In a blow to its plans to grow beyond oncology, Roche has delayed development of its taspoglutide by at least 12 to 18 months after studies showed a higher-than-expected rate of side effects, including skin reactions, digestive symptoms, cardiovascular and respiratory probelms. (here is the statement from Ipsen, which licensed the drug to Roche). More time is needed to identify patients who may be sensitive to the drug and and remove them from the trials.

Roche’s oncology franchise, which includes Herceptin, Avastin and Tarceva, accounts about 50 percent of its pharma sales. The drugmaker planned to seek FDA approval next year for its med, a once-weekly injectable that would compete with Byetta, a twice-a-day injectable from Lilly and Amylin, which hope to transform their own diabetes franchise with Bydureon, another weekly shot. “It’s an important drug for Roche and they need it on the market to balance out the oncology franchise, which is starting to slow down,” Bank Bontobel analyst Andrew Weiss tells Bloomberg News.

Meanwhile, Sanofi-Aventis is lambasting a study in Diabetes Care that suggests its Lantus diabetes drug, a $3.8 billion seller also known as insulin glargine, may be linked to an increased risk of cancer. A year ago, a University of Texas Health Science Center researcher told a conference call that studies would show Lantus was tied to cancer, although a paper in Diabetologia delivered mixed results amid debate over the release of the data and the researcher’s ties to other drugmakers. The FDA later said there was no link (background here).

The latest study is “unclear” and “lacks precision,” Jean-Pierre Lehner, Sanofi’s chief medical officer, tells Bloomberg, and can be “methodologically challenged.” And Nick Turner, an analyst at Mirabaud Securities, says there’s no difference between the latest study and last year’s controversial paper. The new study reviewed health records of 112 people with diabetes who developed cancer to 370 patients matched for age, sex and weight to determine differences that may have contributed to tumor development. A link to cancer was seen only in those getting higher doses of Lantus, but no other insulins.

UPDATE: “The bottom line, both from this new analysis and those that have preceded it, is that nothing can be said for certain about the possibility of a link between Lantus and cancer,” writes Sanford Bernstein analyst Tim Anderson in a note. “This works both ways: it means nothing can be proven with certainty, but it also means nothing can be disproved. In other words, there is still the possiblity that Lantus raises the risk of cancer by a slim margin…the financial impact could be immense.”

He goes on to say there is a biological plausibility. For one, it is commonly accepted that insulin is effectively a growth factor and may cell growth. Also, he recalls that, some years years, an experimental Novo Nordisk insulin analog was scuttled because of a well-documented increase in the risk of cancer in animal studies, providing direct evidence that certain insulin analogs may cause cancer. And Lantus is also an insulin analog, as is Lilly’s Humalog, and Novo Nordisk’s Novolog and Levemir. He notes, meanwhile that Sanofi is conducting two retrospective analyses of medical claims and a case control study to examine the issue.