346 - Scaling biotech and improving global health: lessons from an extraordinary career in medicine

KEYPOINTS

• 01

  Her career was shaped by the early AIDS epidemic.

  Desmond-Hellmann, who went to Catholic school for 12 years in Reno, trained in San Francisco in 1982, a year after the first MMWR report, when the disease was a frightening mystery—she and colleagues even had blood drawn to check whether they'd been infected treating patients. Some patients sold their life insurance expecting to die soon, then regretted it once antiretroviral therapy arrived.

• 02

  Frontline work in Uganda showed the scale of the crisis.

  With her husband, she worked on HIV in Uganda while still UCSF faculty. A 16-year-old girl attending an STD clinic had about a 50% chance of being HIV positive; Peter suggests sky-high viral loads drove heterosexual transmission rates far beyond what treated HIV produces today. She estimates a country of 16 million may have lost around a million people.

• 03

  Oncology in the mid-1980s offered little: durable remissions for perhaps 10%.

  She describes patients with six months to live and few options, which drew her toward new mechanisms. At Bristol-Myers Squibb she taught herself SAS programming and, as the safety lead, helped make Taxol—a microtubule inhibitor with a new mechanism for drug-resistant cancers—safe to give, including managing reactions to its dissolving fluid.

• 04

  Genentech pioneered recombinant biologics and large simple trials.

  In the late 1970s Genentech proved to the FDA its fermented human growth hormone was pure. It 'bet the farm' on TPA for heart attacks, proving benefit in 30-day outcomes and building large simple trials and postmarketing registries. When she arrived in 1995 the company was struggling, and oncology was its future.

• 05

  Herceptin made targeted oncology real, but only for some patients.

  Herceptin targeted HER2-positive breast cancer; about three of four breast cancer patients weren't eligible, which motivated the search for more drugs.

• 06

  Rituxan rescued an antibody skeptics didn't believe in.

  IDEC, which made an anti-CD20 antibody for lymphoma, was going to run out of money around 1995–96 when Genentech partnered. CD20 is on B cells; marrow reserves of immature cells replace depleted ones, so patients can be treated repeatedly. Surprisingly, lymphoma patients didn't develop antibodies against it.

• 07

  Avastin emerged from Judah Folkman's angiogenesis idea.

  Napoleone Ferrara built an anti-VEGF antibody on the same backbone as Herceptin, about 93–94% human. A tough phase 2 in heavily pretreated metastatic breast cancer demanded at least 50% tumor shrinkage; colon cancer then became the phase 3 survival study.

• 08

  Drug development stages explained.

  Phase 1 is dose escalation with an art to choosing the starting dose—'because we are greedy in oncology, we always look to see if anybody responds.' Phase 2 can cost in the tens of millions; phase 3 confirms benefit. Even then, 500 patients is very limited for safety—5–10% may have side effects that stop them taking the drug.

• 09

  Avastin's price tested the value debate.

  At about $100,000 per course and roughly eight months' median survival gain, it failed the UK NHS's cost-per-QALY threshold. She acknowledges patient assistance programs didn't solve overall cost.

• 10

  Leading UCSF and the Gates Foundation required financial and cultural skill.

  She became UCSF chancellor in 2009 amid recession and faculty skepticism of an outsider; grateful patients were the key donor base. As Gates Foundation CEO from 2014, she focused on strategy, decision frameworks and an empowering, values-driven culture.

• 11

  Public trust in science eroded during COVID.

  She isn't convinced we're better prepared for a pandemic than in 2019. On ivermectin claims for stage 4 cancer, she says patients deserve data on dosing and harm; if a cheap drug cured cancer, pharma would modify and patent it immediately. She notes premature dismissal of the lab-leak question in May 2020 hurt credibility.

• 12

  AI could compress trials without cutting corners.

  Desmond-Hellmann argues AI shouldn't change what a trial tests, but it can speed the pieces—site selection, enrollment, data cleaning, monitoring—so a six-year program shrinks. She suggests interim looks: at three years, you might already see the trend. And approvals made in emergencies aren't wrong if later data show less effect—'any port in a storm'.