLC patients with
stored at −80°C to avoid repeated freeze-thaw cycles until testing. the Wilcoxon matched-pairs test.
Serum samples were independently collected by research teams at
the 3 centers. Sample collection and transportation were in accord-
ance with standard operating procedures. 3 | R E S U LT S

3.1 | Sample characteristics

2.3 | Enzyme-linked immunosorbent assay
Sample characteristics of the training (620 cases) and validation (546
Diagnostic kits for the quantitative determination of IDH1 lev- cases) cohorts are shown in Table 1. The distribution of sex between
els (ELISA) are commercially available (Beijing Modern Gaoda the NSCLC and control groups (HC + BPC + OC) was not signifi-
Biotechnology). Researchers performed IDH1 detection indepen- cantly different (Pearson’s χ2 test, P = .488 and .290 in the training
dently at the 3 centers following the manufacturer’s recommenda- and validation cohorts, respectively). Additionally, there was no clear
tions and were blinded to subjects’ clinical information. We used the difference in smoking status between NSCLC patients and disease
same batch of kits, which were kept at 4°C. controls (BPC + OC, Pearson’s χ2 test, P = .633 and .024 in the train-
Before the experiment, samples were thawed at RT. Meanwhile, ing and validation cohorts, respectively).
96-well microtiter plates and associated components were equili- The median age of the 660 NSCLC patients (median age, 60 years;
brated for 30 minutes at RT. Approximately 50 μL serum samples IQR, 14 years) was higher than that in the 506 controls (median age,
and standards (0, 2, 5, 10, 20, 50, and 100 ng/mL) were added to 46 years; IQR, 24 years) (Mann-Whitney U test, P < .001). Non-
wells coated with mAb against IDH1 and incubated at 37°C for small-cell lung carcinoma was dominated by adenocarcinoma (ADC)
2 hours. Then 50 μL solution A (biotinylated anti-IDH1 Ab) and solu- (training cohort, 88.1%; validation cohort, 70.1%) and early-stage can-
tion B (HRP conjugated with avidin) were successively added to each cer in the training cohort (stage 0-I, 61.3%), with obvious distributional
well and incubated at 25°C for 1 hour each time. After every incu- differences between the two cohorts (Pearson's χ2 test, histology
bation, the wells were thoroughly washed at least 5 times with PBS P < .001, stage P < .001, Table 1).
and patted dry. Then, equal volumes of substrate solution A (H2O2) We analyzed the correlation between IDH1 levels and age
and substrate solution B (3,3,5,5-tetramethylbenzidine, TMB) were (Spearman's rank correlation test). Spearman's correlation coeffi-
mixed. Subsequently, 50 μL of the mixed solution was added to each cients were 0.023 (P = .554) in NSCLC patients and 0.028 (P = .532)
well for color development, and the plated were maintained in the in controls. Although the age distribution was different between
dark for 15 minutes at 25°C. Sulfuric acid was then added to stop NSCLC patients and controls, IDH1 levels were not correlated with
the reaction. Optical density was measured at 450 nm/630 nm on a age. Spearman's correlation coefficients between IDH1 levels and
multimode plate reader (Multiskan FC; Thermo Fisher Scientific). We smoking status were 0.020 (P = .613) in NSCLC patients and 0.015
calculated the concentrations of IDH1 with a quadratic polynomial (P = .768) in controls, and IDH1 levels were not correlated with
curve that was fitted using the standard values. Concentrations of smoking status.
quality control samples (QC1 and QC2) were within the linear range
of the curve, with 2-5 ng/mL for QC1 and 10-20 ng/mL for QC2. All
measurements in the clinical trial were undertaken twice in dupli- 3.2 | Serum IDH1 levels in NSCLC patients and
cate, with variable coefficients less than 15%. If the optical density controls in the training cohort
value was higher than 2.0, the sample was diluted and retested. The
concentrations of IDH1 were set at zero when the calculated value The levels of IDH1 are presented as median ± IQR, and nonparametric
was below the detection threshold. tests were used to compare differences in IDH1 levels among groups.
The IDH1 levels in the NSCLC group (6.13 ± 4.80 ng/mL) were sig-
nificantly higher than those in the control groups (HC + BPC + OC,
2.4 | Statistical analysis 1.90 ± 2.81 ng/mL), with P values <.001 (Mann-Whitney U test)
(Figure 2A). The NSCLC group was then separately compared
Data analyses and curve plotting were undertaken with SPSS (ver- with the 3 control groups (Kruskal-Wallis test, P < .001, adjusted
sion 17.0) and MedCalc (version 9.6.2.0). Differences between in- α′ = 0.0083 for post hoc multiple comparisons). The levels of IDH1
dependent groups were tested with the Mann-Whitney U test in the NSCLC group were clearly higher than those in the HC group
(continuous variables and nonparametric analyses). Scatter plots (1.67 ± 2.15 ng/mL, Mann-Whitney U test, P < .001) and the OC
were constructed using GraphPad Prism version 5 for Windows. To group (2.29 ± 3.71 ng/mL, Mann-Whitney U test, P < .001). However,
assess sensitivity, specificity and AUC, ROC curves were constructed there were slightly higher IDH1 levels in the NSCLC group than BPC
SUN et al. | 1743

TA B L E 1 Characteristics of the study cohort

Training cohort (620 cases) Validation cohort (546 cases)

NSCLC Disease controls (patients NSCLC Disease controls (patients

patients HCs with BPCs or OCs) patients HCs with BPCs or OCs)

No. 362 155 103 298 121 127

Age, y
Median 59.0 37.0 53.0 61.0 34.0 59.0
(Q1, Q3) (53, 65) (31, 46) (45, 64) (54, 68) (27, 45) (50, 67)
Range 35-82 23-67 26-84 26-85 18-72 17-86
Sex
Female/male 201/161 87/68 49/54 120/178 64/57 47/80
Smoking status (pack × years)
0 254 141 77 148 – 77
0-20 41 13 9 35 – 18
>20 67 1 17 115 – 32
Histology, n (%)
ADC 319 (88.1) – – 209 (70.1) – –
SCC 42 (11.6) – – 71 (23.8) – –
Others 1 (0.3) – – 18 (6.0) – –
TNM stage, n (%)
0-IA 222 (61.3) – – 66 (22.1) – –
IB 52 (14.4) – – 39 (13.1) – –
II 34 (9.4) – – 22 (7.4) – –
III 45 (12.4) – – 57 (19.1) – –
IV 9 (2.5) – – 62 (20.8) – –
Unclear 0 (0.0) – – 52 (17.4) – –
IDH1 level 6.13 ± 4.80 1.67 ± 2.15 2.69 ± 4.00 5.55 ± 7.11 2.90 ± 1.89 2.18 ± 2.59
(median ± IQR)

–, not applicable; ADC, adenocarcinoma; BPC, benign pulmonary condition; HC, healthy control; IDH1, isocitrate dehydrogenase 1; IQR, interquartile
range; NSCLC, non-small-cell lung cancer; OC, other cancer; Q, quartile; SCC, squamous cell carcinoma.

group (5.49 ± 3.41 ng/mL) (Mann-Whitney U test, P = .037). The level 3.3 | Validation of serum IDH1 levels for NSCLC
of IDH1 was higher in the BPC group than in the HC group (Mann- patients and controls in 2 other centers
Whitney U test, P < .001).
Next, we analyzed the diagnostic performance of IDH1 by Serum IDH1 levels were independently tested in a validation co-
comparing AUCs plotted based on serum IDH1 concentrations hort from 2 other hospitals. However, serum IDH1 concentration
in NSCLC patients and controls. The AUCs (95% CI), sensitivities, in NSCLC patients was not significantly different between the
specificities, predictive values, and probabilities for different validation cohort and the training cohort (5.55 ± 7.11 ng/mL and
comparisons are summarized in Tables 2, S1, and S2. Figure 3A 6.13 ± 4.80 ng/mL, respectively, Mann-Whitney U test, P = .054). The
shows that the AUC for discriminating NSCLC patients from HCs median serum IDH1 levels were 2.90 ± 1.89 ng/mL, 2.17 ± 2.90 ng/
was 0.915 (95% CI, 0.887-0.942) in the training cohort, indicating mL, and 2.18 ± 2.11 ng/mL in the HC, BPC, and OC groups in the
reasonable discriminatory potential for IDH1 in these 2 groups. validation cohort, respectively. The Kruskal-Wallis test revealed that
Additionally, IDH1 showed a strong ability to distinguish be- the concentrations of IDH1 in the NSCLC group were noticeably
tween the NSCLC and control groups (HC + BPC + OC), with an higher than those of the control group (P values <.001 for all) but not
AUC of 0.870 (training cohort, 95% CI, 0.840-0.899) (Table 2 and significantly different among the HC, BPC, and OC groups.
Figure 3D). We chose 5 ng/mL as the cut-off value for IDH1, which The area under the ROC curve for discriminating the NSCLC
showed a high specificity of 92.9% for HCs. This value yielded a group from the control groups (HC + BPC + OC) was 0.745 (95%
sensitivity of 63.3% for distinguishing NSCLC patients in the train- CI, 0.704-0.786) in the validation cohort, and that for discriminat-
ing cohort. The specificity was markedly high at 84.4% for the OC ing NSCLC patients from HCs was 0.730 (95% CI, 0.684-0.776)
group and 86.8% overall (Table 2). (Figure 3B,E). With 5 ng/mL as the cut-off value for IDH1 level,
1744 | SUN et al.

F I G U R E 2 Serum isocitrate dehydrogenase 1 (IDH1) levels in non-small-cell lung cancer (NSCLC) and control groups. Serum IDH1
levels in the (A) training cohort, (B) validation cohort, and (C) whole cohort. Wide horizontal lines represent median values; bars represent
interquartile ranges. *P < .05; **P < .01; ***P < .001. BPC, benign pulmonary condition; HC, healthy control; ns, nonsignificant; OC, other
cancer

sensitivity for discriminating NSCLC patients was 55.0%. The spec- and maintained at the high level with disease progression. The AUC
ificity was markedly high at 89.3% for the HC group, 82.3% for the for discriminating stage 0-IA lung cancer patients and HCs was 0.907
BPC group, 85.4% for the OC group, and 86.3 for the overall group in the training cohort (95% CI, 0.875-0.938), 0.788 in the validation
in the validation cohort (Table 2). cohort (95% CI, 0.711-0.865), and 0.869 in the whole cohort (95%
CI, 0.839-0.899, Figure 4B). Thus, IDH1 is a promising marker for the
identification of patients with stage 0-I lung cancer. As mentioned
3.4 | Combined analysis of the diagnostic above, 5 ng/mL was used as the cut-off value for IDH1 level, and the
performance of IDH1 for early-stage NSCLC sensitivity of NSCLC detection was 58.7% (Table S2).
Isocitrate dehydrogenase 1 had similar diagnostic ability for lung
Combined data from 3 centers were analyzed and are shown in ADC and SCC with equivalent areas under the ROC curves. The
Figure 3 and Table S2. The Kruskal-Wallis test of IDH1 levels in AUCs of the ADC and SCC groups vs control groups (HCs, BPCs, and
the NSCLC, BPC, OC, and HC groups generated P values of less OCs) were 0.822 (95% CI, 0.797-0.848) and 0.813 (95% CI, 0.766-
than .001. The concentration of IDH1 in the NSCLC group was 0.861), respectively, for the whole cohort. Additionally, relative to
6.00 ± 5.78 ng/mL, which was obviously higher than that of control the HC group, the AUCs of ADC and SCC groups were 0.846 (95%
groups (HC, 2.25 ± 2.17 ng/mL; BPC, 2.56 ± 3.61 ng/mL; and OC, CI, 0.819-0.873) and 0.835 (95% CI, 0.786-0.884), respectively,
3.23 ± 3.23 ng/mL). However, there was no difference among the within the whole cohort (Table S2). Despite slightly higher IDH1 lev-
3 control groups. Figure 3C,F shows that the AUC for discriminat- els in the BPC group than in the HC group in the whole cohort, there
ing the NSCLC group from the control groups (HC + BPC + OC) was was no diagnostic significance, and the AUC was 0.549 (95% CI,
0.818 (95% CI, 0.792-0.842), and that for discriminating NSCLC pa- 0.476-0.623). The presence of interfering factors had a low impact
tients from HCs was 0.841 (95% CI, 0.816-0.866). Considering IDH1 on IDH1 detection, and the specificity was 84.2% in the group with
concentration lower than 5 ng/mL as negative, the specificity of dis- interfering factors, similar to that in the control groups (Table S3).
tinguishing NSCLC patients was 91.3% for the HC group, 75.8% for The diagnostic performance of IDH1 was compared with that
the BPC group, 84.4% for the OC group, and 86.6% for the overall of CEA, CA125, and Cyfra21-1, which are lung cancer markers that
control group. The sensitivity was 59.5% for NSCLC patients in the are used in the clinic and had been used for performance compari-
whole cohort (Table S2). son in our previous study. 23 We evaluated the levels of the markers
Next, patients with different stages of lung cancer were sep- for some of the participants. Isocitrate dehydrogenase 1 was supe-
arately compared with controls. Serum levels of IDH1 were not rior to the 3 markers, namely, CEA, CA125, and Cyfra21-1, for lung
different based on disease stages (stage 0-IA, 5.88 ± 5.38 ng/mL; ADC patients. For lung SCC patients, the AUC for IDH1 was smaller
stage IB, 6.69 ± 5.00 ng/mL; stage II, 6.47 ± 7.98 ng/mL; stage III, than that for Cyfra21-1 but greater than that for CEA and CA125
5.84 ± 5.87 ng/mL; and stage IV, 5.18 ± 6.01 ng/mL) (Figure 4A). (Figure S1 and Table S4). Cyfra21-1 was the most effective marker
Isocitrate dehydrogenase 1 was elevated in early-stage lung cancer for lung SCC.
SUN et al. | 1745

TA B L E 2 Performance of serum isocitrate dehydrogenase 1 values for the diagnosis of non-small-cell lung cancer (NSCLC)

Negative
AUC (95% CI) Se% Sp% PPV% NPV% Positive LR LR

Training cohort
NSCLC vs HC, BPC and 0.870 (0.840-0.899) 63.3 86.8 87.1 62.7 4.80 0.42
OC
NSCLC vs HC 0.915 (0.887-0.942) 63.3 92.9 95.4 52.0 8.91 0.40
Stage 0-IA NSCLC vs HC, 0.859 (0.826-0.892) 58.6 86.8 79.3 70.9 4.44 0.48
BPC and OC
Stage 0-IA NSCLC vs HC 0.907 (0.875-0.938) 58.6 92.9 92.2 61.0 8.25 0.45
Stage IB-IV NSCLC vs 0.886 (0.852-0.919) 70.7 86.8 74.4 84.5 5.37 0.34
HC, BPC and OC
Stage IB-IV NSCLC vs HC 0.927 (0.896-0.958) 70.7 92.9 90.0 77.8 9.96 0.32
Validation cohort
NSCLC vs HC, BPC and 0.745 (0.704-0.786) 55.0 86.3 82.8 61.5 4.01 0.52
OC
NSCLC vs HC 0.730 (0.684-0.776) 55.0 89.3 92.7 44.6 5.12 0.50
Stage 0-IA NSCLC vs HC, 0.797 (0.730-0.865) 59.1 86.3 53.4 88.8 4.31 0.47
BPC and OC
Stage 0-IA NSCLC vs HC 0.788 (0.711-0.865) 59.1 89.3 75.0 80.0 5.50 0.46
Stage IB-IV NSCLC vs 0.746 (0.697-0.796) 54.4 86.3 74.2 72.3 3.97 0.53
HC, BPC and OC
Stage IB-IV NSCLC vs HC 0.732 (0.676-0.788) 54.4 89.3 88.3 56.8 5.07 0.51

AUC, area under the receiver operating characteristic curve; BPC, benign pulmonary condition; HC, healthy control; LR, likelihood ratio; NPV,
negative predictive value; OC, other cancer; PPV, positive predictive value; Se, Sensitivity; Sp, Specificity.

3.5 | Isocitrate dehydrogenase 1 showed specificity associated with surgical resection in NSCLC patients. Therefore, cir-
for NSCLC diagnosis culating IDH1 was mainly released by the tumor tissue.

We examined serum IDH1 levels in patients in common types of

cancer, including esophageal, stomach, liver, colorectal, breast, 4 | D I S CU S S I O N
kidney, and cervical cancer, and SCLC. Levels of IDH1 in patients
with various cancers were not different from those of HCs (Kruskal- In our previous study, IDH1 expression was higher in lung cancer
Wallis test, P = .088, Figure 5A). The AUCs relative to HCs for pa- tissues than in adjacent normal tissues at both transcriptional and
tients with these common cancers were 0.582, 0.523, 0.530, 0.572, translational levels, and plasma IDH1 was identified as a biomarker in
0.594, 0.733, 0.589, and 0.527, respectively (Figure 5B and Table S5). a large cohort from a single center. In the present study, we recruited
Therefore, IDH1 shows no diagnostic value in patients with OCs and 1223 participants from 3 centers. This study is thus a further valida-
is a relatively specific marker for the diagnosis of NSCLC. tion of the clinical utility of IDH1 as an early diagnostic biomarker
for lung cancer.
This is the first study comparing IDH1 levels using HC, BPC,
3.6 | Isocitrate dehydrogenase 1 reduced in and OC groups as controls in a multicenter-based cohort. The
postoperative serum of lung cancer patients IDH1 levels are significantly increased in NSCLC patients, and ROC
curve analysis showed the robust ability of IDH1 to distinguish
During the process of cancer development, tumor biomarkers are NSCLC patients from HCs, with rather high specificity and positive
generated from cancer cells, the tumor microenvironment, or the predictive values (Figure 2 and Table 2). The expression level of
host. Reductions in IDH1 levels after treatment were evaluated to BPCs (5.49 ± 3.41 ng/mL) was higher than HCs (1.67 ± 2.15 ng/
confirm the correlation between circulating markers and lung can- mL) in the training cohort, but the number of cases (16 cases) was
cer load. Indeed, IDH1 levels were significantly reduced in the post- too few to draw a conclusion. Slightly increased IDH1 levels were
operative serum of lung cancer patients (Wilcoxon matched-pairs detected in patients with BPCs; nevertheless, these levels were
test, P < .0001, Figure 6). The rate of positive detection of IDH1 obviously lower than those in NSCLC patients in the whole cohort
was 70.5% before operation (74/105), and it decreased to 5.7% (Figure 2C). Patients with BPCs had benign tumors and some in-
(6/105) within 1-3 days after operation. Thus, IDH1 was significantly flammatory diseases, which can be precancerous lesions or can
1746 | SUN et al.

F I G U R E 3 Receiver operating characteristic (ROC) curve analysis of the diagnostic performance of serum isocitrate dehydrogenase
1 (IDH1) levels for differentiating non-small-cell lung cancer (NSCLC) patients and controls in the (A, D) training cohort, (B, E) validation
cohort, and (C, F) whole cohort. Black circles mark the cut-off of values at 5 ng/mL. A-C, ROC curves of NSCLC group vs healthy control
(HC) + benign pulmonary condition + other cancer groups. D-F, NSCLC group vs HC group

F I G U R E 4 Diagnostic performance
of isocitrate dehydrogenase 1 (IDH1) for
early-stage non-small-cell lung cancer
(NSCLC). A, IDH1 levels in patients
with different stages of lung cancer. B,
Receiver operating characteristic curve
analysis of IDH1 levels in early-stage
NSCLC patients vs healthy controls (HCs).
Black circle indicates the cut-off value at
5 ng/mL. ***P < .001. ns, nonsignificant

initiate inflammatory responses. However, we could not identify IDH1 detected also includes various subclasses. Additionally, IDH1
a diagnostic value of IDH1 in patients with other common cancers levels dramatically decreased after cancer resection and were sig-
(Figure 5). Therefore, IDH1 is a specific marker for NSCLC. Unlike nificantly related to lung cancer load (Figure 6). Isocitrate dehy-
the classical protein markers CEA and Cyfra21-1, IDH1 level was drogenase 1 was superior to CEA, CA125, and Cyfra21-1 for lung
already high in the early stages; this interesting phenomenon also ADC patients, yet weaker than Cyfra21-1 for lung SCC (Figure S1
reminds us to pay attention to the sources of IDH1 in addition to and Table S4). These results were consistent with those of previous
tumor release in the early stage or dig deeper whether all the serum studies. 23
SUN et al. | 1747

F I G U R E 5 Diagnostic value of isocitrate dehydrogenase 1 (IDH1) for common human tumors. A, Serum IDH1 levels in patients with
common human cancers. B, Receiver operating characteristic curve analysis for patients with esophageal, stomach, liver, colorectal, breast,
kidney, cervical, or small-cell lung cancer (SCLC) vs healthy controls (HCs)

NSCLC patients from healthy individuals with 58.7% sensitivity and

91.3% specificity.
According to our results, IDH1 has many potential clinical
uses. First, IDH1 is a potentially practical marker for lung can-
cer screening. The combined detection of tumor markers and CT
screening is an important diagnostic strategy. Isocitrate dehydro-
genase 1 could help to accurately identify high-risk individuals and
thus reduce the need for repeated screening, or IDH1 levels could
even be used for secondary screening of LDCT-positive individu-
als. Many lung nodules identified in CT scans remain undiagnosed
in clinical practice. The evaluation of IDH1 levels might assist in
F I G U R E 6 Serum isocitrate dehydrogenase 1 (IDH1) level is
lung cancer diagnosis by supplementing other tumor biomarkers
significantly reduced in lung cancer patients within 1-3 days after
lung cancer surgery and routine clinical examinations. Furthermore, IDH1 could be
used for recurrence monitoring of NSCLC patients because of
its significant reduction in postoperative serum. However, a fol-
In addition to these confirmatory results, we found that IDH1 low-up study for post-treatment IDH1 levels monitoring is still
level is already increased in the early-stage of lung cancer. Early- needed.
stage NSCLC is hard to diagnose due to its rare clinical manifestation Nevertheless, our study has several limitations. We did not col-
and lack of regular screening. Low-dose CT screening can improve lect follow-up information to determine outcomes, and no monitor-
early detection rate, but its use is limited because of cost and need ing tests were undertaken to analyze dynamic changes in IDH1 levels.
for irradiation. In contrast, testing for IDH1 using a standardized in Moreover, the number of patients with OCs was relatively small in
vitro blood-based diagnostic kit is simple, inexpensive, and easy with this study. Because of its important role in the TCA cycle, IDH1 level
minimal stress. The IDH1 levels can be used to distinguish stage 0-IA might increase in cancers of many organs, especially the liver, which
1748 | SUN et al.

is the main metabolic organ. Further verification of IDH1 levels in a 3. Siegel R, Naishadham D, Jemal A. Cancer statistics, 2013. CA Cancer
J Clin. 2013;63:11-30.
larger number of samples from patients with OCs is thus required.
4. DeSantis CE, Lin CC, Mariotto AB, et al. Cancer treatment and sur-
Metabolic reprogramming has emerged as a new hallmark of cancer. vivorship statistics, 2014. CA Cancer J Clin. 2014;64:252-271.
A recent study reported that WT IDH1 is involved in the regulation 5. Aberle DR, Adams AM, Berg CD, et al. Reduced lung-cancer mortal-
of metabolism of branched-chain amino acids promoting glioma cell ity with low-dose computed tomographic screening. N Engl J Med.
2011;365:395-409.
proliferation. 25 Thus, the number of studies on the role of WT IDH1
6. Yousaf-Khan U, van der Aalst C, de Jong PA, et al. Final screening
in tumors is gradually increasing; IDH1 in glioma and acute myeloid round of the NELSON lung cancer screening trial: the effect of a
leukemia should be tested in the next stage of clinical experiment. 2.5-year screening interval. Thorax. 2017;72:48-56.
However, regarding NSCLC, the underlying mechanism maintaining 7. Patz EF Jr, Pinsky P, Gatsonis C, et al. Overdiagnosis in low-dose
high expression of WT IDH1 remains unclear as does the mechanism computed tomography screening for lung cancer. JAMA Intern Med.
2014;174:269-274.
by which increased WT IDH1 levels promote tumor occurrence and
8. Horeweg N, van der Aalst CM, Vliegenthart R, et al. Volumetric
development. Thus, mechanistic investigations on WT IDH1 are still computed tomography screening for lung cancer: three rounds of
needed. In addition, we hope to increase the sample size and add the NELSON trial. Eur Respir J. 2013;42:1659-1667.
follow-up information to further validate our observations. In view 9. Goulart BH, Bensink ME, Mummy DG, Ramsey SD. Lung cancer
screening with low-dose computed tomography: costs, national
of the satisfactory preliminary results in early-stage lung cancer, we
expenditures, and cost-effectiveness. J Natl Compr Canc Netw.
plan to determine IDH1 levels in a screening cohort to verify diag- 2012;10:267-275.
nostic potential of IDH1 combined with LDCT screening. 10. Nakamura H, Nishimura T. History, molecular features, and clinical
This is the first multicenter large-scale validation trial in- importance of conventional serum biomarkers in lung cancer. Surg
Today. 2017;47:1037-1059.
vestigating IDH1 levels in NSCLC. As a diagnostic marker for
11. Molina R, Marrades RM, Auge JM, et al. Assessment of a combined
NSCLC, IDH1 showed excellent performance with high specificity. panel of six serum tumor markers for lung cancer. Am J Respir Crit
Isocitrate dehydrogenase 1 was a potential biomarker for identify- Care Med. 2016;193:427-437.
ing benign and malignant tumors and monitoring recurrence. This 12. Wang L, Wang D, Zheng G, et al. Clinical evaluation and therapeutic
monitoring value of serum tumor markers in lung cancer. Int J Biol
study is thus important for the field of lung cancer screening and
Markers. 2016;31:e80-e87.
diagnosis. In summary, this study verified and revealed the value 13. Boyle P, Chapman CJ, Holdenrieder S, et al. Clinical validation of an
of IDH1 in clinical practice. Isocitrate dehydrogenase 1 possesses autoantibody test for lung cancer. Ann Oncol. 2011;22:383-389.
potential clinical utility as a serum protein biomarker for the early 14. Murray A, Chapman CJ, Healey G, et al. Technical validation of an
autoantibody test for lung cancer. Ann Oncol. 2010;21:1687-1693.
diagnosis of NSCLC.
15. Chapman CJ, Healey GF, Murray A, et al. EarlyCDT(R)-Lung test:
improved clinical utility through additional autoantibody assays.
AC K N OW L E D G M E N T S Tumour Biol. 2012;33:1319-1326.
This work was supported by the National Key R&D Program of China 16. Boeri M, Verri C, Conte D, et al. MicroRNA signatures in tis-
(2018YFC1315000, 2018YFC1315003, and 2016YFC0905400), sues and plasma predict development and prognosis of com-
puted tomography detected lung cancer. Proc Natl Acad Sci USA.
National Natural Science Foundation of China (81871885), PUMC
2011;108:3713-3718.
Youth Fund (2017320013), Non-profit Central Research Institute 17. Sestini S, Boeri M, Marchiano A, et al. Lung cancer screening in
Fund of Chinese Academy of Medical Sciences (2018RC320010), high-risk subjects: early detection with LDCT and risk stratification
National Key Basic Research Development Plan (2016YFC0905400), using miRNA-based blood test. Epidemiol Prev. 2016;40:42-50.
18. Sozzi G, Boeri M, Rossi M, et al. Clinical utility of a plasma-based
CAMS Innovation Fund for Medical Sciences (2017-I2M-1-005,
miRNA signature classifier within computed tomography lung
2016-I2M-1-001), and Special Foundation for Central Committee cancer screening: a correlative MILD trial study. J Clin Oncol.
Health Care (W2017BJ39). The authors thank the patients who par- 2014;32:768-773.
ticipated in the study and the staff of clinical laboratories of all par- 19. Dang L, White DW, Gross S, et al. Cancer-associated IDH1 muta-
tions produce 2-hydroxyglutarate. Nature. 2009;462:739-744.
ticipating sites for their valuable efforts.
20. Yan H, Parsons DW, Jin G, et al. IDH1 and IDH2 mutations in glio-
mas. N Engl J Med. 2009;360:765-773.
D I S C LO S U R E 21. Mardis ER, Ding L, Dooling DJ, et al. Recurring mutations found
The authors declare that no competing interest exists. by sequencing an acute myeloid leukemia genome. N Engl J Med.
2009;361:1058-1066.
22. Tan F, Jiang Y, Sun N, et al. Identification of isocitrate dehydroge-
ORCID nase 1 as a potential diagnostic and prognostic biomarker for non-
Shuangshuang Mao https://orcid.org/0000-0001-6799-5813 small cell lung cancer by proteomic analysis. Mol Cell Proteomics.
Jie He https://orcid.org/0000-0002-0285-5403 2012;11:M111 008821.
23. Sun N, Chen Z, Tan F, et al. Isocitrate dehydrogenase 1 is a novel
plasma biomarker for the diagnosis of non-small cell lung cancer.
REFERENCES Clin Cancer Res. 2013;19:5136-5145.
1. Chen W, Zheng R, Baade PD, et al. Cancer statistics in China, 2015. 24. Travis WD, Brambilla E, Nicholson AG, et al. The 2015 world health
CA Cancer J Clin. 2016;66:115-132. organization classification of lung tumors: impact of genetic, clini-
2. Ettinger DS. Ten years of progress in non-small cell lung cancer. J cal and radiologic advances since the 2004 classification. J Thorac
Natl Compr Canc Netw. 2012;10:292-295. Oncol. 2015;10:1243-1260.
SUN et al. | 1749

25. Tonjes M, Barbus S, Park YJ, et al. BCAT1 promotes cell prolifera-
tion through amino acid catabolism in gliomas carrying wild-type How to cite this article: Sun N, Sun S, Gao Y, et al. Utility of
IDH1. Nat Med. 2013;19:901-908. isocitrate dehydrogenase 1 as a serum protein biomarker for
the early detection of non-small-cell lung cancer: A

S U P P O R T I N G I N FO R M AT I O N multicenter in vitro diagnostic clinical trial. Cancer Sci.

Additional supporting information may be found online in the 2020;111:1739–1749. https://doi.org/10.1111/cas.14387

Supporting Information section.