Two groups of investigators are interested in the problem of esophageal cancer in a population that has a high prevalence of hot beverage consumption. One group conducted study A, which showed that hot beverage consumption is associated with esophageal cancer, with an odds ratio (OR) of 1.51 and a 95% confidence interval of 1.19 to 1.81. The other group conducted study B, which demonstrated an OR of 1.46 and a 95% confidence interval of 0.95 to 1.97. Which of the following statements is most likely true concerning the results of these 2 studies?
A)The odds ratio estimate obtained in study A is likely not valid.
B)The p-value for the odds ratio estimate in study B is likely

Question

Two groups of investigators are interested in the problem of esophageal cancer in a population that has a high prevalence of hot beverage consumption.  One group conducted study A, which showed that hot beverage consumption is associated with esophageal cancer, with an odds ratio (OR) of 1.51 and a 95% confidence interval of 1.19 to 1.81.  The other group conducted study B, which demonstrated an OR of 1.46 and a 95% confidence interval of 0.95 to 1.97.  Which of the following statements is most likely true concerning the results of these 2 studies?
A)The odds ratio estimate obtained in study A is likely not valid.
B)The p-value for the odds ratio estimate in study B is likely < 0.05.
C)The results of study A are not statistically significant.
D)The sample in study B is likely poorly selected.
E)The sample size in study B is likely too small to detect a difference.

Quizplus · Accepted Answer

Explanation: Although both studies have almost the same point estimate of association, with an odds ratio (OR) of ~1.5 (suggesting increased odds of esophageal cancer in patients who consume hot beverages), the confidence interval (CI) of study B is wider than that of study A; it also includes the null value (1.0 for OR), indicating that the result is not statistically significant. The wide CI in study B is most likely due to a smaller sample size leading to insufficient power to detect the difference between exposed and unexposed subjects, thereby resulting in a lack of statistical significance. In other words, studies A and B reached similar values for the OR point estimate, but because of the smaller sample size of study B, its estimate did not reach statistical significance. Increasing the sample size of study B would increase its power and make the CI narrower. By doing so, the null value would likely have been excluded (as in study A), thereby reaching statistical significance. Although there are several causes of a wide CI, small sample size is a common one. (Choices A and D) Although the accuracy or validity of the studies cannot be fully evaluated without any information on how the studies were designed and conducted, the fact that the ORs obtained in both studies are similar makes it less likely that the results of study A are invalid (as the OR obtained in study B would be invalid as well) or that the sample in study B was poorly selected (as the OR point estimate would likely be different from that seen in study A). (Choice B) If the CI of an estimate reflects lack of statistical significance, then the corresponding p-value should also reflect the same conclusion. Therefore, if the 95% CI includes the null value, the p-value should be > 0.05 (not statistically significant). Conversely, if the 95% CI does not include the null value, the p-value should be < 0.05 (statistically significant). corresponding p-value (Choice C) The 95% CI in study A does not include the null value; therefore, the OR achieves statistical significance. Educational objective: The power of a study is the ability to detect the difference between 2 groups (eg, treated versus nontreated, exposed versus nonexposed). Increasing the sample size increases the power of a study. As a result, the confidence interval of the point estimate (eg, odds ratio) becomes narrower.

Two Groups of Investigators Are Interested in the Problem of Esophageal