Cross design synthesis : a new strategy for medical effectiveness research : report to Congressional requesters / United States General Accounting Office.

  • United States. General Accounting Office.
Date:
[1992]
Catalogue details

Licence: Public Domain Mark

Credit: Cross design synthesis : a new strategy for medical effectiveness research : report to Congressional requesters / United States General Accounting Office. Source: Wellcome Collection.

    45/128 (page 43)
    Part 2: Checking Linkage of Baseline Characteristics to Treatment Effects Chapter 2 Methods for Assessing Randomized Studies’ Generalizability to the Patient Population (Task 1) nonrepresentativeness is simply a question of relative numbers. In addition to comparing single-characteristic distributions, joint distributions of multiple patient characteristics (for example, age by stage of disease) could be compared (randomized study versus data base). Such a comparison might reveal, for example, that a randomized study includes no patients who are both old and severely ill, even though some old and some severely ill patients were included. However, reviewers are limited by available information. If several randomized studies are being reviewed, the combined coverage of the set of randomized studies should be considered, particularly if complementary portions of the patient population are covered by different randomized studies. The reason is that even if no single randomized study covers the entire patient population, the combined set of existing randomized studies may come close to doing so. For example, if one randomized study covers male patients and a second randomized study covers female patients, together the two randomized studies cover both. This situation differs greatly from that in which all studies fail to cover the same patient group. Once the reviewer has assessed the representativeness of patient baseline characteristics, he or she addresses the next logical question: whether any nonrepresentativeness that was detected for these characteristics actually translates into nonrepresentativeness with respect to the treatment effect. The danger is quite simply that a treatment may be more or less effective for the kinds of patients participating in randomized studies than for other kinds of patients. As previously noted, patient nonrepresentativeness on such measurable variables as sex, age, or known prognostic factors may or may not translate into a lack of generalizability for the estimated treatment effect. For example, suppose that a randomized study’s patient pool does not represent all patient ages or that it does not represent all stages of the disease being treated. If the treatment has different effects on patients in different age groups or on those with different stages of the disease, then the results of this study will not be generalizable to the full population of patients.!! But if the treatment in question is equally effective for patients ‘This would be the case if, for example, younger patients responded well and received a great deal of benefit, but older patients were helped little or not at all.