StATS: Contradictory research (created 2006-03-30).

The Washington Post published an article that gets to the heart of the difficulty with Evidence Based Medicine. The URLs for these newspaper sites are always a bit problematic, because each newspaper has a different policy on how long they keep these URLs active and who is allowed to see them.

It starts out with a rather provocative statement, but one that is very hard to argue with:

The public is bombarded with messages about diet and cancer prevention. Unfortunately, the advice is pretty inconsistent. One day a diet prevents cancer, the next day it doesn't. In the early '90s, beta carotene (a vitamin A precursor present in fruits and vegetables) was said to prevent lung cancer. But several years later, headlines read, "Beta carotene pills yield no benefit" (The Post, 1996). And while people have been told for years to eat a high-fiber diet to reduce the risk of colon cancer, recently we were told "High-fiber diets are not anti-cancer miracle" (Montreal Gazette, 2005).

The most recent contradictory research studied the whether a low fat diet could be linked with a decrease in the risk of breast cancer. One study examined 48,835 post menopausal women without prior breast cancer.

The researchers randomly assigned 19,541 women to an intervention group where women were encouraged to consume on 20% of their calories from fat and increase servings of vegetables and fruits. The remaining women (29,294) were a control group and no dietary advice was provided.

The women in the intervention group did follow this advice to some extent. In the intervention group, the consumption of fat was reduced from 76 grams per day on average (SD=34) to 41 grams per day (SD=21). The control group saw little change with a mean consumption of 76 grams (SD=34) at baseline and 63 grams (SD=31) after one year.

The researchers followed these patients for an average of 8.1 years. The annualized incidence rate was 0.42% in the treatment group and 0.45% in the control group. This difference was not statistically significant (Hazard Ratio = 0.91, 95% CI 0.83-1.01) though you might argue that the result was borderline significant since the confidence interval just barely includes the value of 1. The p-value is 0.07.

The authors draw the following conclusion

Among postmenopausal women, a low-fat dietary pattern did not result in a statistically significant reduction in invasive breast cancer risk over an 8.1-year average follow-up period. However, the nonsignificant trends observed suggesting reduced risk associated with a low-fat dietary pattern indicate that longer, planned, nonintervention follow-up may yield a more definitive comparison.

This contradicts finding of another study, the Women's Intervention Nutrition Study. These findings were presented at the 2005 meeting of the American Society of Clinical Oncology. Although I could not find the original presentation, there are many summaries on the web:

This study recruited 2,437 women who had breast cancer and were treated either by masectomy or lumpectomy plus radiation. These women were randomly assigned to receive an intervention (975 women) or a control group (1,462 women). The intervention was a recommendation to eat a low fat diet. The intervention group had multiple coaching sessions from a professional dietician.

The intervention did have an impact, since the average women in the intervention group consumed 33 grams of fat per day (SD=17), considerably better than the control group (51 grams of fat per day on average, SD=24).

The results were impressive in this study. The recurrence of breast cancer was 12.4% in the control group and only 9.8% in the treatment group. Most summaries did not report a confidence interval or p-value, though the Washington Post article puts the p-value at 0.03.

Is it fair to say that two studies are contradictory when one has a statistically significant result (p=0.03) and the other has a borderline result (p=0.07). Is it fair to expect diet to have the same effect on women who are trying to prevent their cancer from recurring as it would on women who were trying to avoid their first diagnosis of cancer?

Another problem is the tendency for researchers and the media to overstate the impact of a positive finding. Why is it that a p-value that is just barely below the threshold of 0.05 is not also called a borderline finding?

age was written by Steve Simon while working at Children's Mercy Hospital. Although I do not hold the copyright for this material, I am reproducing it here as a service, as it is no longer available on the Children's Mercy Hospital website. Need more information? I have a page with general help resources. You can also browse for pages similar to this one at Category: Critical appraisal.