| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Letters to the Editor |
Department of Radiology, Hospital Santiago Apóstol, Olaguíbel 29, 01004 Vitoria, Spain e-mail: eanorbe@hsan.osakidetza.net
Editor:
We have read with great interest the article by Patz and coworkers (1) about screening for lung cancer with computed tomography (CT). They commented on the biases (lead-time bias, length-time bias, and overdiagnosis bias) that may appear in any study in which the effectiveness of screening is attempted to be determined. The measurement they propose to avoid such biases is the disease-specific mortality rate (ratio of the number of deaths due to target disease to the total number of persons in the trial). Naturally, patients who die of the complications of treatment should also be included among those who die as a consequence of the disease (1).
It is evident that in some cases it may be difficult to know the cause of death, especially in patients who develop a second neoplasia. The long-term effects of treatment can likewise be difficult to determine. This was revealed in a large meta-analysis involving 20,000 women, in which there seemed to be an increase in mortality due to cardiovascular causes in patients given radiation therapy for early breast cancer compared with a control group who did not receive radiation therapy (2). The deaths attributed to vascular disorders would be the consequences of treatment and as such should be included among the deaths related to the target disease. However, it is unlikely that this happens, and the death would probably be mistakenly classified as not related to the disease. This wrong classification may lead to a belief in a beneficial effect of screening, as deaths secondary to treatment and therefore originating in the screening itself are reduced.
It could be better to use the total mortality rate (ratio of the number of deaths for any cause to the number of people taking part in the study), and not the mortality rate attributable to the target disease, as a parameter for determining the effectiveness of screening. The use of this rate would prevent the errors mentioned earlier.
In an ideal situation, the randomized clinical trial, the only difference between the two groups—the control group and the screened group—would be participation in the screening. Variations in the total mortality rate of each group would then only be attributable to the screening and would enable its beneficial effects, if any, to be known with more precision.
REFERENCES
and
Philip C. Goodman, MD*
Department of Radiology, Duke University Medical Center, Box 3808, Durham, NC 27710* e-mail: patz0002@mc.duke.edu
Department of Radiology, Dartmouth-Hitchcock Medical Center, Lebanon, NH
Drs Añorbe and Aisa raise excellent points about the evaluation of screening, and we welcome the opportunity to respond.
First, we completely agree that deaths caused by complications of treatment should be counted as deaths from the target disease.
Second, we agree that it may be difficult to determine whether a death has resulted from treatment, especially when the death occured long after the treatment. Furthermore, failure to properly attribute such deaths to the target disease would bias the disease-specific mortality end point in favor of screening. This bias has been recently described and given the name "slippery linkage bias" (1). There is another bias related to the determination of cause of death that biases the disease-specific mortality end point against screening. Because subjects in the screening group are more likely to be diagnosed with the target disease than are subjects in the control group, deaths from other cause may be more often falsely attributed to the target disease in the screened group than in the control group. This bias has been termed "sticky diagnosis bias" (1). These two biases can coexist in a randomized trial of screening (and in any other study design) and probably did in the Mayo Lung Project (1). However, that total mortality was higher in the screened group suggests that slippery-linkage bias had a greater effect than did sticky-diagnosis bias. In other words, more deaths caused than prevented by screening were probably obscured because of the combination of these two biases.
Finally, we agree that both of these biases could be completely avoided by using total mortality as the main end point in place of disease-specific mortality. However, no screening test findings have ever been shown to decrease total mortality to a significant degree, because deaths from the target disease have composed only a small proportion of all deaths (3%–16% in previous trials of cancer screening [1]). Perhaps, in the future, we will be able to identify cohorts at sufficient risk from a certain disease that a significant reduction in total mortality with screening would be demonstrable.
Regardless, we think that total mortality should always be reported along with disease-specific mortality so that readers can assess the validity of the latter end point and put the effect of screening into proper perspective.
REFERENCES
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
