Starting cholesterol-lowering statin therapy within six months of a cancer diagnosis does not improve patients’ survival at three years, according to a study that found scientific bias in previous research.
The study, “Examining Bias in Studies of Statin Treatment and Survival in Patients With Cancer,” was published in JAMA oncology. It dealt with the findings of previous observational research — studies that are more prone to bias because they do not include control groups.
A number of previous studies have reported that cancer patients who use statins do significantly better than those who don’t. In fact, research has shown that 18-44 percent of prostate, breast, and colorectal cancer patients who are on statins survive longer than those who aren’t on them.
Statins also lengthen the lives of people without cancer, studies have shown. This has intrigued the scientific community and ignited further research in the area.
Researchers decided to look at whether two kinds of scientific bias could have played a role in studies that suggested statins increased cancer patients’ survival.
Actually, three types of bias could have creeped in, they said.
One is bias from confounding variables. This could have occurred if the results were not properly adjusted for factors associated with the start and completion of statin therapy.
Another possibility is selection bias. Those conducting the studies could have included many people who were using statins at the time of their cancer diagnosis.
The third possibility is immortal-time bias. It could have surfaced when researchers classified study participants who died before receiving statins as non-statin users. This bias would make a connection between statins and cancer survival appear stronger than it actually is.
While confounding variables bias is inherent in an observational study, the other two kinds would have come from flaws in the study design.
The researchers decided to investigate whether selection bias and immortal-time bias could explain the correlation between statins and cancer patient survival.
They looked at the records of 17, 372 cancer patients in the Surveillance, Epidemiology, and End Results (SEER)-Medicare database. The registry covers people with colorectal, breast, prostate, or bladder cancer.
The team first analyzed the data in a way that would reduce selection bias. Then they analyzed it in a way that would reduce immortal-time bias. Finally, they analyzed it in a way that would reduce both selection and immortal-time bias.
They discovered that when selection and immortal-time bias is eliminated from the study design, there is no difference in cancer survival between statin users and non-users.
The bottom line, the team said, was that bias can lead to the wrong conclusions in observational studies.