The speech analysis approach to determining onset of improvement under antidepressants
Introduction
How do effective antidepressant drugs really act? This question, while being of great practical and theoretical relevance, is still unsolved – despite decades of research into the putative mechanisms of action of various kinds of antidepressants. The facts in this field of research are indeed puzzling: (1) Antidepressants of large biochemical and pharmacological differences have virtually the same efficacy in terms of the proportion of patients in whom they induce a therapeutic response, (2) None of the newer treatments available to date, though having better tolerance and safety profiles, offers an advantage over the first-generation agents of tricyclics, (3) There is a large proportion of patients (35–45%) suffering from a refractory depression who are therapy-resistant under all drug treatment modalities, (4) There is a large proportion of placebo responders in active treatment groups (typically 30–40% among moderately to severely depressed patients), and (5) Not more than 20–25% of patients are, on average, “true” drug responders.
In view of the obvious dominance of placebo responders under all kinds of antidepressant treatment, an answer to the question of what makes “true” drug responders distinguishable from placebo responders is expected to imply an important step towards understanding antidepressant drug effects. Abrupt and fluctuating improvement has been hypothesized to be an indicator of placebo response, since patients eliciting nonpersistent, abrupt response to antidepressants have a higher relapse rate than persistent responders (Quitkin et al., 1987, Quitkin et al., 1993). Accordingly, an increasing number of investigations into the time course of recovery from depression under active drugs, as well as under placebo, has been published during recent years (e.g. Quitkin et al., 1987, Katz et al., 1987, Khan et al., 1989, Stassen et al., 1993, Tollefson and Holman, 1994, Entsuah et al., 1994, Montgomery, 1995, Möller et al., 1996, Stassen et al., 1996).
Findings, now replicated across several different antidepressant drug classes and placebo (Stassen et al., 1996, Stassen and Angst, 1997), suggest that differences between the efficacy of active drugs and placebo are reflected only by the total number of responders, and by the number and the time distribution of premature withdrawals, while the time course of improvement among responders appears to be independent of the treatment modality. In terms of the time characteristics of improvement, the therapeutic effect of antidepressants seems to be virtually independent of the biochemical mode and of the primary site of action within the monoaminergic systems, and effective antidepressants seem to merely trigger and maintain the conditions necessary for improvement. Once triggered, the recovery appears to follow its “natural” course identical to that of spontaneous remission, as observed under placebo treatment (Parker, 1996).
All this, at odds with “delayed onset of antidepressants”, challenges much of our current understanding of the course of recovery from depression for those receiving antidepressant drugs. On the other hand, the standard design of antidepressant drug trials, as used in the above studies, limits the value of the corresponding results: the assessment of the patients' responses are made at fixed points in time by their doctors who might be inclined toward reporting positive outcomes. For obvious reasons, self ratings are not suited to solve this problem. In consequence, results may be explained, to a certain extent, by the expectations of raters and patients. To overcome these methodological difficulties, “objective” assessments of the patients' improvement are required at a sufficiently dense time resolution throughout the first 2 weeks of treatment.
A computerized analysis of the patients' speaking behavior and voice sound characteristics throughout the time course of recovery from depression is often regarded as a promising way to revealing “objective” indicators of clinical change. Indeed, clinicians routinely observe that the speech of depressed patients is uniform, monotonous, without expression, and that depressed patients speak in a low voice, slowly, hesitatingly (Greden and Caroll, 1980). The results of previous studies in the field – most of which evaluated the speech characteristics of depressive patients at the beginning of hospitalization and at the time point of hospital release – generally underlined the potential value of computerized speech analyses with respect to monitoring the time course of improvement under antidepressants (Hargreaves and Starkweather, 1964, Greden and Caroll, 1980, Nilsonne et al., 1988, Alpert et al., 1993, Kuny and Stassen, 1993, Baer et al., 1995, Stassen et al., 1995, Sobin and Sackeim, 1997).
Our present investigation was designed as a model-finding study, aimed at assessing changes in speaking behavior and voice sound characteristics in hospitalized depressive patients throughout the first 2 weeks of antidepressant treatment, and at evaluating the relationships between speech parameters and psychopathology scales. Within the scope of this study, we especially aimed at developing a “practical”, computerized speech analysis procedure which would be routinely applicable to most patients, would be easy to carry out in a standard form by a technician, and would be relatively inexpensive and not too time-consuming. To this end, we relied upon a normative study on 192 healthy volunteers, stratified according to age, gender and education. The specific design of this study with three different types of text and two repeated assessments at 14-day intervals allowed us to establish the thresholds necessary to distinguish between “normal” fluctuations and “significant” changes of speech parameters. Specifically, our study addressed the following questions: (1) What proportion of patients show improvement at early stages of treatment, i.e. a prespecified reduction of depressive symptomatology within the first 12 days of study; (2) To what extent does early improvement predict response to treatment; (3) To what extent does measured early improvement depend on the criteria used to assess improvement; (4) To what extent does early improvement depend on the severity of depression at baseline; (5) Can early improvement be independently validated through objective parameters.
Section snippets
Patient sample
Our sample2 was composed of 43 hospitalized depressive patients (20 males and 23 females) with a mean age of 48.3 years (SD 15 years, range 24–85 years). Of these patients, 33 were diagnosed as manic-depressive psychoses, depressed type or currently depressed (ICD9: 296.1/296.3), 5 as schizoaffective psychoses, depressive state (ICD9: 295.7), 2 as depressive states (ICD9: 311), 2 as neurotic depressions
Time course of improvement in terms of psychopathology scores
Crucial to all investigations into onset and time course of improvement is the appropriate definition of “improvement”. In the literature, authors mostly rely on differences in the total score of depression scales, such as the HAMD, and define “improvement” through a significant baseline score reduction: (1) Relative improvement, is said to occur if a patient achieves a prespecified percentage baseline score reduction, regardless of the subsequent course of depressive illness, which may involve
Discussion
Lacking specific knowledge about the biochemical mechanisms underlying effective antidepressive drug treatment, the onset of action of those drugs can be studied only indirectly through an analysis of the time course of improvement in a sufficiently representative sample of depressive patients. If there were a 3 week latency of antidepressive drug action this should become visible in the distribution of time spans to onset of improvement. The present study yielded no evidence for a delayed
Acknowledgements
Supported in part by a grant of the Swiss National Science Foundation (SNF 32-42542.94)
References (25)
- et al.
Mapping schizophrenic negative symptoms onto measures of the patient's speech: set correlational analysis
Psychiat. Res.
(1993) - et al.
Cumulative mean change procedure: application to a comperative trial of venlafaxine, imipramine, and placebo in the treatment of major depression
Progr. Neuropsychopharmacol. Biol. Psychiatry
(1994) - et al.
Onset of response in relation to outcome in depressed outpatients with placebo and imipramine
J. Affect. Disord.
(1989) - et al.
Speaking behavior and voice sound characteristics in depressive patients during recovery
J. Psychiat. Res.
(1993) - et al.
Time course of improvement under antidepressant treatment: A survival-analytical approach
Eur. Neuropsychopharmacol.
(1993) - et al.
Speaking behavior and voice sound characteristics associated with negative schizophrenia
J. Psychiat. Res.
(1995) - Angst et al., 1969. AMDP: Das Dokumentationssystem der Arbeitsgemeinschaft für Methodik und Dokumentation in der...
- et al.
Is a cut-off score a suitable measure of treatment outcome in short-term trials in depression? A methodological meta-analysis
Hum. Psychopharmacol.
(1993) - et al.
Automated telephone screening survey for depression
J. Am. Med. Assoc.
(1995) - et al.
Decrease in speech pause times with treatment of endogenous depression
Biol. Psychiatry
(1980)
HAMD: A rating scale for depression
J. Neurosurg. Psychiat.
Voice quality changes in depression
Language Speech
Cited by (0)
- 1
This publication is dedicated to Dr. Stephen Kuny, who died August 13, 1995.