Dr. Kessler and Colleagues Reply

To the Editor: Dr. Olfson cites the global kappa value from our overview article on the validity of the National Comorbidity Survey diagnoses (Kessler et al., 1998). He apparently did not read our more detailed report on the validity of the National Comorbidity Survey assessment of generalized anxiety disorder (1), in which we showed that the low concordance between a modified version (UM-CIDI) of the Composite International Diagnostic Interview and the SCID was due to a controversial component of DSM-III-R criterion A, that the worries must be excessive or unrealistic. This requirement was revised in DSM-IV. As reported in our 1995 article (1), kappa increases to 0.66 (with a positive predictive value of 0.63) when the excessive/unrealistic requirement is taken out of consideration. This kappa value is as high as the test-retest consistency of the SCID (2).

On the basis of this observation, I find it difficult to accept Dr. Olfson’s conclusion that the failure of his small primary care study to replicate the National Comorbidity Survey’s finding of significant impairment in pure generalized anxiety disorder was due to the imprecision of the National Comorbidity Survey’s generalized anxiety disorder assessment. A more plausible explanation, in my view, is the one proposed in our article, but not mentioned in Dr. Olfson’s letter: that Dr. Olfson’s small study contained only four respondents with pure generalized anxiety disorder, making it impossible to assess the impairment of pure generalized anxiety disorder with any precision. Consistent with this interpretation, a much larger primary care study cited in our article, but not mentioned in Dr. Olfson’s letter (3), yielded results consistent with those of the National Comorbidity Survey and inconsistent with Dr. Olfson’s data.

Dr. Olfson also commented on the difference between statistical and substantive significance. He is, of course, correct in noting that this is an important difference. However, he is incorrect in saying that the clinical significance of our finding regarding the impairment of pure generalized anxiety disorder is unclear. The fact that 78.2% of community respondents with generalized anxiety disorder reported no work impairment means that 21.8% did report work impairment. This percentage is not meaningfully different from the 24.7% rate of work impairment found among respondents with pure major depression. Both these percentages were dramatically higher than the 5.3% rate of work impairment found among respondents with neither generalized anxiety disorder nor major depression. These effects on work impairment are much larger than those found for the vast majority of other chronic conditions (4) and account for literally millions of dollars in lost productivity in the United States each year.