Quantitative methods in research on complementary and alternative medicine. A methodological manifesto. NIH Office of Alternative Medicine.

OBJECTIVES: This article summarizes the deliberations of the Quantitative Methods Working Group convened by the National Institutes of Health (NIH) in support of the NIH Office of Alternative Medicine. METHODS: The working group was charged with identifying methods of study design and data analysis that can be applied to empirical research on complementary and alternative medicine. This charge was broad and inclusive and addressed the evaluation of alternative therapies, the investigation of the basic science of complementary medical systems, studies of health promotion and disease prevention, and health services research. RESULTS: The working group produced a "methodological manifesto," a summary list of seven recommended methodological guidelines for research on alternative medicine. These recommendations emphasize the robustness of existing research methods and analytic procedures despite the substantive unconventionality of alternative medicine. CONCLUSIONS: Contrary to the assertions of many researchers and alternative practitioners, established methodologies (eg, experimental trials, observational epidemiology, social survey research) and data-analytic procedures (eg, analysis of variance, logistic regression, multivariate modeling techniques) are quite satisfactory for addressing the majority of study questions related to alternative medicine, from clinical research on therapeutic efficacy to basic science research on mechanisms of pathogenesis and recovery.

Causal modeling and alternative medicine.

Many problems of health promotion or prevention call for an understanding of relations among variables embedded in complex causal webs that may include psychosocial, cultural, or environmental factors as well as biological dysfunction. Experimental investigation of these kinds of research problems is frequently impossible or not feasible. Causal modeling, particularly latent variable structural modeling, can provide a useful alternative to manipulative experimentation when one is trying to build and test explanatory models in a rigorous and systematic fashion. A hypothetical model of fibromyalgia is presented to illustrate how latent variable models can prove useful when the malady under investigation is of relatively complex multifactorial origin. Topics discussed include the fundamental notion underlying causal models, how such models solve problems due to measurement error, and why both cross-sectional and longitudinal models deserve consideration. A number of applications in medical research are described.

Backward masking, information processing, and schizophrenia.

Many researchers have used backward masking to examine information-processing speed in schizophrenic subjects. The validity of this approach rests upon two main assumptions. One is that the mask effectively limits the time a previously presented stimulus is available for processing. The other is that the components of the masking mechanism in schizophrenic subjects are comparable to those in control groups. It is argued that the masking procedures used in these studies fail to meet either assumption. Alternative interpretations of these investigations, however, suggest important hypotheses for further research in our quest to understand information-processing deficits associated with schizophrenia.

Construct validity of various memory testing procedures.

The construct validity of Wechsler Memory Scale Logical Memory, Paired Associate Learning and Visual Reproduction subtests was evaluated, as well as the validity of the Benton Visual Retention Test, and the two memory factor scales of the Luria-Nebraska Neuropsychological Battery (LNNB). The results of a series of factor analyses based on test performances of 102 subjects indicated that delayed reproduction measures of visual memory were more valid than the traditional immediate reproduction administrations, which were more closely associated with visual-perceptual-motor abilities. Construct validity was also demonstrated for Logical Memory and Paired Associate Learning. Although both LNNB measures loaded on a memory factor, item heterogeneity and brief sampling of items raised serious questions about the clinical utility of these scales.

Gender differences in schizophrenia.

Lewine (Psychol. Bull., 90: 432-444, 1981) has proposed sex differences, specific to schizophrenics, in age at first psychiatric hospitalization, age at first reported symptoms, and premorbid social competence. To evaluate Lewine's hypothesis we collected data on 64 schizophrenic and 30 nonschizophrenic psychiatric outpatients. As no interaction between sex and diagnostic groups was found, our data failed to demonstrate sex differences specific and unique to schizophrenia. It is possible that Lewine's evidence is due to a cultural artifact.

Factor analysis of the WAIS and Wechsler Memory Scale: an analysis of the construct validity of the Wechsler Memory Scale.

The WAIS and Wechsler Memory Scale subtest scores of 256 neurologic and nonneurologic subjects were factor analyzed. The results supported the construct validity of the Wechsler Memory Scale as a measure of verbal learning and memory, attention and concentration, and orientation. Construct validity was not demonstrated for the Visual Reproduction subtest as a measure of visual memory. Suggestions are offered for future development and research on measures of visual memory.

Proprioception in schizophrenia.

B. Ritzler (J. Abnorm. Psychol., 86: 501-504, 1977) failed to replicate previous weight-lifting studies showing a proprioceptive deficit in schizophrenics. However, Ritzler did not use the same standard weights that the previous studies employed. Could this difference in procedure have caused his failure to replicate? Two experiments were completed to provide additional information about this question and related issues. In Experiment I, chronic hospitalized schizophrenics and normals were tested with light and heavy weights. Rosenbaum's original modified method of limits and the method of constant stimuli were utilized. Both methods provided comparable results, which, like Ritzler's data, failed to show a differential impairment in the discrimination of the light weights for paranoid schizophrenics. In Experiment II, outpatient paranoid schizophrenics, outpatient nonparanoid schizophrenics, outpatient psychiatric patients, and aged normals were evaluated utilizing the same psychophysical procedure and standard weights employed in the original studies. No differential impairment was found for the light weights for any group. These results provide strong support for Ritzler's position that a proprioceptive deficit is not unique to schizophrenia.

A study of schizophrenics' ability to localize the source of a sound.

Twelve process, nonparanoid schizophrenics, 12 nonschizophrenic psychiatric controls, and 12 normal controls were given an auditory localization task. Performance was measured by the number of correct localizations and was studied as a function of three variables: diagnosis, position of the apparatus in relation to the subject's median plane (right or left), and the degree of displacement of the auditory stimulus from a fixation point (4.5 degrees, 3.0 degrees, and 1.5 degrees). A three-way analysis of variance with two repeated measures resulted in a significant main effect for each of the three independent variables. Orthogonal comparisons indicated no difference in performance between normal and psychiatric controls but a significant difference between schizophrenics and the combined controls. The degree of displacement of the auditory stimulus resulted in a significant linear trend in performance. There were no significant interactions. The present findings strongly suggest that process, nonparanoid schizophrenics suffer from a specific deficit in auditory localization.