Biotech, in brief

Side effects deciding treatment

Serious lung inflammation hit 0.9% of patients on Roche's new lung cancer drug, against 3.9% and 4.4% on two rival drugs. [9] In Italy and Spain, colon cancer patients given lighter chemotherapy after a blood test had far less nerve damage. [11] All 13 patients in a Texas gut study had stopped their cancer drug because it inflamed their bowel. [12]

Cancer trials run in China

Akeso's lung cancer drug beat Merck's Keytruda on survival in a trial run in China, and analysts question whether that carries over elsewhere. [8] Roche's partner MediLink ran its own final-stage lung cancer trial in China too. [9]

New drugs added to old ones

Isembyld is approved only for patients already taking an older SMA drug. [1] AstraZeneca's breast cancer pill, added to a common drug, did not hold cancer back significantly longer than the standard pairing. [7] Cellectis dropped two blood cancer therapies because newer medicines had shrunk the number of patients. [10]

Computers entering the clinic

The US health department is backing projects in which AI programs diagnose patients and prescribe treatment. [15] Some of its own officials worry there is too little evidence that this is safe. [15] Servier, a French drugmaker, licensed an AI system from Owkin to guide its cancer research. [14]

2 pressures seen across the day run through Biotech too.

• Worries that AI is going too fastBuilding ↑ · seen in 4 beats→
• Treatments falling short in careful testsBuilding ↑ · seen in 2 beats→

1. 1The US regulator approved the first drug for spinal muscular atrophy that works on the muscle itself
2. 2A lung cancer drug outlives Keytruda
3. 3AstraZeneca's new breast pill fails a wider test
4. 4US health department pushes AI that prescribes
5. 5A blood test chose who got harder chemotherapy

The FDA, the US drug regulator, approved Isembyld on Friday 11 September to treat spinal muscular atrophy, or SMA. [1]

The next tests are what insurers agree to pay against a $310,000 list price, and how the broken-bone warning shapes which patients doctors give it to. [4][1]

Akeso, a Chinese drugmaker, and its US partner Summit Therapeutics said people with lung cancer lived a median 30.8 months on their drug ivonescimab, against 22.6 months on Merck's Keytruda. [8]

Keytruda is the immunotherapy that much of lung cancer care is built around. [8] Ivonescimab blocks two targets, PD-1 and VEGF, where Keytruda blocks one. [8]

The trial, HARMONi-2, had already shown the drug held cancer back for longer, and the survival figures came out at a lung cancer conference in South Korea. [8]

Analysts have doubted whether trials run in China carry over to patients elsewhere, and the trials that will decide approval outside China are still running. [8]

One analyst noted that Merck's Sac-TMT cut the risk of death by 45% in a similar group, against 27% for ivonescimab. [8]

AstraZeneca said on Friday that its pill Etcamah, also called camizestrant, failed a major trial called SERENA-4. [7]

Given with a common drug to people newly diagnosed with advanced breast cancer, it did not hold the cancer back significantly longer than the standard pairing. [7]

The US regulator had approved it earlier this month for patients whose tumours develop ESR1 mutations, which help cancer resist treatment. [7] AstraZeneca has not released the numbers. [7]

The result could limit how widely the pill is used. [6] AstraZeneca executives had talked of it one day earning $5 billion a year. [7]

Also seen in Mind & body.

The New York Times reported on Monday that the US health department is speeding up projects in which AI programs diagnose patients, offer therapy and prescribe medicine. [15]

Some officials in the department worry the change is moving too fast, with too little evidence that it is safe and works. [15]

Vinod Khosla, a billionaire investor whose son runs a health AI company, has been influential with top officials, people close to the matter said. [15]

Those officials include Mehmet Oz, who runs Medicare and Medicaid, the US government's health insurance for older and poorer people. [15]

Prescribing is the step where software stops helping a doctor with paperwork and starts choosing a patient's treatment. [15]

Khosla told a start-up event in July that AI would replace human doctors. [15]

Also seen in Cyber, Finance and Tech.

After colon cancer surgery, 135 patients at 11 hospitals in Italy and Spain had their blood tested for DNA shed by leftover tumour. [11]

Patients with no tumour DNA got a single chemotherapy drug instead of the usual combination. [11]

The 26% who had it got a stronger combination, stepped up again if the DNA stayed. [11]

Three years on, 83% of those without tumour DNA were free of cancer, against 58% of those with it. [11] The results were published on Monday in Nature Cancer. [11]

The trial missed its own goal: 88% of the test-negative patients stayed free of relapse at two years, below the 92% set in advance. [11]

Against a matched group from an older trial, results were similar with much less nerve damage, and the authors say a proper randomised trial is still needed. [11]

When several treatments help about as much, the one that hurts less gets used, because a patient can only benefit from a drug they can keep taking.

Its survival figure sits below GSK's, so Roche points to serious lung inflammation in 0.9% of its patients against 3.9% and 4.4% on the two rival drugs.

Patients with no tumour DNA got a single drug instead of a combination, and the step that changed was nerve damage, which fell sharply against an older group.

Every one of the 13 patients had already stopped a cancer drug because of the harm, and settling the harm let some of them start it again.

Each of the three is deciding from a different piece of paper.

MediLink, which licensed the drug to Roche for sale outside China, reported a final-stage trial of tam-peli in small cell lung cancer. [9]

Tam-peli is an antibody that carries a cancer-killing drug to cells showing a protein called B7-H3. [9]

In the trial, run in China, patients lived a median 13.3 months, against 9.4 months on the chemotherapy drug topotecan. [9]

GSK's rival drug reported 18.5 months in its own trial, and Merck and Daiichi Sankyo's reported 12 months in a smaller, earlier-stage study. [9]

Comparing separate trials run in different places can be unreliable, so the three companies are also competing on side effects. [9] Roche says it will start a worldwide final-stage trial quickly. [9]

Checkpoint inhibitors are cancer drugs that free the immune system to attack tumours. They commonly cause diarrhoea and colitis, an inflamed bowel. [12]

At MD Anderson Cancer Center in Texas, 13 patients got a transplant of donor stool as the first treatment for it. [12]

Ten improved, and among them the typical wait was a day and a half. [12]

All 13 had stopped their cancer drug because of the gut damage, and some were able to restart it. [12]

The study had no comparison group, and the authors say some patients might have recovered with ordinary care or by stopping the drug. [12]

They want trials against steroids, the usual treatment, which dampen the immune system. [12]

Cellectis, a gene-editing company built on technology from France's Institut Pasteur, said it will seek partners for its two cell therapies for leukaemia and lymphoma and move into heart disease. [10]

It had worked to make these therapies from donors' cells, as a simpler option than CAR-T treatments custom-made from each patient's own cells. [10]

New medicines have changed how blood cancers are treated, leaving fewer patients for each therapy and slowing its trials. [10]

It will now work on two heart-disease gene-editing treatments not yet tested in people. [10]

Both heart treatments aim at targets, APOC3 and PCSK9, where approved medicines already exist. [10] That leaves Cellectis facing an uphill battle in a crowded field. [10]

Researchers built a vaccine for H5N1 bird flu on VSV, a virus that does not cause disease in people, by swapping in the bird flu virus's surface genes. [13]

They changed the carrier's genes to make it safer and then gave mice two doses into the nose. [13]

Every vaccinated mouse survived a dose of H5N1 that would otherwise have killed it. [13] A vaccine built on the same kind of carrier is already sold against Ebola. [13]

The study was published on 9 September. [13]

H5N1 bird flu is treated as a pandemic threat, and the nose spray raised defences in the lining of the airways as well as in the blood. [13]

This is a mouse result, and most things that protect mice never reach people. [13]

Owkin, a company that builds AI for drug research, licensed its system K Pro to Servier, a French drugmaker, for cancer discovery and trial design. [14]

The deal also opens Owkin's MOSAIC collection to Servier's scientists. [14] MOSAIC maps tumour samples by their molecules and by where each cell sits in the tissue. [14]

The two companies first worked together in 2023, and Servier signed an $888 million cancer research deal with Insilico Medicine in January. [14]

Servier says the value is having patient data and the analysis in one place. [14] Owkin has earlier signed deals with AstraZeneca, Merck and Sanofi. [14]