The US Food and Drug Administration approved atezolizumab with chemotherapy on 8 October 2026 for adjuvant treatment of stage III colon cancer that is mismatch repair deficient, known as dMMR. The approval is American. As of 13 October 2026 no authorisation for this use was found from the UK or EU regulators.
One randomised phase 3 study, the ATOMIC trial, carries the decision. Published in the New England Journal of Medicine on 26 March 2026, it was funded by the US National Cancer Institute and by Genentech, the drug's maker. Among 712 patients, 86.3% of those given atezolizumab with chemotherapy were alive and free of disease at 3 years, against 76.2% of those given chemotherapy alone. That is an absolute difference of 10.1 percentage points on the pre-registered primary endpoint.
Disease-free survival is a surrogate measure. Longer life has not been shown, and serious adverse events were more common with the combination. What follows reads the FDA notice, the journal abstract and the registry entry side by side. The review ran under Project Orbis, the FDA scheme described in our earlier report on an Orbis submission.
What the FDA approved on 8 October 2026
The FDA approved atezolizumab, sold as Tecentriq by Genentech, in combination with a fluoropyrimidine and oxaliplatin for the adjuvant treatment of adults and children aged 2 years and over with stage III dMMR colon cancer. Adjuvant means treatment given after surgery to lower the chance of the cancer returning.
A second product was approved in the same notice. Tecentriq Hybreza, a version of atezolizumab given under the skin, received the same indication for adults and for children aged 12 and over who weigh at least 40 kilograms.
That age range deserves a second look. By the agency's own summary, the study enrolled 711 adults and 1 paediatric patient. How the evidence was extended to children from 2 years of age is not explained in the notice, which says the full prescribing information is still to be posted on the Drugs@FDA database. That reasoning will sit in the review documents.
Warnings and precautions are listed for immune-mediated adverse reactions, infusion-related reactions, complications of allogeneic stem-cell transplantation and embryo-fetal toxicity. Both applications were granted priority review.
Atezolizumab is a prescription-only medicine. Nothing here is advice about treatment, which is a matter for a patient's own oncology team.
The ATOMIC trial in absolute numbers
Start with who was in it.
ATOMIC randomly assigned 712 patients whose stage III colon tumours had been surgically removed and tested as dMMR. Of these, 355 were assigned to atezolizumab plus a standard chemotherapy regimen called mFOLFOX6 and 357 to mFOLFOX6 alone, according to the paper by Sinicrope and colleagues. Median age was 64, 55.1% were women, and 53.9% had tumours classed as high risk.
Consider the comparator. Patients in the control group received the established standard of care, not a placebo, so the result measures what atezolizumab adds to chemotherapy. Everyone knew who was receiving what, because the design was open-label. Treatment also lasted different lengths of time in the 2 groups: 6 months of chemotherapy in one, and 12 months of therapy in the other, the second half being atezolizumab alone.
After a median follow-up of 40.9 months, 3-year disease-free survival was 86.3% with the combination, with a 95% confidence interval of 81.8 to 89.8, and 76.2% with chemotherapy alone, interval 70.9 to 80.6. In absolute terms the difference is 10.1 percentage points. Put another way, about 10 patients would need to receive the combination for 1 more to be alive and disease-free at 3 years. That last figure is our arithmetic from the published rates, not a number the authors report.
Most headlines carry the relative figure: a hazard ratio of 0.50, interval 0.35 to 0.73, meaning the rate of recurrence or death was halved. Both statements describe one result. Oddly, the 3 documents print 3 different p-values for it, below 0.001 in the journal, 0.0002 on the registry and 0.0001 in the FDA notice. All are far below the threshold the protocol set.
Completion is harder to summarise. The registry records that 163 of 355 patients in the atezolizumab group completed their assigned treatment, against 214 of 357 given chemotherapy alone. Because the planned courses differed in length, those 2 rates are not a like-for-like comparison.
What the registry adds to the headline
Headlines say that adding immunotherapy halves the risk of recurrence. The registry entry, NCT02912559, lets a reader test what that does and does not cover.
First, the endpoint was not switched. Registered in 2016, the protocol named disease-free survival at 3 years as the primary outcome and overall survival at 5 years as a secondary outcome, and the journal reports that same primary endpoint. That is how the system is meant to work, and it is worth saying when it does.
Second, survival. Results first posted to the registry on 14 April 2026 give overall survival as 89.7% in the atezolizumab group and 87.9% with chemotherapy alone, a difference of 1.8 percentage points, with a hazard ratio of 0.90 and a confidence interval from 0.55 to 1.47. An interval that wide spans both a meaningful benefit and a meaningful harm.
Deaths from any cause numbered 31 of 355 against 33 of 357. Atezolizumab raised 3-year disease-free survival from 76.2% to 86.3% in ATOMIC, with no survival difference shown so far.
Third, who left before treatment began. According to the registry, 25 patients assigned to chemotherapy alone withdrew before starting, compared with 4 in the atezolizumab group. In an open-label study that pattern is a recognised hazard, since a patient who learns they will not receive the new drug may leave. All 712 randomised patients were counted in the primary analysis, which limits the damage, but the imbalance belongs in any honest account.
Disease-free survival is a surrogate endpoint. It counts recurrences and deaths, on the reasoning that fewer recurrences should mean fewer deaths later. Whether that holds for immunotherapy after surgery in dMMR colon cancer is what the overall survival follow-up exists to answer, and the registry still lists the study as active.
Harms compared with chemotherapy alone
More toxicity came with the benefit, and the numbers are public.
Adverse events of grade 3 or higher, the severe end of the standard scale, occurred in 291 of 346 treated patients in the atezolizumab group and 240 of 334 given chemotherapy alone, according to the registry. That is 84.1% against 71.9%, matching the percentages in the journal, and a difference of 12.2 percentage points. Events are counted whether or not doctors attributed them to treatment.
Serious adverse events, a separate regulatory category, affected 197 of 346 patients, or 56.9%, against 138 of 334, or 41.3%. Stopping treatment because of an adverse event was recorded for 62 patients in the atezolizumab group and 32 in the chemotherapy group.
So the trade can be stated plainly. For every 100 patients given the combination in place of chemotherapy alone, the data suggest about 10 more were disease-free at 3 years, about 12 more had a severe adverse event and about 16 more had a serious one. Those are group averages from 1 study and say nothing about an individual patient.
Against that comparison, 2 cautions apply. Patients in the atezolizumab group were on treatment for twice as long, which gives more time for events to be recorded. And the chemotherapy backbone, which contains oxaliplatin, accounts for much of the toxicity in both groups, as the 71.9% figure in the control group shows.
In the UK, suspected adverse reactions to any medicine can be reported through the MHRA's Yellow Card scheme.
Is atezolizumab approved for colon cancer in the UK or EU
No authorisation of atezolizumab for stage III dMMR colon cancer was found in the UK or the EU as of 13 October 2026. The FDA decision applies only in the United States. The European Medicines Agency's page for Tecentriq, read on that date, makes no mention of colon cancer.
Project Orbis, under which the FDA reviewed the application, lets partner regulators assess a submission at the same time. For this review it names 4 partners: Australia's Therapeutic Goods Administration, Health Canada, Israel's Ministry of Health and Swissmedic. Reviews at those agencies are continuing, the notice says. The MHRA, which takes part in some Orbis reviews, is not on the list.
That absence is a fact about 1 review and not a forecast. A UK decision, if one is sought, would come through a separate application. Whether a NICE appraisal is planned was not established for this article.
For UK and European readers the position is therefore unchanged. According to the paper, fluoropyrimidine and oxaliplatin chemotherapy is the standard adjuvant treatment for stage III colon cancer, and that was the control arm. Use of atezolizumab in this setting outside the United States would be off-label.
What is still unknown
Several questions are open, and each has a document that will answer it.
Overall survival at 5 years, the registered secondary endpoint.
The FDA's reasoning for the paediatric indication.
The decisions of the 4 Orbis partner agencies.
Whether any application is made to the MHRA or the EMA.
Whether the benefit differs between lower-risk and higher-risk tumours. The subgroup analyses were not read for this article, and a subgroup finding is not a result.
Watch 2 documents. Once posted, the FDA's full prescribing information and review package will show the reviewers' own reservations. Longer survival follow-up will appear on the registry entry, last updated on 1 October 2026, as it matures. Readers who want the same method applied to another recent decision can turn to our reading of the FDA's imlunestrant approval.
Until then the accurate summary is short. One study, funded by a public institute and the manufacturer, showed a 10.1 point absolute gain in 3-year disease-free survival, no demonstrated survival gain so far, and more severe adverse events. This article was written from the journal abstract, the registry and the FDA notice, not the full paper. It will be reviewed by 13 January 2027, or sooner if a UK or EU decision is published.
