Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2023 Sep 4;14(1):5391.
doi: 10.1038/s41467-023-41011-4.

The Oncology Biomarker Discovery framework reveals cetuximab and bevacizumab response patterns in metastatic colorectal cancer

Alexander J Ohnmacht #  1   2 , Arndt Stahler #  3 , Sebastian Stintzing #  3   4 , Dominik P Modest  3 , Julian W Holch  4   5 , C Benedikt Westphalen  5 , Linus Hölzel  1 , Marisa K Schübel  1   2 , Ana Galhoz  1   2 , Ali Farnoud  1 , Minhaz Ud-Dean  1 , Ursula Vehling-Kaiser  6 , Thomas Decker  7 , Markus Moehler  8 , Matthias Heinig  1 , Volker Heinemann  9 , Michael P Menden  10   11   12
Affiliations

The Oncology Biomarker Discovery framework reveals cetuximab and bevacizumab response patterns in metastatic colorectal cancer

Alexander J Ohnmacht et al. Nat Commun. .

Abstract

Precision medicine has revolutionised cancer treatments; however, actionable biomarkers remain scarce. To address this, we develop the Oncology Biomarker Discovery (OncoBird) framework for analysing the molecular and biomarker landscape of randomised controlled clinical trials. OncoBird identifies biomarkers based on single genes or mutually exclusive genetic alterations in isolation or in the context of tumour subtypes, and finally, assesses predictive components by their treatment interactions. Here, we utilise the open-label, randomised phase III trial (FIRE-3, AIO KRK-0306) in metastatic colorectal carcinoma patients, who received either cetuximab or bevacizumab in combination with 5-fluorouracil, folinic acid and irinotecan (FOLFIRI). We systematically identify five biomarkers with predictive components, e.g., patients with tumours that carry chr20q amplifications or lack mutually exclusive ERK signalling mutations benefited from cetuximab compared to bevacizumab. In summary, OncoBird characterises the molecular landscape and outlines actionable biomarkers, which generalises to any molecularly characterised randomised controlled trial.

PubMed Disclaimer

Conflict of interest statement

A.S. served on advisory boards for BMS and Novocure, received honoraria for talks by Roche, Servier and Taiho Pharmaceuticals and received reimbursement for travel by Roche, Merck KGaA, MSD Sharp & Dohme, Pfizer, Lilly Oncology, and Amgen. V.H., S.S. and D.P.M. received honoraria for talks, advisory boards and travel expenses by Merck KGaA, Amgen, Roche, Pfizer, BMS, MSD, AstraZeneca, Novartis, Terumo, Oncosil, Nordic, Seagen, GSK, Takeda, Servier, Pierre Fabre, Taiho, Lilly Oncology, Servier, Sanofi and Bayer Pharmaceuticals. M.P.M. is a former employee at AstraZeneca, academically collaborates with AstraZeneca, GSK and Roche, and receives funding from GSK and Roche. J.W.H. served on an advisory board for Roche, has received honoraria from Roche, and travel support from Novartis. M.M. received honoraria for advisory boards or talks by Amgen, BMS, Roche, Merck KGaA, MSD Sharp & Dohme, Lilly Oncology, Servier, Pierre Fabre, Taiho Sanofi and Bayer Pharmaceuticals and serves as officer for the European Organisation on Research and Treatment of Cancer (EORTC), and Arbeitsgemeinschaft internistische Onkologie (AIO). C.B.W. has received honoraria from Amgen, Bayer, Chugai, Celgene, GSK, MSD, Merck, Janssen, Ipsen, Roche, Servier, SIRTeX, Taiho; served on advisory boards for Bayer, BMS, Celgene, Servier, Shire/Baxalta, Rafael Pharmaceuticals, RedHill, Roche, has received travel support by Bayer, Celgene, RedHill, Roche, Servier, Taiho and research grants (institutional) by Roche. C.B.W. serves as an officer for the European Society of Medical Oncology (ESMO), Deutsche Krebshilfe (DKH) and AIO. The remaining authors declare no competing interests.

Figures

Fig. 1
Fig. 1. The Oncology Biomarker Discovery (OncoBird) workflow.
a Patients in clinical trials were treated with (T) two treatment regimens with measured clinical endpoints. Subsequently, their tumours are characterised according to (M) tumour genetic alterations (somatic mutations and copy number alterations) and (S) tumour subtypes. b With this input, OncoBird outlines c the molecular landscape and d the biomarker landscape. For the latter, e somatic alterations are explored for a differential patient prognosis for each treatment arm. f Consecutively, for each treatment arm, subtype-specific biomarkers are derived. g Finally, interactions between treatment arms are examined. The grey shadings indicate the data included in the previous analysis step. Here, this is exemplified in the FIRE-3 clinical trial using Kaplan–Meier plots, including 95% confidence intervals (CI) and summary statistics of the Cox regression models. h RAS mutations are established biomarkers of cetuximab resistance. i Patients with RAS wild-type tumours showed a better prognosis when treated with cetuximab within left-sided tumours compared to right-sided tumours. In addition, j the RAS wild-type subpopulation in left-sided tumours showed benefits when treated with cetuximab compared to bevacizumab.
Fig. 2
Fig. 2. Molecular landscape of the FIRE-3 clinical trial.
a Oncoprint of 373 mCRC tumours, including mutations and copy number alterations detected in more than 12 tumours. b The mutually exclusive alteration patterns were derived with the Mutex algorithm. Gene expression profiles of 451 mCRC tumours are annotated by c the consensus molecular subtypes (CMS) and d the primary tumour side. e Venn diagram showing all enriched somatic alterations for CMS1 and right-sided tumours, and f enriched somatic alterations for CMS2 and left-sided tumours. g Frequently altered cancer genes tested for enrichment in left- or right-sided tumours, and h tested against CMS subtypes using one-sided hypergeometric tests. Source data for the figure panels are provided as Source Data file.
Fig. 3
Fig. 3. Identification of genetic biomarkers for FOLFIRI plus cetuximab or bevacizumab.
a Volcano plot for genetic biomarkers of cetuximab in the form of mutually exclusive gene modules or single gene mutations. Each point shows the effect of a particular group of alterations summarised by its hazard ratio derived by the Cox regression models and its raw p-value derived by a Wald test. Exemplifying the most significant associations, Kaplan–Meier plots, including 95% confidence intervals (CI) and summary statistics of the Cox regression models, are shown for b the mutually exclusive module consisting of RAS and BRAF mutations, and c the amplification of TOP1 treated with cetuximab. For investigating the biomarker composition, we focus on d resistance biomarkers of FOLFIRI plus cetuximab with FDRcet < 0.1, showing their hazard ratios and 95% CIs. For these, e the composition of mutually exclusive genes is indicated by dark grey colour, and f an oncoprint highlighting mutational frequencies of biomarker combinations is shown. In like manner, g cetuximab sensitivity biomarkers, shown by their hazard ratios and 95% CIs, and h their composition are summarised. i Karyoplot showing transcription start sites of co-amplified genes on chromosome 20q. j Volcano plot of the genetic biomarkers of bevacizumab with FDRbev < 0.3, shown in brown colour by their hazard ratios derived by the Cox regression models and their raw p-values derived by a Wald test. Kaplan–Meier plot including 95% CIs and summary statistics of the Cox regression models of k mutations in KRAS or BRAF and l APC mutations treated with bevacizumab. The compositions of bevacizumab biomarkers are shown in Supplementary Fig. 10a, b. A Source Data file is provided, which contains the source data for the figure panels and the sample sizes of the conducted statistical tests.
Fig. 4
Fig. 4. Identification of subtype-specific genetic biomarkers for FOLFIRI plus cetuximab or bevacizumab.
Subtype-specific genetic biomarkers for OS of a cetuximab and b bevacizumab using hazard ratios including 95% confidence intervals (CI) derived from single Cox regression models. Subtypes are defined by either the primary tumour side, CMS or unstratified (reference model). Kaplan–Meier plots including 95% CIs, hazard ratios and raw p-values derived by Wald tests from the Cox regression models of subtype-specific genetic biomarkers for c ARFRP1 in CMS2, d KRAS in CMS4, e KRAS in CMS2 and f KRAS in CMS1 in either the cetuximab or bevacizumab treatment arm. A Source Data file is provided, which contains the source data for the figure panels and the sample sizes of the conducted statistical tests.
Fig. 5
Fig. 5. Predictive biomarkers in the context of tumour subtypes.
Overview of interaction biomarkers (FDRint < 0.2) focusing on a mutant and b wild-type tumours when comparing cetuximab and bevacizumab treatment, using hazard ratios including 95% confidence intervals (CI) derived from single Cox regression models fitted on OS. Triangle points and confidence intervals were obtained from the bootstrap-based bias-correction of treatment effects. For the conducted statistical tests, the sample sizes are given in Supplementary Data 3. Here exemplified, Kaplan–Meier plots including 95% CIs, hazard ratios and raw p-values derived by Wald tests from the Cox regression models compare treatments in subgroups for c ARFRP1 amplifications in CMS2 and d, e KRAS mutations in CMS4. Source data for the figure panels are provided as Source Data file.
Fig. 6
Fig. 6. Stability analysis and benchmark with other methods.
a The ten most significant biomarkers across 25 models of five times repeated 5-fold cross-validation. b Boxplots of treatment effects in terms of hazard ratios for the predicted subgroups in the 25 test sets for the benchmarked methods, including standard treatment guidelines (std) and overall across all patients (null). The centre line depicts the median; the box represents the inter-quartile range (IQR) and the whiskers the interval 1.5 times the IQR. c Oncoprint showing identified subgroups for the benchmarked methods, including std, CMS subtypes, tumour sidedness and mutations in KRAS and NRAS. d Forest plot showing hazard ratios including 95% confidence intervals (CI) and amount of patients in the subgroup for which standard treatment is not recommended and which was found by subgroup analysis methods (new-std-negative). A Source Data file is provided, which contains the source data for the figure panels and the sample sizes of the conducted statistical tests.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

References

    1. Ting, N., Cappelleri, J. C., Ho, S. & Chen, D.-G. (eds) Design and Analysis of Subgroups with Biopharmaceutical Applications (Springer, 2020).
    1. European Medicines Agency. Guideline on the Investigation of Subgroups in Confirmatory Clinical Trials. Draft. European Medicines Agency/Committee for Medicinal Products for Human Use. EMA/CHMP/539146/2013 (EMA, 2014).
    1. Lipkovich I, Dmitrienko A, D'Agostino Sr BR. Tutorial in biostatistics: data-driven subgroup identification and analysis in clinical trials. Stat. Med. 2017;36:136–196. doi: 10.1002/sim.7064. - DOI - PubMed
    1. Zhang Z, Seibold H, Vettore MV, Song W-J, François V. Subgroup identification in clinical trials: an overview of available methods and their implementations with R. Ann. Transl. Med. 2018;6:122. doi: 10.21037/atm.2018.03.07. - DOI - PMC - PubMed
    1. Loh W, Cao L, Zhou P. Subgroup identification for precision medicine: a comparative review of 13 methods. Wiley Interdiscip. Rev. Data Min. Knowl. Discov. 2019;9:e1326. doi: 10.1002/widm.1326. - DOI

Publication types

MeSH terms