Trial design and comparator choices shape the wave of phase 3 failures

AstraZeneca’s HICKORY trial in hypophosphatasia, whose data landed in March 2026, failed after 25 weeks because the enrolled patients were not the patients who would show an effect on the Six-Minute Walk Test. The analysis called the outcome 'a design failure, not an efficacy failure'.

That distinction now frames several other readouts. Novo Nordisk said ziltivekimab missed the primary endpoint in the ZEUS phase 3 trial in people with atherosclerotic cardiovascular disease, chronic kidney disease and inflammation. The trial recorded a hazard ratio of 0.99 (95% confidence interval, 0.88 to 1.11) for major adverse cardiovascular events against placebo. Drug Discovery and Development reported that the protocol required the drug to demonstrate a relative risk reduction of no less than 20%, with 95% power. The company said ziltivekimab produced the expected biological effect but no MACE benefit in this population.

AstraZeneca also stopped the EVOLVE-Lung02 phase 3 trial of volrustomig after an Independent Data Monitoring Committee concluded the bispecific was unlikely to meet either dual primary endpoint, progression-free survival or overall survival, in people with PD-L1-negative non-small cell lung cancer. A total of 895 people had been randomized to volrustomig or Merck’s Keytruda in combination with chemotherapy. AstraZeneca’s statement did not include efficacy data.

In September, Novartis reported two more late-stage misses. On Friday after U.S. markets closed, the company and Ionis Pharmaceuticals said pelacarsen failed the phase 3 Lp(a)HORIZON trial’s primary endpoint of time to first cardiovascular death, nonfatal myocardial infarction, nonfatal stroke or urgent coronary revascularization requiring hospitalization, compared with placebo. The trial enrolled over 8,000 participants already on optimized background care. Novartis did not disclose the magnitude of cardiovascular risk reduction. Citi analysts said the Lp(a) hypothesis is 'weakened, but not disproven' because details beyond the topline miss remain scarce.

GEN reported that del-desiran, an antibody oligonucleotide conjugate Novartis acquired with Avidity, failed the phase 3 HARBOR trial in myotonic dystrophy type 1. The trial missed its primary endpoint of statistically significant improvement versus placebo in video Hand Opening Time through week 54. Novartis stated that exploratory analyses and secondary endpoints suggested some clinical activity, though details were not disclosed. The same article noted that pelacarsen had failed on the previous Friday and that Novartis paused rap-cel CAR T studies on Sept. 2 after three patient deaths.

What the protocol choices show

These events share a common structure. The ZEUS protocol set a high bar in a placebo-controlled population. In EVOLVE-Lung02, the comparator was Keytruda plus chemotherapy, an established standard, in a PD-L1-negative population where single-agent checkpoint inhibitors have limited responses. Both choices shaped what the trials had to show.

The Lp(a) failure shows how standard-of-care changes can blunt a trial’s ability to detect an added benefit. Jefferies analysts wrote that improving standards of care are reducing cardiovascular events, which raises the bar for experimental drugs to show added benefit. BMO Capital Markets said differences in trial design mean Lp(a) as a target has not been fully invalidated. Truist Securities noted that Amgen’s olpasiran produced deeper Lp(a) reductions in earlier work and enrolled a more high-risk population.

None of these results proves a class is dead. Ziltivekimab trials in heart failure and after acute heart attack are planned to continue, with data expected in the first half of 2027. AstraZeneca is continuing phase 3 studies of volrustomig for cervical cancer, for head and neck squamous cell carcinoma, and for mesothelioma; the lung cancer study showed no new safety signals. Amgen’s olpasiran and Lilly’s lepodisiran remain in late-stage outcome trials, with topline readouts expected in 2028 and 2029 respectively.

The HICKORY analysis argues that sponsors should run a pre-activation enrichment review before site qualification visits begin. The publication says feasibility questionnaires typically ask whether sites can enroll patients, not whether the enrolled distribution will produce a readable signal at the primary endpoint. It cites a Tufts Center analysis putting the median direct cost of a substantial phase 3 protocol amendment at $535,000. According to the publication, a protocol that enrolls a population outside the endpoint’s sensitivity window is an oversight failure, regardless of whether each site executed its initiation visit correctly.