READERS' FORUM

Trial dropouts and study findings: Getting the analysis right

I thoroughly enjoyed "Your turn to learn to read: Evaluating articles about treatment in the medical literature" (May 2003) by Dr. Dimitri A. Christakis. Everything was just to the point. I would like to further clarify the problem of dropouts and emphasize the need to always consider dropouts in randomized controlled trials (called the "intention to treat" analysis). Investigators may fail to include, or even mention, the number of patients that could not complete the investigation for various reasons, often resulting in biased findings.

Consider, for instance, a hypothetical situation in which a new drug is being tried for intractable seizures for several months. Forty patients with seizures are randomly assigned to receive either the new drug or another, commonly used medication. Suppose that five patients (25%) of those randomized to receive the new drug cannot complete the study (see Table below). In the analysis, however, these five patients are not included, resulting in a 10/15 (67%) improvement rate for those completing the study, as compared with only 5/18 (28%) for the patients receiving conventional medical therapy (X² = 6.3; P <.01). The authors may, therefore, erroneously conclude that the new drug is more effective than the conventional drug.

If, on the other hand, the authors include the "intention to treat" group of patients in their analysis, the outcome is completely different. The proportion of successfully treated patients is now only 10/20 (50%) as compared with 5/20 (25%, X² = 2.0; P >.05). The conclusion this time is that the new drug is no different from the conventional drug.

In reading the literature, therefore, it is very important to read between the lines and determine if the authors include or even mention the percentage of dropouts in their work. I absolutely agree with Dr. Christakis that, when more than 10% to 15% of patients drop out in any arm of the study, the validity of the findings must be called into question. Furthermore, if nothing is mentioned about the dropout rate in the study, the study findings should be viewed with some reservation.

Philip Kum-Nji, MD, MPH
Pittsburgh, Pa.

Cultural similarities, cultural differences

As a pediatrician of Puerto Rican descent and an anthropology major, I looked forward to reading "The culturally competent pediatrician: Respecting ethnicity in your practice" (June 2003). Having gone to medical school in Wisconsin, my experiences with Hmong families were similar to those described in the article. The section on Latinos, however, was of concern, because it lumped all Latinos into one group. Latinos are such a diverse group that the cultural practices seen in families from Mexico, Puerto Rico, Guatemala, and throughout Latin America can be significantly different. If you apply the same principles to all these patients, you are stereotyping, which in itself can be insulting. In my experience, you are better off specifically asking families about practices with which you may be unfamiliar.

J. Rebecca Daumen, MD
Raleigh, N.C.

Author reply: We appreciate Dr. Daumen's concerns regarding the section on Latinos in our article. It was in no way our intent to suggest that all Latinos are the same or that they always share the same health beliefs and practices. In fact, we emphasize this very issue at the beginning of the section entitled "The Latino patient." We agree completely that Latinos are a heterogeneous group. However, there are, in fact, many normative cultural values, as well as folk illnesses, that are common to many Latino cultures. "Mal de ojo" is an example of one folk illness shared by many Latinos, including those from Mexico as well as regions of Central and South America. We believe that many health beliefs and practices of Latinos are similar enough to justify their presentation in an overview format.

In putting the article together, we struggled with the issue of "stereotyping" and did our best to minimize the potential. A discussion of each Latino "subgroup" was beyond the scope of our article. We believe that proper use of the "awareness-assessment-negotiation" model proposed by Pachter automatically eliminates the element of stereotyping from the interaction with the patient, because through "assessment" one is able to discuss whether an individual patient relates to a particular belief system.

Lori Taylor, MD
Lindia Willies-Jacobo, MD

Gloves for vaccinations? The jury is split

Just as many pediatricians never wear gloves when administering vaccinations as always wear them when giving these shots. So suggest the results of a Contemporary Pediatrics Web site instant poll. The poll, which received 319 responses, was prompted by readers' comments about the cover of the February 2003 issue of Contemporary Pediatrics (see "Pardon our appearance," May 2003, Readers' Forum). Voting in the Web site instant polls is anonymous and—like use of the site overall—is not restricted to pediatricians.

Readers' Forum.

Contemporary Pediatrics

September 2003;20:149.