Fishbain, David A. MD, FAPA; Cutler, R. B. PhD; Rosomoff, H. L. MD, DMedSc; Rosomoff, R. Steele RN, MBA
CP = chronic pain; CPPs = chronic pain patients; SSRIs = selective serotonin reuptake inhibitors; DSM = Diagnostic Statistical Manual of Mental Disorders; DSM-IV = Diagnostic Statistical Manual of Mental Disorders, 4th Edition; Meta = meta-analysis.
Meta-analysis is a statistical approach developed for the purpose of research integration, where the results of several studies in a specific area are combined mathematically and analyzed [1,2]. The most important feature of meta-analysis is the conversion of individual study outcomes to a common metric, such as standardized mean differences (or effect size), followed by statistical analysis on the collection of outcomes . This makes it possible to estimate the magnitude of the relationship found so that the clinical importance of the effect is more evident . This type of review has become necessary, because there is often disagreement in traditional reviews over the conclusions reached by different reviewers in the same field . There is now some empirical evidence that meta-analysis reviews may be superior to the traditional review . Thus, the meta-analysis review has become the "supreme court" in determining whether an effect exists and some authors believe that meta-analysis will be the way in which many arguments about treatment effectiveness will be ended .
The analgesic effect of antidepressants for the treatment of chronic pain has been evaluated in many double-blind studies [7-10]. Previously, these studies have been subjected to two meta-analyses and two analytic reviews. In the first analytic review, Goodkin and Gullion  reviewed 53 antidepressant trials. Of these, 16 trials met minimal criteria but none of these trials met all four types of validity criteria. Goodkin and Gullion , therefore, did not perform a meta-analysis on these trials, which is an attempt to pool the results across trials . However, they reported that 56% of these studies reported a significant improvement in pain when the active drug was compared with placebo. In addition, 71% of the headache treatment studies reported an improvement in the frequency and severity of pain . In 1992, Onghena and Van Houdenhove  performed a meta-analysis on 39 placebo-controlled trials on the analgesic effect of antidepressant treatment for CP. The results indicated that antidepressant-treated CPPs had less pain than 74% of the CPPs who received a placebo . In addition, the results indicated that antidepressants acted through an analgesic effect and not via masked depression, manifest depression, sedation, or a placebo effect . In 1995, Goodkin et al. , again in an analytic review, reported on 17 post-1987 trials of antidepressant treatments for chronic pain. In this review, only 29% of the trials met design and protocol criteria and a meta-analysis, therefore, was not performed. Goodkin et al.  wrote that "antidepressants cannot be concluded to be a generally effective treatment modality (for chronic benign pain syndromes), although we previously have reported  that they might be effective for headache." However, this study was severely criticized [11,12] for several reasons. First, the study was "selective" and not a systematic review and did not statistically pool results across studies, but yet the authors claimed that it was a meta-analysis . Second, the design and protocol criteria used as inclusion criteria were too rigid  and the "vote counting" procedure used was statistically inadequate . The second meta-analysis  on the analgesic efficacy of antidepressants used only CPPS suffering from "neuropathic pain." This is usually pain associated with some form of nerve injury or degeneration, such as diabetic polyneuropathy. Here, 21 placebo-controlled trials were included, 17 of which were randomized controlled trials . In 6 of 13 diabetic neuropathy and in 2 of 3 postherpetic neuralgia studies, the odds ratio showed significant benefit compared with placebo. The authors concluded that antidepressants are effective in relieving neuropathic pain and that compared with placebo in 100 patients with neuropathic pain, 30 patients will obtain more than 50% pain relief . The results of this meta-analysis on neuropathic pain supported the conclusions of other reviewers in this area .
This literature contains a significant number of antidepressant treatment studies of CPPs diagnosed as suffering from psychogenic pain or somatoform pain disorder. The above meta-analyses results, although only partially consistent, raised the issue of whether an antidepressant analgesic effect could be demonstrated for patients diagnosed with psychogenic pain or somatoform pain disorder. We, therefore, undertook a meta-analytic study of the analgesic effect of antidepressants in the treatment of psychogenic pain or somatoform pain disorder. This meta-analytic study is described below and, to the authors knowledge, is the first such study in the literature.
To locate all the relevant references, the following procedure was observed. Medline, Psycho. Info., Science Citation Index, and the National Library of Medicine databases were reviewed, using the MESH (medical subject heading) terms: Pain and antidepressants, psychogenic pain or somatoform pain disorder, and antidepressants. The searches were conducted back to 1966 except for Science Citation Index which was conducted back to 1974. A manual search was also performed, using pain treatment outcome literature, pain books, and pain meetings abstracts. English and non-English language literature was reviewed.
Chronic pain is now broken down into two large categories: neuropathic (neurogenic) and nonneuropathic (nociceptive) . Neuropathic pain includes such conditions as central pain, postherpetic neuralgia, avulsion of plexus, neuroma formation, phantom limb, lancinating neuralgias, nerve compression, painful polyneuropathy, and reflex sympathetic dystrophy/causalgia . Nonneuropathic pains would include such conditions as myofascial pain syndrome, fibromyalgia, spinal stenosis-associated low back pain, osteoarthritis-associated low back pain, psychogenic pain, and somatoform pain disorder . As such, any reports of pain treatment with antidepressants were first sorted into neuropathic and nonneuropathic pain conditions. Reports dealing with nonneuropathic pain conditions were then sorted additionally by type of pain treated. Reports dealing with the treatment of psychogenic pain or somatoform pain disorder were then isolated and sorted into a combined group. Studies dealing with tension headache were placed into the psychogenic pain/somatoform disorder group if the authors claimed that the treated patients' tension headaches were psychogenic. Case reports and nonplacebo-controlled, nonrandomized studies were then eliminated from the psychogenic pain disorder/somatoform pain disorder group. The neuropathic pain disorder group was also reviewed for any reports on psychogenic pain or somatoform pain disorder.
The isolated psychogenic pain disorder/somatoform pain disorder studies were examined independently by the first two authors of this study for inclusion in the analysis as well as for coding of relevant statistics. Discrepancies were resolved in conference between the two investigators. Studies were additionally eliminated unless they contained the minimum information necessary: a p value that compared the effects of antidepressants with placebo on pain intensity. If a study, however, did not have this p value, but one could be calculated, that study was not eliminated.
After this process was completed, the remaining studies were judged to be representative of the psychogenic pain and somatoform disorder studies. This was because we believed that we had located all of the antidepressant treatment studies of this population and used all of these in the meta-analysis except for those eliminated by exclusion criteria.
The remaining studies were analyzed statistically in the following fashion. For each study, a single p value for the drug/placebo comparison was found or calculated . All calculations were based on change scores between predrug pain intensities and postdrug pain intensities, for both the antidepressant and placebo drugs. If more than one statistic was reported for the comparison, the statistics were combined if the patients were different. However, if the patients overlapped, only one statistic was used. In these cases, the statistic selected was the one in which the length of time on the drug was greatest. Some studies reported only significant p levels and left it impossible to estimate p levels for other antidepressant-placebo comparisons; these studies were not used. One study did not provide a p value but stated that antidepressants had no effect. Here we took the conservative approach and set the p value at 1.00. Finally, to combine p values, it is necessary that they be one-tailed. This is because a two-tailed p value does not indicate direction of effect. For example, if two identical two-tailed p values (eg, p = .01) show results in the opposite direction, they should cancel each other out, but if they were in the same direction they should result in an increased significance level . Two-tailed values were converted to one-tailed. When it was impossible to determine whether a p value was one-tailed or two-tailed, the more conservative assumption, that the value is one-tailed, was used.
After a single overall p level was obtained for each study, z scores were calculated. For the study for which we set the p value at 1.00, the z score was 0. Because significant heterogeneity suggests that caution should be used in interpreting results, the z scores were checked for heterogeneity . A chi squared test for heterogeneity was used to test the hypothesis that the results of all the studies were similar . Next, the z scores were combined into an overall z score and an overall p value was obtained.
Each study was given equal weight in the analysis. The correlation between effect size and sample size was r = .24. This was primarily due to the zero effect size conservatively assigned to the Onghena and Van Houdenhove study . When this outlier was removed, the correlation was r = .09.
Effect sizes were then calculated from the overall z score and the z scores of each individual study. Finally, the number of hypothetical studies, with findings of no effect, necessary in order to render the overall p value nonsignificant (the "file drawer number") was calculated, and the maximum number of unpublished studies that might reasonably exist ("tolerance") was estimated . The file drawer number was calculated based on the overall z score obtained and the number of studies included in the analysis. Tolerance was calculated by multiplying the number of studies by five and adding 10 .
The selected studies were then coded for methodological problems as per the method used by Onghena and Van Houdenhove . These methodological and reporting problems were the following:
1. The drug side effect either invalidated double-blindness or were not reported;
2. Differential drop-outs either invalidated randomization or were not reported;
3. Depression was neither an exclusion or an inclusion criterion;
4. The assessments psychometric properties were not documented and doubtful;
5. The duration of the drug condition was less than 4 weeks;
6. There were problems with the statistical analysis;
7. The sample was too small to have a powerful test;
8. There was no dose specification reported;
9. Compliance was not monitored (either by counting the returned capsules or monitoring plasma levels of the drugs or metabolites) or the compliance as not reported; or
10. The use of analgesics was not reported.
These analyses were repeated on the five studies dealing primarily with headache pain and the remaining group of studies to determine whether the overall results could be confounded by the results of the antidepressant effect on headache pain.
The reference search yielded 155 reports in which antidepressants had been used for the treatment of pain. These references broke down into the following pain location/etiological categories: back pain, N = 11; neck pain, N = 1; pelvic pain, N = 2; cancer pain, N = 23; arthritic/rheumatic pain, N = 22; fibromyalgia, N = 8; facial pain, N = 9; ulcer pain, N = 6; headache (except tension), N = 29; mixed etiologies, N = 19; pain as a symptom of depression, N = 4; and psychogenic pain and somatoform pain disorder, N = 19. Of the 19 psychogenic pain and somatoform disorder antidepressant treatment reports [16-34], one  was a case report. Four studies [27,29,30,33] were not placebo-controlled. One study  did not contain information on controls, and two studies were repeats [31,32] of previously published studies [16,19]. The case report, nonplacebo controlled studies, no-information studies, and repeat studies were then eliminated from the psychogenic pain and somatoform pain disorder group. The remaining 11 studies [16-34] fulfilled the other meta-analytic inclusion criteria (see Methods section) and were, therefore, subjected to detailed analysis.
(Table 1) presents the characteristics of the 11 studies included within the meta-analysis. These characteristics are labeled: author/year study, type of pain (location), type of study, active drug dosage per day, sex distribution of patients treated with active drug, length of treatment trial, control for presence of depression, and type of pain rating scale. The significance of these items will be discussed below.
(Table 2) presents the results of the analysis comparing pain reduction as a result of treatment with antidepressants versus treatment with placebo. The 11 p values were not significantly heterogeneous (chi squared = 7.92, 10 df, NS). The combined difference showed that antidepressants decreased pain intensity significantly more than placebo (z = 5.71, p < .0001). Effect sizes near 0.1 are small, near 0.3 are medium, and near or greater than 0.5 are large . The overall effect size was large (mean = .48) and ranged from 0 to .91. The number of unpublished studies (file drawer number) showing no difference between drug and placebo that would be necessary to make these results nonsignificant was estimated to be 133. The number of nonsignificant studies that could be reasonably argued to exist (tolerance level) was calculated to be 65. These results indicate that a bias in reporting publishable data does not account for the significant difference between drug and placebo.
The combined effect size (ES) for headache pain was much greater than that for mixed pain, but both were at least moderate in size and significant (for headache, z = 4.75, p < .01, ES = .59; for mixed pain, z = 3.08, p < .01, ES = .39).
A significant amount of evidence indicates that some antidepressants have an analgesic effect . This evidence is supported by the meta-analyses reported by Onghena and Van Houdenhove  and McQuay et al. . Interestingly, the results reported by Onghena and Van Houdenhove  also indicated that in CPPs, antidepressants acted through an analgesic effect rather than through improvement in manifest depression, improvement in masked depression, or sedation or a placebo mechanism. Our meta-analysis results also indicated that in CPPs diagnosed with psychogenic pain or somatoform pain disorder, antidepressant treatment resulted in a reduction in chronic pain that was significantly greater than that of placebo. Such results were not expected, inasmuch as according to the definitions of these disorders, the analgesic response should be no greater than that of a placebo response. There are several potential explanations for these results that will be discussed below.
The first possible explanation is that within the reviewed studies the diagnostic criteria of psychogenic pain and somatoform pain disorder were not used correctly. There is some evidence from some other DSM diagnoses that this is not an insignificant problem . We have no way of determining whether CPPs with nonpsychogenic or nonsomatoform pain disorder were included in the studies used within the meta-analysis. This then could have served as a source of error.
The second possible explanation is that the criteria for psychogenic pain and somatoform pain disorder were not sensitive and specific enough to separate CPPs with those disorders from those CPPs with nonpsychogenic and somatoform pain disorder. The diagnostic sensitivity and specificity of these two diagnoses has never been tested. However, there is some evidence from a report on the outcome of emergency room patients diagnosed with DSM-III conversion disorder, which is relevant to the issue discussed here. This report [37,38] found that the psychological criteria for conversion disorder did not protect the psychiatric examiner from a false diagnosis [37,38]. In this study, all three patients were later found to have organic disease that was responsible for their conversion symptoms and this resulted in dire medicolegal consequences in all cases. It is to be noted here that the psychological criteria for DSM-III conversion disorder were exactly the same as the psychological criteria for psychogenic pain. As such, it is very possible that some CPPs with nonpsychogenic pain were diagnosed erroneously as having psychogenic pain. This then could also have served as a source of error.
The third possible explanation is that the other criteria of psychogenic pain and somatoform pain disorder other than the psychological criteria did not adequately discriminate between nonpsychogenic pain and nonsomatoform pain disorder versus CPPs with these disorders. In psychogenic pain, these other criteria stated that the pain had to be inconsistent with anatomic distribution, lack organic pathology to account for it, or to be grossly in excess of that expected from physical findings. In somatoform pain disorder, these criteria were changed to "lack of organic pathology to explain the pain or the complaint of pain had to exceed that expected from the physical findings." Both of these sets of criteria present the following difficulties. Few CPPs totally lack demonstrable organic pathology as demonstrated by soft tissue physical findings . However, clinicians are not necessarily adept at identifying these soft tissue syndromes . Such a problem could have led to incorrect information being transmitted to the psychiatrist. In addition, even if physical findings indicative of soft tissue syndromes are identified, it is often unclear whether the pain demonstrated or experienced by the CPP exceeds that expected from the physical findings . Thus, the psychiatric examiner would then need to make a value judgment. These criteria problems could have then served as source of error in the analyzed studies.
The fourth possible explanation was that the apparent analgesic effect was actually a function of the antidepressant effect of the drug. Although, Onghena and Van Houdenhove  demonstrated in their meta-analysis that antidepressants acted through an analgesic effect rather than through improving manifest or masked depression, we were not able to draw the same unequivocal conclusion. Only three studies [16,19,20] included in this Meta controlled for depression. However, most of the studies either had treated their CPPs with subtherapeutic doses of antidepressant (Table 1) [17,19-25] or the treatment had been for less than 6 weeks [18,20,21]. These observations make it unlikely that the demonstrated analgesic effect was a function of the antidepressant properties of these drugs. Nevertheless, this issue could have confounded the results.
The fifth possible explanation is that of the sex ratio of the CPPs in the included studies. In most of the studies [16-19,23,24], there was a preponderance of female CPPs, whereas in the rest of the studies the sex ratio was not reported [21,22,25,34] (Table 1). Women may metabolize antidepressants differently from men. One study has demonstrated a significant sex-related difference in the nortriptyline to amitriptyline ratio, which was higher in female CPPs . The preponderance of women in these studies could then have also been a confounding factor, especially if the same sex distribution were not present in the placebo group.
The sixth possible explanation is that the CPPs, diagnosed with psychogenic pain or somatoform pain disorder, actually had forms or neuropathic pain. As pointed out in the Introduction, neuropathic pain has been shown to respond to antidepressants . As such, if CPPs with neuropathic pain were misdiagnosed in the included studies, this could have confounded the results of our meta-analysis.
A major inclusion criteria for the study selection of this Meta was the placebo-controlled randomized or placebo-controlled crossover study. Nevertheless, many of the selected studies demonstrated methodological problems (Table 2). Ten methodological problems were recorded. Most studies contained one or more of these problems, the average being four problems. The most common problem was that of assessment of psychometric properties not being documented. The second most common and serious problem was that of side effects invalidating double-blindness. The third most common problem was that of differential drop-outs, either invalidating randomization or not being reported. In all, these three methodological problems could have significantly impacted on the results of the studies included within the Meta and, thereby, could have confounded Meta results.
Of the 11 studies analyzed in this meta-analyses, five related to headache pain. Because it has been a clinical impression supported by analytic review  that headache pain may respond more robustly to antidepressants, we isolated these five studies and used them as a subgroup repeating our analyses. The results indicated that the combined effect size (2 = 4.75, p < .01, ES = 0.59) for the headache pain studies was much greater than for the remaining pain studies (2 = 3.08, p < 01, ES = 0.39). However, both effect sizes were greater than what would be considered a medium effect size . As such, we do not believe that the preponderance of headache studies within the psychogenic pain/somatoform disorder group confounded our results.
A final potential explanation for the obtained results relates to problems in implementing or performing any meta-analysis, thereby leading to faulty data. These problems have been outlined by many researchers and methods to circumvent these procedural problems have also been delineated [1-4,6,14,15]. In performing this meta-analysis study, we were careful in following the recommendations of these researchers by doing the following. We searched various sources of information to ensure adequacy of retrieval for all studies. This was especially important here, because many of the studies found were old and none had used SSRIs. This result would lead one to suspect that perhaps some recent studies were missed. However, this is not the case, inasmuch as it seems that recent antidepressant treatment studies for chronic pain usually use diagnostic groups other than psychogenic pain/somatoform pain disorder , eg, fibromyalgia. This is likely the case because of the diagnostic difficulties in actually identifying and being sure that patients with chronic pain actually do have a diagnosis of psychogenic pain/somatoform pain disorder . Therefore, we are confident that we had retrieved all or almost all of the applicable studies. Another potential source of error, that of publication bias, was controlled for by performing the file drawer calculation. We also specified and had strict inclusion/exclusion criteria. Two independent raters read the studies, making independent judgments about meta-analysis inclusion/exclusion. Meta quality was maximized by including only randomized placebo-controlled studies. Finally, tests for homogeneity were conducted. It is, therefore, unlikely that there were Meta implementation problems. Therefore, if the results of this Meta are incorrect, we would attribute this to the other potential explanations discussed above.
The results of this meta-analysis indicate that for CPPs diagnosed with psychogenic pain or somatoform pain disorder antidepressants have an analgesic effect above that of placebo. The potential reasons for this result have been discussed. Three of the potential reasons relate to the criteria of these disorders and to issues of the validity of these disorders in separating nonpsychogenic pain and nonsomatoform pain disorder CPPs from CPPs with these disorders.
Meta-analysis can be abused and nothing takes the place of well designed and adequately powered randomized clinical trial. Such trials now need to be conducted with the DSM-IV diagnosis of pain disorder to determine whether CPPs diagnosed with this disorder respond with an analgesic effect to antidepressants.
This work was supported, in part, by Grant H133A00032 from the National Institute of Disability and Disability Research. The authors thank Ms. Marilyn Green for typing the manuscript.
1. Pollick VE: Meta-analysis in literature reviews. Biol Psychiatry 34:345-347, 1993
2. Sacks HS, Berrier J, Reitman D, et al: Meta-analyses of randomized controlled trials. N Engl J Med 316:450-455, 1987
3. Schmidt FL: What do data really mean? Research findings, meta-analysis and cumulative knowledge in psychology. Am Psychol 47:1173-1181, 1992
4. Oxman AD, Guyatt GH: Guidelines for reading literature reviews. Can Med Assoc J 138:697-703, 1988
5. Beaman AL: An empirical comparison of meta-analytic and traditional reviews. Pers Soc Psychol Bull 17:252-257, 1991
6. Mann C: Meta-analysis in the breech. Science 249:476-480, 1990
7. Onghena P, Van Houdenhove B: Antidepressant-induced analgesia in chronic nonmalignant pain: A meta-analysis of 39 placebo-controlled studies. Pain 49:205-209, 1992
8. Goodkin K, Gullion C: Antidepressants for the relief of chronic pain: Do they work? Ann Behav Med 11:83-101, 1989
9. McQuay HJ, Tramer M, Nye BA, et al: A systematic review of antidepressants in neuropathic pain. Pain 68:217-227, 1996
10. Goodkin K, Vrancken MAE, Feaster D: On the putative efficacy of the antidepressants in chronic, benign pain syndromes. Pain Forum 4:237-247, 1995
11. Turner JA: Efficacy of antidepressant medication for chronic pain problems. Pain Forum 4:254-255, 1995
12. Max MB: Thirteen consecutive well-designed randomized trials show that antidepressants reduce pain in diabetic neuropathy and postherpetic neuralgia. Pain Forum 4:248-253, 1995
13. Waid SJ, Paitenay RK, Yaksh TL: Nociceptive models: Relevance to clinical pain states. In Basbaum AI, Beggon JM (eds), Toward a New Pharmaco-Therapy of Pain. Wiley, New York, 1991, 381-392
14. Rosenthal R: Judgement studies: Design, analysis and meta-analysis. New York, Cambridge University Press, 1987
15. Rosenthal R: The "file drawer problem" and tolerance for null results. Psychol Bull 86:638-641, 1979
16. Feinmann C: Psychogenic facial pain: Presentation and treatment. J Psychosom Res 27:403-410, 1983
17. Pilowsky I, Barrow CG: A controlled study of psychotherapy and amitriptyline used individually and in combination in the treatment of chronic intractable, "psychogenic" pain. Pain 40:3-19, 1990
18. Sjaastad O: So-called "tension headache"-The response to a 5-HT uptake inhibitor: Femoxetine. Cephalalgia 3:53-60, 1983
19. Zitman FG, Linssen ACG, Edelbrook PM, et al: Low dose amitriptyline in chronic pain: The gain is modest. Pain 42:35-42, 1990
20. Johansson F, Von Knorring L: A double-blind controlled study of a serotonin uptake inhibitor (Zimelidine) versus placebo in chronic pain patients. Pain 7:69-78, 1979
21. Lascelles RG: Atypical facial pain and depression. Br J Psychiatry 112:651-659, 1966
22. Loldrup D, Langemark M, Hansen HJ, et al: Clomipramine and mianserin in chronic idiopathic pain syndrome. A placebo controlled study. Psychopharmacology (Berl) 99:1-7, 1989
23. Fogelholm R, Murros K: Maprotiline in chronic tension headache: A double-blind crossover study. Headache 25:273-275, 1985
24. Gobel H, Hamouz V, Hansen C, et al: Chronic tension-type headache: Amitriptyline reduces clinical headache-duration and experimental pain sensitivity but does not alter pericranial muscle activity readings. Pain 59:241-249, 1994
25. Lance JW, Curran DA: Treatment of chronic tension headache. Lancet 1:1236-1239, 1964
26. Eisendrath SJ, Kodama KT: Fluoxetine management of chronic abdominal pain. Psychosomatics 33:227-229, 1992
27. Gessel AH: Electromyographic biofeedback and tricyclic anti-depressants in myofascial pain-dysfunction syndrome: Psychological predictors of outcome. J Am Dental Assoc 91:1048-1052, 1975
28. Okasha A, Ghaleb HA, Sadek A: A double-blind trial for the clinical management of psychogenic headache. Br J Psychiatry 122:181-182, 1973
29. Valdes M, Garcia L, Treserra J, et al: Psychogenic pain and depressive disorders: An empirical study. J Affec Disord 16:21-25, 1989
30. Pilowsky I, Barrow G: Predictors of outcome in the treatment of chronic "psychogenic" pain with amitriptyline and brief psychotherapy. Clin J Pain 8:358-362, 1992
31. Zitman FG, Linssen ACG, Edelbroek PM, et al: Does addition of low-dose Flupentixol enhance the analgetic effects of low-dose Amitriptyline in somatoform pain disorder? Pain 47:25-40, 1991
32. Feinmann C, Harris M, Cawley R: Psychogenic facial pain: Presentation and treatment. Br Med J 288:436-438, 1984
33. Eberhard G, Von Knorring L, Nilsson HL, et al: A double-blind randomized study of clomipramine versus maprotiline in patients with idiopathic pain syndromes. Neuropsychobiology 19:25-34, 1988
34. Onghena P, De Cuyper H, Van Houdenhove B, et al: Mianserin and chronic pain: A double-blind placebo-controlled process and outcome study. Acta Psychiatr Scand 88:198-204, 1993
35. Cohen J: Statistical power analysis for the behavioral sciences. New York, Academic 1969
36. McGorry PD, Mihalopoulos C, Henry L, et al: Spurious precision: Procedural validity of diagnostic assessment in psychotic disorders. Am J Psychiatry 152:220-223, 1995
37. Fishbain DA, Goldberg M: The misdiagnosis of conversion disorder in a psychiatric emergency service. Gen Hosp Psychiatry 13:177-181, 1991
38. Fishbain DA, Goldberg M: Response to "The misdiagnosis of conversion disorder in a psychiatric emergency service." Gen Hosp Psychiatry 14:145-148, 1992
39. Rosomoff HL, Fishbain DA, Goldberg M, et al: Physical findings in patients with chronic intractable benign pain of the neck and/or back. Pain 37:279-287, 1989
40. Fishbain DA: DSM-IV: Implications and issues for the pain clinician. APS Bull 5:6-18, 1995
41. Edelbroek PM, Linssen AC, Zitman FG, et al: Analgesic and antidepressive effects of low-dose amitriptyline in relation to its metabolism in patients with chronic pain. Clin Pharmacol Ther 39:156-62, 1986