No one can predict who responds to immunotherapy

For patients with advanced melanoma, immunotherapy offers a realistic chance of long-term survival and probably cure, and the ten-year data show that patients who are alive and progression-free at three years rarely die of melanoma afterwards. Nivolumab plus ipilimumab gives the best long-term results but at a high price in serious side effects; nivolumab alone or nivolumab plus relatlimab are alternatives for patients at lower risk or with autoimmune concerns. The trial is also a caution about surrogate endpoints: the survival plateau took years to become visible.

For the first time a new drug has beaten pembrolizumab, the global first-line standard, in a randomised lung cancer trial, and it did so by combining checkpoint blockade with anti-angiogenesis in one molecule. For patients outside China nothing changes yet: the drug is not approved in the West, the trial was single-country, and survival benefit has not been shown. If confirmed in the global HARMONi-3 and HARMONi-7 trials, PD-1 x VEGF bispecifics could replace PD-1 antibodies as the immunotherapy backbone.

Patients with limited-stage small-cell lung cancer who complete chemoradiotherapy without progression should now be offered up to two years of durvalumab consolidation, which extends life by almost two years on average. This is the first survival improvement for limited-stage disease since twice-daily radiotherapy and prophylactic cranial irradiation, and small-cell lung cancer is no longer a disease where immunotherapy gives only marginal gains.

For the first time a randomised trial suggests that a vaccine tailored to an individual's tumour can reduce relapse when combined with immunotherapy, which is a proof of concept for a field that had failed for decades. Nothing changes for patients yet: the trial was small, the confidence interval crossed one, and the phase 3 trial in melanoma (and parallel trials in lung and other cancers) must confirm it. If it does, personalised mRNA vaccines could become a routine adjunct to checkpoint inhibitors after surgery.

Women with locally advanced cervical cancer that is node-positive or stage III-IVA should now be offered pembrolizumab alongside and after chemoradiotherapy, which improves the chance of cure. The result matters most in countries where cervical cancer is common but immunotherapy access is poorest, so its global impact depends on pricing and health-system capacity. It does not apply to early-stage disease treated with surgery or to lower-risk locally advanced disease without nodal involvement.

Patients fit enough for cisplatin whose bladder cancer has invaded the muscle wall should now be offered durvalumab with their pre-operative chemotherapy and for about a year after surgery, which improves the chance of cure without compromising the operation. The trial cannot say whether the adjuvant phase is necessary, or how to treat cisplatin-ineligible patients, for whom other trials are ongoing.

For colon cancer that is mismatch-repair deficient (about 10-15% of colon cancers, more in older patients), a single short course of immunotherapy before surgery is now a reasonable standard and is far more effective than chemotherapy, which barely works in this subtype. It requires testing every colon cancer for mismatch repair at diagnosis, before surgery. Whether some patients can safely skip surgery, as in dMMR rectal cancer, is the next question.

Pancreatic cancer was thought to be immunologically inert because of its low mutation burden; this study showed that with the right vaccine platform its few neoantigens can still be targeted. It is the strongest human evidence so far that personalised cancer vaccines can generate durable, tumour-specific immunity, and it justifies the randomised trials now running in pancreatic cancer and melanoma. Benefit is not yet proven, because responders may simply have had more immunogenic tumours.

Women with newly diagnosed advanced or recurrent endometrial cancer should receive a PD-1 antibody (dostarlimab or pembrolizumab) with their chemotherapy, and mismatch repair testing is now essential because women with dMMR tumours gain a very large and durable benefit. The gain in mismatch-repair-proficient tumours is real but smaller, and molecular classification (POLE, p53, MMR) is increasingly used to decide who benefits most.

Patients with operable stage II-III lung cancer without EGFR or ALK alterations should have chemo-immunotherapy discussed before surgery rather than only afterwards. Three pre-operative cycles do not compromise the operation and improve cure rates. Whether to continue immunotherapy after surgery, as the perioperative trials do, and whether patients with pCR need any further treatment, remain open questions.

Patients with rectal cancer whose tumour is mismatch-repair deficient (about 5-10% of rectal cancers) can now be offered immunotherapy alone with the realistic expectation of avoiding surgery, radiotherapy and a permanent stoma. This requires mismatch repair testing on the diagnostic biopsy, close endoscopic and MRI surveillance, and treatment in an experienced centre. It does not apply to the 90% of rectal cancers that are mismatch-repair proficient.

For stage II-III triple-negative breast cancer, chemotherapy plus pembrolizumab before surgery and pembrolizumab alone afterwards is now the standard approach worldwide, and the survival gain is real, not just a surrogate. It does not apply to stage I disease or to hormone-receptor-positive or HER2-positive cancers. The price is a year of immunotherapy with a meaningful chance of a permanent endocrine side effect such as hypothyroidism or adrenal insufficiency.

Patients with newly diagnosed advanced melanoma have a dual-checkpoint option that improves on nivolumab alone with only a modest increase in serious side effects, making it attractive for those unable to tolerate or unwilling to risk the toxicity of ipilimumab. It did not prove superior survival, and it has not been compared with nivolumab plus ipilimumab, which remains preferred for patients with brain metastases or other high-risk features. LAG-3 is now an established target under study in many other cancers.

Lifileucel proved that Steven Rosenberg's decades-old TIL concept could be industrialised into a licensed product and gave patients with checkpoint-refractory melanoma, who otherwise have few options, a chance of durable remission. It is the first cell therapy approved for any solid tumour. The treatment requires surgery to harvest tumour, hospitalisation for lymphodepletion and IL-2, and specialised centres, so its reach is limited.

Patients with newly diagnosed advanced stomach or oesophageal adenocarcinoma whose tumour is HER2-negative and PD-L1 positive (CPS 5 or more, or at least 1 in some regions) should receive chemotherapy with nivolumab (or pembrolizumab, from KEYNOTE-859), which adds about three months of median survival and doubles the chance of being alive at three years. The benefit in PD-L1-negative tumours is doubtful, and these patients may be better served by chemotherapy alone or by trials.

Patients with advanced liver cancer and good liver function should be offered atezolizumab plus bevacizumab (or durvalumab plus tremelimumab) rather than sorafenib as first treatment; median survival is now around 19 months and about a quarter of patients respond. Endoscopy to treat varices before starting bevacizumab is essential because of bleeding risk. Patients with poorer liver function (Child-Pugh B) or autoimmune disease or transplants were not studied.

Every colorectal cancer should be tested for mismatch repair deficiency, because patients whose metastatic tumour is dMMR should receive pembrolizumab rather than chemotherapy as first treatment, gaining a much better chance of durable remission with fewer side effects. About a third of dMMR tumours do not respond initially, so early scans are essential and chemotherapy remains available. The trial does not apply to the 95% of colorectal cancers that are mismatch-repair proficient.

Patients with head and neck squamous cell cancer that has recurred or spread should be treated first with pembrolizumab: alone if their tumour is strongly PD-L1 positive and they can wait for a slower response, or with chemotherapy if the tumour is bulky or PD-L1 low. Cetuximab-based chemotherapy is no longer the default. Long-term follow-up shows a small but real group of patients alive at four to five years, which was almost unheard of before.

Most patients with newly diagnosed advanced non-squamous lung cancer that lacks a targetable mutation should receive chemotherapy plus pembrolizumab; those with PD-L1 of 50% or more may reasonably receive pembrolizumab alone. About one in five patients is alive at five years, compared with roughly one in ten with chemotherapy alone. Patients with EGFR or ALK alterations were excluded and should have targeted therapy first.

Patients with stage III lung cancer that cannot be removed surgically should receive a year of durvalumab after completing chemoradiotherapy, provided they have not progressed. This roughly doubles the chance of being alive without progression at five years. Whether the benefit extends to PD-L1-negative tumours is contested, and the EGFR-mutated subgroup is better served by osimertinib (LAURA).

This small trial explained why colorectal cancer had seemed immune-resistant (most is MMR-proficient) and established the principle that a genomic feature, not the tissue of origin, can predict immunotherapy response. It led directly to the 2017 tissue-agnostic approval of pembrolizumab and to routine MMR/MSI testing of many cancers. Every patient with advanced dMMR cancer should now be considered for checkpoint blockade.

This study is why PD-1 inhibitors were developed across cancers rather than in melanoma alone: unexpected activity in lung cancer, historically thought immune-resistant, changed drug development priorities industry-wide. It also introduced PD-L1 immunohistochemistry as a candidate biomarker and pneumonitis as a signature toxicity. Within five years PD-1 blockade was approved in more than ten cancers.

This paper launched the immunotherapy era. It was the first randomised evidence that taking a brake off the immune system could extend life in a solid cancer, and it introduced clinicians to immune-related adverse events and to responses that arrive late or after apparent progression. Ipilimumab alone has since been superseded by PD-1 antibodies and combinations, but every checkpoint programme traces back to this result.

Tumours hide from T cells by displaying PD-L1; blocking that interaction lets the immune system attack. This is the mechanism of pembrolizumab, nivolumab, atezolizumab and their relatives, which now treat more than 20 cancer types.

Every checkpoint inhibitor, from ipilimumab to pembrolizumab, rests on this idea: the immune system can already recognise cancer and just needs its brakes released. It changed the goal of immunotherapy from vaccinating against tumours to unleashing existing T cells.