Premium
“Absolute” or “added” predictive value?
Author(s) -
Koole O.,
Van den Ende J.
Publication year - 2010
Publication title -
tropical medicine and international health
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.056
H-Index - 114
eISSN - 1365-3156
pISSN - 1360-2276
DOI - 10.1111/j.1365-3156.2009.02465.x
Subject(s) - predictive value , medicine , test (biology) , pre and post test probability , statement (logic) , value (mathematics) , statistics , mathematics , political science , law , paleontology , biology
This is a classical example of misuse of predictive values. Although the authors admit in the discussion that ... the negative predictive values were high not because the immunological criteria were a powerful test, but because only few patients developed virological failure , they conclude that these criteria are (more) appropriate for ruling out. What is important in validating predictors is the added value they offer compared to pre-test probability. In this study, the prevalence of treatment failure is 1.7%, the prevalence of success is 98.3%. With the WHO criteria CD4 results (compared to >10,000 copies on two measurements as reference test), the probability of success increases from 98.3% to 98.5%, an increase of a mere 0.2%. This is clearly a negligible gain, invalidating the statement that (CD4 criteria) are more appropriate for ruling out . When pre-test probability almost equals posttest probability, a test is of no value (Table 1). Notwithstanding this critique, we think this study is invaluable, in confirming that (i) CD4 count, contrary to our former beliefs and worldwide efforts of providing testing, has little place in evaluating treatment failure or success; (ii) we are in need of affordable, point of care tests