Immunoscore for predicting risk of colon cancer relapse

Study size, design and location

Retrospective clinical validity study assessing the prognostic value of Immunoscore in 2,681 people with stage 1 to 3 colon cancer, in 14 centres across 13 countries in Europe (including Belgium, Sweden, France, Germany, the Netherlands, Czech Republic, Italy, Switzerland), North America and Asia.

Intervention and comparator

Immunoscore (high compared with low), no comparator.

Key outcomes

The Immunoscore test showed a high level of reproducibility between observers and centres (r=0.97 for colon tumour; r=0.97 for invasive margin; p<0.0001). In the training set (n=700), people with a high Immunoscore had the lowest risk of recurrence at 5 years. Recurrence rates at 5 years were 8%, 19% and 32% for people with high, intermediate and low Immunoscore respectively (hazard ratio [HR] for high compared with low Immunoscore was 0.20; p<0.0001). Findings were confirmed in the internal (n=636) and external (n=1,345) validation sets. The association between Immunoscore and TTR was independent of a person's age, sex, T stage, N stage, microsatellite-instability status and existing prognostic factors (p<0.0001). Of the 1,434 people with stage 2 cancer, a high Immunoscore was associated with the lowest risk of recurrence and the highest DFS and OS. The difference in risk of recurrence at 5 years was significant (HR for high versus low Immunoscore was 0.33; p<0.0001), even when adjusting for potential confounders in a Cox multivariable analysis (p<0.0001). Immunoscore had the highest relative contribution to survival and risk of recurrence compared with all the other clinical parameters, including the TNM classification system.

There are also published conference abstracts presenting data from various subgroup analyses of this study, including:

Strengths and limitations

This was a large international multicentre study that included a training data set (n=700) and internal validation data set (n=636) using samples from centres across Switzerland, Germany, France, US, Czech Republic and Canada. The demographic and clinical characteristics of the patients were well balanced between the training set and internal validation set. Study pathologists and immunologists used standardised operating procedures and staining quality and intensity was validated to ensure consistency among samples. The statistical analyses were done by an independent external group.

People who had preoperative treatment were excluded and adjuvant chemotherapy use was not considered in the study. The study did not include any centres from the UK which may limit generalisability to the NHS. One of the authors is a co-founder of the company.

Study size, design and location

Retrospective clinical validity study of data from a phase 3 randomised trial of 1,322 people with stage 3 colon cancer who were having oxaliplatin-based treatment in France. This study was an ancillary biomarker analysis of the IDEA France Study.

Intervention and comparator

Immunoscore (low compared with intermediate or high scores), no comparator.

Key outcomes

In total, 1,322 people (66%) from the overall IDEA France study's modified intention-to-treat population with available samples were included in the analysis. Overall, 1,062 (85.6%) samples reached the quality control; 43.6%, 47.0% and 9.4% of people in this analysis had low, intermediate and high Immunoscores respectively. A low Immunoscore was associated with a higher risk of relapse or death compared with an intermediate or high Immunoscore (low compared with intermediate or high score; HR=1.54; p<0.0001). The 3‑year DFS was 66.8% for people with a low Immunoscore and 77.1% for people with an intermediate or high Immunoscore. Immunoscore remained significantly independently associated with DFS (p=0.003) when adjusted for sex, histological grade, T or N stage and microsatellite instability. For people having treatment with FOLFOX (92% of people in the study), Immunoscore showed statistically significant predictive value for treatment duration (3 compared with 6 months) in terms of DFS (p=0.057). People with intermediate or high Immunoscore significantly benefitted from 6 months of FOLFOX therapy compared with 3 months (HR=0.53; p=0.0004), including people with clinically low- and high-risk stage 3 colon cancer (all p<0.001). People with low Immunoscore did not significantly benefit from 6‑month FOLFOX therapy compared with 3‑month treatment (HR=0.84; p=0.27).

Strengths and limitations

The study used data from a large multicentre, randomised trial in 129 centres. The Immunoscore test was done blinded to clinical data. Baseline clinical and histopathological characteristics and the Immunoscore categorisations were well balanced among people having 3 and 6 months of FOLFOX therapy.

Most of the people (90%) in the trial had treatment with FOLFOX so no conclusions can be made about people having CAPOX. The median follow up was 4.3 years, meaning the longer-term treatment benefit of using Immunoscore could not be determined. The study was done in France which may limit generalisability to the NHS. One of the authors is the co‑founder of the company and 2 authors are employees of the company.

Study size, design and location

Retrospective clinical validity study of data from an international multicentre phase 3 randomised trial in 559 people with stage 3 colon cancer who had treatment with adjuvant FOLFOX. Data from the FOLFOX arm of the phase 3 randomised North Central Cancer Treatment Group (NCCTG) N0147 trial were used to construct Cox models for predicting DFS. The NCCTG N0147 trial included centres from the US, Canada and Puerto Rico.

Intervention and comparator

Immunoscore (low compared with high score), no comparator.

Key outcomes

A low Immunoscore was significantly associated with a shorter DFS (adjusted HR=1.69; p=0.001) after adjusting for age, tumour location, T and N stage, BRAF and KRAS gene mutations, and mismatch repair status. The 3‑year DFS for people with a low Immunoscore was 66.6% compared with 82.6% for those with a high Immunoscore. In terms of the relative contributions of variables to DFS risk, the number of positive lymph nodes had the largest impact (43.1%), followed by T stage (18%), BRAF and KRAS status (16.1%) and then Immunoscore (14.9%). Among people in the low‑risk group (people with T1 to T3 N1 tumours; n=296), Immunoscore was the strongest predictor of DFS and the only variable to remain statistically significant (HR=0.87; p=0.02).

Strengths and limitations

This study included people with uniform cancer stage and treatment. Data came from a multicentre randomised phase 3 clinical trial, reducing risk of selection bias. The study used a multivariable Cox model to adjust for confounding variables.

The trial did not include UK centres which may limit generalisability to the NHS. Everyone in the trial received adjuvant chemotherapy so the ability of Immunoscore to predict chemotherapy response could not be determined. Two of the authors are co-founders of the company.

Study size, design and location

Multicentre, retrospective cohort study evaluating the prognostic value of Immunoscore in 763 people with stage 3 colon cancer. It included people from 2 cohorts (cohort 1, North America; cohort 2, Europe and Asia) of the SITC‑led study reported by Pagès et al. (2018).

Intervention and comparator

Immunoscore (high compared with low scores), no comparator.

Key outcomes

Recurrence‑free rates at 3 years were 56.9%, 65.9% and 76.4% in people with low, intermediate and high Immunoscores respectively (HR [high compared with low], 0.48; p=0.0003). High Immunoscore was significantly associated with prolonged TTR, OS and DFS (all p<0.001). A statistically significant association between a high Immunoscore and prolonged TTR was also shown in people with MSS status (HR [high compared with low], 0.36; p=0.0003). Immunoscore had the strongest contribution for influencing survival (TTR and OS). Chemotherapy was significantly associated with survival in people with a high Immunoscore with either low-risk cancer (T1 to 3, N1; HR [chemotherapy compared with no chemotherapy] 0.42; p=0.0011) or high-risk cancer (T4 or N2, or both; HR [chemotherapy compared with no chemotherapy] 0.50; p=0.0015).

Strengths and limitations

This was an international study that included data from 14 centres across 13 countries. The ability of Immunoscore to predict response to chemotherapy demonstrated in this study supports IDEA France study findings (see Pagès et al. 2020). Biomarker quality control was blinded to clinical data, and clinical data quality control was blinded to biomarker data. Immunoscore categories were previously defined independently of clinical data. The study used multivariable Cox models stratified by centre and adjusted for potential confounders.

The study population was heterogenous in terms of treatment and follow up but is reflective of real-world practice. People having neoadjuvant treatment were not included. Only 65% and 16% of people in the study had microsatellite instability and mutational status respectively. The study did not include any centres from the UK, which may limit generalisability to the NHS. Authors had financial interests in the company.

Study size, design and location

Retrospective cohort study of 595 people with stage 1 to 3 colon cancer to assess the prognostic value of Immunoscore. Samples from 595 people with stage 1 to 3 colon cancer taken from 3 different studies (see Pagès et al. 2018, ImmuCol and ImmuCol2) were analysed. The interlaboratory reproducibility of Immunoscore was assessed using a second cohort of 100 people with stage 1 to 3 colon cancer treated in Romania. The analytical precision outcomes were determined using 13 anonymous FFPE colon cancer blocks.

Intervention and comparator

Immunoscore, no comparator.

Key outcomes

Manual and automatic counts for CD3+ and CD8+ T cells were strongly correlated (r=0.94, p<0.001, and r=0.92, p<0.001 respectively). The intensity of the histological staining was not affected by the age of the tumour sample block over a period of 30 years. Immunoscore was not affected by the tumour block selected or the position of the tested tissue section within the tumour block. Consistency of Immunoscores between selected or randomised tumour tissue blocks was 93% and between distant tissue sections from the same block was 95%. Interlaboratory reproducibility was assessed between 2 centres and was shown to have 93% agreement in Immunoscores. Reproducibility of the test was also unaffected by other variables such as antibody lots, staining kits, immunohistochemistry automators and operators. The prognostic validity of Immunoscore was assessed in a cohort of 229 people with stage 1 to 3 colon cancer. The relative proportion of variance for TTR explained by Immunoscore was 53%. This was greater than other prognostic factors included in the model including T stage, microsatellite-instability status and total number of lymph nodes.

Strengths and limitations

Data used to assess the prognostic value of the Immunoscore came from 3 previously published studies, 2 of which were prospective in design. The study tested for multiple variables of sample preparation including lot-to-lot variability of primary antibodies, the choice of tumour block, the reproducibility of the sectioning process, and age and storage of the paraffin block.

Retrospective data used to assess prognostic value of the Immunoscore came from French cohorts only, limiting generalisability to the NHS. Only 1 of the 2 test centres did immunostaining of CD3+ and CD8+ T cells using the CE-marked HalioDX Immunoscore Kit and quantification of cells using the Immunoscore Analyser. Two of the study authors are co-founders of the company.

Study size, design and location

Clinician survey on 10 people with stage 2 colon cancer, in the US.

Intervention and comparator

There were 25 medical oncologists who were presented with 10 patient cases and asked for their recommendations (adjuvant chemotherapy and frequency of surveillance) through an online survey. At a live event, clinicians were presented with Immunoscore data and asked to complete the same patient survey.

Key outcomes

On average, clinicians chose to change their chemotherapy or surveillance recommendations, or both, in 56% of cases. All but 1 clinician (96%) changed their recommendations for at least 1 case, whereas 92% (23/25) changed their preference for chemotherapy in at least 1 case. The rate of change for chemotherapy prescription was 36% per patient case (range: 7 to 13 changes). Surveillance strategies were rarely altered when chemotherapy recommendations changed.

Strengths and limitations

The study was sponsored by the company. The surveys were completed using different formats (online survey versus live event), which may have influenced physician decisions. Also, the study was available as abstract only with limited methodological information such as the time between completing each survey, as well as the level of expertise of the medical oncologists enrolled.