Lung cancer: the failed and stopped programmes, and why

More is not better with radiotherapy. RTOG 0617 raised the dose in stage III disease from 60 to 74 Gy and shortened survival (median 28.7 against 20.3 months in favour of the lower dose), a result attributed to cardiac dose and treatment interruptions, and 60 Gy remains the standard. Lung ART tested postoperative mediastinal radiotherapy in 501 patients with proven N2 involvement after complete resection and did not improve three-year disease-free survival: mediastinal relapse fell, cardiopulmonary toxicity rose, and the two cancelled out.

Threshold and design failures account for much of the immunotherapy list. CheckMate 026 gave first-line nivolumab to patients selected at a PD-L1 threshold of 5 percent and found median progression-free survival 4.2 against 5.9 months (hazard ratio 1.15) and overall survival 14.4 against 13.2 months (hazard ratio 1.02), while KEYNOTE-024, run at the same time with a 50 percent threshold, succeeded; the 50 percent cut-off in use today is the direct consequence. MYSTIC randomised 1,118 patients at 203 centres in 17 countries to durvalumab, durvalumab with tremelimumab, or platinum doublet chemotherapy and missed its primary overall survival endpoints (JAMA Oncology 2020); NEPTUNE, which tested the same doublet against chemotherapy in patients selected by blood tumour mutational burden, also failed. That is why the surviving durvalumab and tremelimumab regimen, in POSEIDON, includes chemotherapy and a limited course of the CTLA-4 antibody rather than dual blockade alone. PACIFIC-2 added durvalumab concurrently with chemoradiotherapy rather than after it and did not improve progression-free survival, so consolidation, not concurrent treatment, remains the sequence.

Adjuvant circulating tumour DNA selection has not worked either: MERMAID-1 and MERMAID-2, which used minimal residual disease to pick patients for adjuvant durvalumab, were both stopped, and no phase 3 trial has yet shown that treating a positive blood test before a scan changes anything in lung cancer.

Among drugs, ATLANTIS, the confirmatory trial of lurbinectedin with doxorubicin in relapsed small-cell lung cancer, missed its overall survival endpoint, leaving an accelerated approval unconfirmed for years. Mobocertinib, approved in 2021, was withdrawn across 2023 and 2024 after its confirmatory trial in EGFR exon 20 insertion disease failed. Necitumumab, approved in 2015 for squamous disease, added about a month of median survival and is almost never used. In access rather than efficacy, NICE has refused five regimens that are standard elsewhere: nivolumab with ipilimumab and chemotherapy (TA724), pralsetinib (TA812), amivantamab after platinum chemotherapy (TA850), ramucirumab with docetaxel (TA403) and tarlatamab (TA1091).

The pattern is consistent. In lung cancer, a biomarker threshold chosen one notch too low, a drug given at the same time as radiotherapy rather than after it, or a dose raised on the assumption that more kills more, each turns a positive programme into a negative one.