CommentaryBMJ 1995; 311 doi: https://doi.org/10.1136/bmj.311.7015.1273a (Published 11 November 1995) Cite this as: BMJ 1995;311:1273
- Julie Morris
The findings reported by Boone and Watters are seriously limited by the low percentage (43%) of subjects traced for follow up. Although there are inevitable practical problems in following up an itinerant population in the tropics, consideration has to be given to how the results should be interpreted.
If only a low proportion of subjects are followed up biased results will be obtained if the success of follow up depends on either the original treatment (splenectomy or splenic conservation in this study) or the outcome measures (history of malaria, upper respiratory tract infection, etc). For example, suppose that for the subjects who had undergone splenectomy those who subsequently contracted malaria were more likely to remain in New Britain (and be available for follow up) than those who did not contract malaria. The estimated rate of malaria among the follow up group would therefore be a gross overestimate of the true rate. The comparison of rates for the splenectomy and splenic conservation groups would then be severely confounded if follow up of the latter group was not similarly related to a history of malaria.
Interpretation of the results for follow up subjects may be helped by calculating the minimum and maximum rates for the various outcome measures based on the assumption that all or none of the untraced subjects have the outcome. For the malaria rates in this study the minimum and maximum values are, respectively, 55% and 100% for the splenectomy group and 21% and 82% for the splenic conservation group, allowing for both the untraced and deceased subjects. The large range of values here reflects the high proportion of untraced subjects.