Roche Holding AG is focusing part of its pipeline on the riskiest area of drug development: the human brain.

The Swiss drugmaker had its last top-selling brain therapy when Valium revolutionized anxiety treatment in the 1960s. Now Roche is betting it can use the tools that have since made it the world’s biggest maker of cancer drugs to penetrate the “black box” of Alzheimer’s and autism, CEO Severin Schwan said.

“If you think back 15 years, a cancer tumor was like a black box,” Schwan said in an interview at the company’s Basel headquarters. “The brain has been a black box until recently. And now we start to understand what is happening on a molecular basis in the brain.”

The stakes are different now than when Valium inventor Leo Sternbach was cooking up his new class of medicines. Roche’s brain drugs are just one component in a late-stage pipeline packed with risky projects, said Karl Heinz Koch, a Helvea SA analyst. Treatments for central nervous system diseases also have an “enormous” potential payoff, said Andrew Baum, a London-based analyst for Citigroup Inc.

“The central nervous system will remain the highest of the high-hanging fruit,” Baum said.

A hint of whether the gamble may pay off is due in the second half of this year. Indianapolis-based Eli Lilly and Co. and Pfizer Inc. plan to announce results for Alzheimer’s drugs that attack the same protein as Roche’s experimental drug, called gantenerumab. If the Lilly therapy succeeds, its sales may exceed $9 billion a year by 2020, according to analysts at Sanford C. Bernstein Ltd.

The path to patients may be rockier than Valium’s. Sternbach’s anxiety drug took just five years to progress from the test tube to its approval in 1963. That’s less than half as long as the process is likely to take under current testing standards, which are more stringent, according to Tufts Center for the Study of Drug Development research.

Roche needs new pipeline prospects after the failure this month of a good-cholesterol pill that UBS AG analysts might have had estimated peak sales of $6.8 billion.

“For the company to create value medium- to longer-term, you need the ex-oncology business to take shape,” Koch said. Yet investors have even lower expectations for Roche’s central nervous system pipeline than they did for the cholesterol drug, he said. “That pipeline is just very risky.”

The Roche brain drug furthest along in the patient testing process, a schizophrenia medicine, didn’t have statistically significant results in an earlier trial if all patients who entered the trial were measured, Koch said. The compound had an effect on patients who stuck with it and who followed study protocols perfectly.

The Swiss drugmaker is also running clinical trials on treatments for depression and autism. Meanwhile, the field of Alzheimer’s, where four of Roche’s 10 brain drugs in patient testing are concentrated, is littered with failures.

The last new therapy for Alzheimer’s was Namenda, a Forest Laboratories Inc. medicine approved in 2003, according to data compiled by Bloomberg. Namenda, like all other Alzheimer’s drugs on the market, addresses symptoms without slowing the disease’s march through the brain.

Lilly’s solanezumab, which is similar to Roche’s Alzheimer’s drug, has at most a one-in-five chance of success, according to Bernstein’s Anderson. Lilly said on April 25 that one final-stage trial of the drug had finished that month, with a second slated to complete in June. The first look at the results could come in July, Anderson said in a note to clients.

Both the Lilly and Roche medicines target a reduction in beta amyloid, a protein that accumulates in the brains of Alzheimer’s patients. Pfizer, Johnson & Johnson and Elan Corp. are also expecting to release results in the third quarter for a drug to target beta amyloid, bapineuzumab, and have said they’ll seek regulators’ approval by the end of the year if it’s successful.

The cloud hanging over all three drugs is the earlier failure of a Lilly drug called semagacestat, which tried to zap beta amyloid by inhibiting an enzyme called gamma secretase that’s tied to the production of the protein. The pill actually worsened patients’ ability to do day-to-day activities, Lilly said in 2010. The U.S. company gave up on the medicine that year.

Lilly has said that semagacestat failed not because beta amyloid isn’t the right target, but because gamma secretase wasn’t the right way to go after the protein.

Still, investors expect both the new Lilly and Pfizer drugs to fail just as the older Lilly compound did. Two-thirds of 209 analysts and fund managers in a survey conducted last month by New York-based ISI Group thought neither would succeed in this year’s trials.

“There was a lot of hope riding on these types of approaches,” said Cord Dohrmann, chief scientific officer of Evotec AG, a German biotech working with Roche on another therapy for the disease. “There was no obvious success. People are wondering, are we really on the right track.”

The Evotec-partnered drug started out with Roche, which licensed the compound to Evotec in 2006. The German biotech tested it unsuccessfully as an anti-smoking drug. After it failed to help smokers quit, Roche signed a deal last September to finish developing the drug in Alzheimer’s instead.

“The original plan was always Alzheimer’s,” Dohrmann said. Smoking was an easier indication for a small biotech to pursue, with shorter, less expensive clinical trials, he said. And the drug did show itself to be safe--an important characteristic, he said.

By taking the drug back, Roche gave itself a candidate that doesn’t go after beta amyloid, he said. Instead it targets monoamine oxidase type B, or MAO-B, an enzyme that breaks down dopamine in the brain and contributes to the production of free radicals.

“It is such a wide-open market,” Dohrmann said. “People have made estimates that just slowing progression for a year would have a huge economic impact. That is saying a lot for a disease that people have for many, many years.”

Roche is forging ahead with beta amyloid as well. Gantenerumab isn’t as far along in the testing process as the Lilly and Pfizer drugs--it’s in the second of the three stages of clinical trials usually required for regulatory approval in the U.S. Roche is recruiting patients for a two-year, 360-person trial. If that’s successful, the company will test the compound in a larger patient group, with the aim of having it on the market after 2016.