Manuscript published in (2002) Ethical Human Sciences and Services, 4 (3), pp. 163-187.

The Benefits of Individual Psychotherapy for People Diagnosed with Schizophrenia: A Meta-Analytic Review

William H. Gottdiener	Nick Haslam
New School University	University of Melbourne
New York, NY	Melbourne, Australia

Abstract

A comprehensive meta-analytic review was undertaken to determine the efficacy of individual psychotherapy for people diagnosed with schizophrenia. Mean effect sizes were calculated for 37 studies conducted on 2642 patients. Possible moderator variables included (1) randomization, (2) source of data (between-groups and within-groups), (3) type of individual psychotherapy, (4) use of conjoint antipsychotic medication, (5) chronicity of the disorder, (6) treatment context, and (7) diagnostic criteria. Psychodynamic, cognitive-behavioral, and non-psychodynamic supportive therapies were all associated with improvement in functioning. Similar effect sizes were found between individual psychotherapy combined with antipsychotic medication and individual psychotherapy used without medication, for chronic schizophrenia compared with acute schizophrenia, and between studies that used randomization and those that did not. Larger effects existed for within-groups data compared with between-groups data, for outpatients compared with inpatients, and for studies conducted after the publication of DSM-III. Limitations of this review and suggestions for future research are discussed.

The Benefits of Individual Psychotherapy for People Diagnosed with Schizophrenia: A Meta-Analytic Review

For nearly 100 years, individual psychotherapy has been used in the treatment of schizophrenia, but its efficacy remains one of the most hotly debated subjects in the history of psychiatry. During that time, a large clinical literature has accrued from multiple theoretical perspectives. Although psychotherapists going back to Freud (1904) have questioned the utility of individual psychotherapy for people diagnosed with schizophrenia, most of the clinical literature describes creative interventions that psychotherapists and clinical researchers have developed to achieve positive results. Some of the more prominent psychotherapists and clinical researchers include psychoanalysts such as Arieti (1974), Boyer and Giovacchini (1980), Federn (1952), Fromm-Reichmann (1960), Lotterman (1996), Karon and VandenBos (1981), Robbins (1993), Searles (1965), and Sullivan (1962); cognitive-behavioral therapists such as Fowler, Garety, and Kuipers (1995), Haddock and Slade (1996), Kingdon and Turkington (1994), and Perris (1989), and clinical researchers such as Hogarty, G. E., Kornblith, S. J., Deborah, G., DiBarry, A. L., Cooley, S., Flesher, S., Reiss, D., Carter, M., & Ulrich, R. (1995) who have developed non-psychodynamically oriented supportive therapies.

In contrast to the clinical literature, research on the efficacy of individual psychotherapy for people diagnosed with schizophrenia has been marked by contradictory findings. Two frequently cited studies exemplify these mixed results (Karon & VandenBos, 1981; May, 1968). Both were randomized controlled clinical trials (RCT) that compared individual psychotherapy used with and without medication to standard psychiatric treatment in which antipsychotic medication was the primary intervention. May (1968) found that patients treated with supportive psychodynamic psychotherapy and conjoint antipsychotic medication, or those treated solely with medication, had significantly greater improvement rates than patients who received only supportive psychodynamic psychotherapy. In the Karon and VandenBos (1981) study, also conducted in the 1960s (see Karon & O’Grady, 1969, 1970; Karon & VandenBos 1970, 1972, 1975), the opposite results were obtained. Psychotherapists treated two groups of patients with individual exploratory psychodynamic psychotherapy. One group did not receive medication and the other did, but only in small doses that were ended within the first few weeks of treatment. These two treatment groups were compared to a third group of patients who received standard hospital care, with antipsychotic medication as the primary treatment. The researchers found that significantly more patients treated with individual psychotherapy (including those treated with psychotherapy and a brief course of medication) improved compared with those patients who received medication only.

Since these two landmark studies were conducted, many others have been conducted from different theoretical perspectives, the results of which have been summarized in a number of qualitative reviews (Fenton, 2000; Gomes-Schwartz, 1984; Heinrichs & Carpenter, 1981; Liberman, 1994; Mosher & Keith, 1980; Mueser & Berenbaum, 1990; Scott & Dixon, 1995). Most reviewers have concluded that little evidence exists to support the efficacy of psychodynamic psychotherapy, but evidence exists to support the efficacy of cognitive-behavioral and non-psychodynamic supportive therapies.

When reviewing a body of literature, the decision to use meta-analysis, or not, has potentially far-reaching implications. Two major policy papers that have established guidelines for the treatment of schizophrenia relied on qualitative reviews of the empirical literature. They came to somewhat different conclusions about the efficacy of individual psychotherapy for people diagnosed with schizophrenia, and made somewhat different recommendations about its use. These guidelines include the Patient Outcomes Research Team (PORT) report (Lehman & Steinwachs, 1998) and the guidelines developed by the American Psychiatric Association (APA) (1997). Both sets of guidelines recommend that individual psychotherapy using supportive interventions is an effective treatment for schizophrenia when combined with antipsychotic medications, but only the APA guidelines state that exploratory interventions (e.g., psychodynamic) might be useful for at least some patients. The PORT report, however, states that

Individual and group psychotherapies adhering to a psychodynamic model (defined as therapies that use interpretation of unconscious material and focus on transference and regression) should not be used in the treatment of persons with schizophrenia.

Rationale. The scientific data on this issue are quite limited. However, there is no evidence in support of the superiority of psychoanalytic therapy to other forms of therapy, and there is a consensus that psychotherapy that promotes regression and psychotic transference can be harmful to persons with schizophrenia. This risk, combined with the high cost and lack of evidence of any benefit, argues strongly against the use of psychoanalytic therapy, even in combination with effective pharmacotherapy. (Lehman & Steinwachs, 1998, pp. 7-8)

To date four reviews have employed meta-analysis (Cormac, Jones, and Campbell, 2002; Malmberg & Fenton, 2001; Mojtabai, Nicholson, & Carpenter, 1998; Smith, Glass, & Miller, 1980). Meta-analysis is the quantitative synthesis of similar empirical reports. Although like all statistical techniques, meta-analysis is not without its limitations (see LeLorier, Gregoire, Benhaddad, Lapierre, & Derderian, 1997; Wilson & Rachman, 1983), it is the most effective way to resolve controversial findings across a body of literature and has distinct advantages over qualitative literature reviews (see Hunter & Schmidt, 1990; Rosenthal, 1991).

Previous meta-analyses have come to different conclusions on this topic. Meta-analyses conducted by Smith et al. (1980) and Mojtabai et al. (1998) found that many schizophrenic patients improved who were treated with a variety of psychosocial treatments and antipsychotic medication. Of particular relevance to the current discussion, Mojtabai et al. (1998) found that individual psychotherapy was the most efficacious psychosocial treatment they reviewed. Malmberg and Fenton (2001) reviewed RCTs and concluded that the methods of existing research were too poor to draw definitive conclusions for or against the efficacy of individual psychodynamic psychotherapy. Cormac, Jones, and Campbell (2002) looked at cognitive-behavioral interventions for schizophrenia and concluded that there was no effect over standard treatment.

The Need for a New Meta-Analytic Review

Schizophrenia remains a significant public mental health problem, and any treatment that could help people who suffer from it ought to be known and available. Because meta-analysis is the most effective way to clarify conflicting results in a body of empirical literature, a comprehensive meta-analytic review could resolve the debate surrounding the efficacy of individual psychotherapy for people diagnosed with schizophrenia.

The four meta-analyses cited above have not resolved the debate, because of methodological differences. First, none of the meta-analyses have been comprehensive. Smith et al. (1980) and Mojtabai et al. (1998) investigated the efficacy of individual psychotherapy only when it was combined with medication. No review established the relative efficacy of individual psychotherapy when used without medication because none examined this issue. Three out of the four meta-analyses analyzed data only from between-groups designs (e.g., two-group experimental designs); only Smith et al. also included studies with within-group designs (e.g., quasi-experimental single group pretest-posttest designs and single group posttest only designs). However, Smith et al., did not examine these studies separately. Furthermore, Malmberg and Fenton (2001) and Cormac et al. (2002) only reviewed RCTs. Within-groups data is important because it provides information on treatment efficacy in absolute terms, whereas between-groups data provides information in relative terms. Second, findings from the Smith et al. meta-analysis are over twenty years old. The number of studies these authors reviewed is also unclear from their report and they combined together results of individual, group, and family treatments. Third, although relatively up to date, the meta-analysis by Mojtabai et al. included only 10 reports of individual psychotherapy when about forty have been published. Malmberg and Fenton included only three reports and Cormac et al. included 22 reports, but not all of the studies reviewed by Cormac et al. were of individual psychotherapy. Fourth, none of the reviews reported the relative efficacy of the three major forms of individual psychotherapy for schizophrenia—psychodynamic, cognitive-behavioral, and non-psychodynamic supportive.

A new meta-analysis is also warranted because some empirical evidence indicates that psychosocial treatments enhance the recovery of people diagnosed with schizophrenia. According to Hegarty, Baldessarini, Tohen, Waternaux, and Oepen (1994), and Warner (1994), approximately 50% of people diagnosed with schizophrenia and treated during the past 100 years have been shown to improve to at least the level of social recovery. A person in this category is able to be self-sufficient and live independently. Yet, as suggested by Warner's review, this level of recovery appears only to occur with those patients who have received psychosocial treatments. Prior to 1985 the overall success rate for somatic treatments (medication, electroconvulsive therapy, etc.) rose to just under 50%, but without a clear reason, it has since dropped to approximately 35% despite the more widespread use of antipsychotic medications (Hegarty et al., 1994). These findings, culled from hundreds of outcome studies and long-term follow-up reports, indicate that somatic only treatments are often insufficient by themselves and that psychosocial treatments are frequently necessary for the highest levels of improvement to occur. Clinical experience with people diagnosed with schizophrenia and empirical research also reveal that most patients with schizophrenia believe psychosocial treatments and individual psychotherapy in particular enhance their lives (see Coursey, Keller, & Farrell, 1995).

In order to determine the efficacy of individual psychotherapy for people diagnosed with schizophrenia we conducted a comprehensive meta-analytic review of the empirical literature, which included review of any study in which an effect size was calculable regardless of study design. (An effect size is 'the degree to which the phenomenon is present in the population,' or 'the degree to which the null hypothesis is false.'…the null hypothesis always means that the effect size is zero" (Cohen, 1988, pp. 9-10).) This procedure enabled us to determine the overall efficacy of individual psychotherapy for schizophrenia. It also enabled us to determine (a) the effects of random assignment of participants, (b) of source of the data (within-groups and between-groups), (c) the relative efficacy of psychodynamic, cognitive-behavioral, and non-psychodynamic supportive therapies, (d) the efficacy of conjoint antipsychotic medication, (e) the effects of the chronicity of the disorder, (f) the effects of treatment context, and (g) the potential effects of the narrowing of the schizophrenia diagnostic criteria that occurred with the publication of the Diagnostic and Statistical Manual of Mental Disorders-Third Edition (DSM-III); (APA, 1980).

Method

Inclusion Criteria

In order to conduct a comprehensive review, we used broad inclusion criteria to select reports. We chose to consider any outcome study that tested the efficacy of individual psychotherapy for people diagnosed with schizophrenia in which an effect size estimate was calculable. Therefore, we did not limit our review to studies that employed random assignment of participants and we reviewed reports that contained data from within-groups and between-groups designs. Inclusion of studies that did not use random assignment is justified because many clinical reports do not employ random assignment. Furthermore, non-random assignment does not necessarily invalidate a study's outcome. Although we expand on this point in the discussion section, it is worth noting that the effect of randomization is an empirical one and it has been pointed out recently that effect sizes and confidence intervals in studies that employ random assignment are often comparable to those that do not employ random assignment (Benson & Hartz, 2000; Concato, Shah, & Horowitz, 2000).

As noted, with the exception of Smith et al. (1980), past reviews of this topic have only considered the results obtained from between-groups data (i.e., experiments) and have not reviewed results obtained from within-groups data (pre-post), even if they were available. Such inclusion criteria can create a potentially serious bias in the review process. Review of between-groups data can only provide information about the results of experimental group vs. control group, or as in the case of most psychotherapy outcome studies, experimental treatment vs. alternative treatment. The results of such studies can only provide information about relative efficacy. They cannot provide information about the magnitude of change in the treatment group relative to where those patients began treatment or the proportion of patients successfully treated at outcome compared with those who were not successfully treated. Most reviewers exclude from review data or studies that were conducted using a within-groups design in the belief that lack of control or alternative treatment groups, poses too serious a threat to the study’s internal validity. This, however, is not necessarily the case because when the reliability of the dependent variable measure(s) is high, alternative explanations become less plausible (Hunter & Schmidt, 1990. This topic is more fully addressed in the discussion section). In most psychotherapy research, the reliability of dependent variable measures is high because standardized measures are used. Exclusion of information that comes from within-group study designs means that potentially valuable information is lost from consideration in the review.

To conduct this review we operationally defined individual psychotherapy broadly as any system of psychotherapy where one individual (the therapist) attempts to help another individual diagnosed with schizophrenia (the patient) obtain a relatively self-sufficient level of functioning by focusing on remedying cognitive and/or emotional difficulties in the patient without necessarily attempting to change directly the patient’s interpersonal or social-occupational functioning. We therefore excluded from review reports in which the primary interventions were case management, social skills training, or psychoeducation. We categorized the psychotherapies included for review as psychodynamic, cognitive-behavioral, or non-psychodynamic supportive therapies. All studies we reviewed were published in English; we located no pertinent studies published in other languages.

Literature Search

We obtained reports by perusing past review articles, the introductions to reports that were eventually used in this study, reference sections of books and articles, the tables of contents of various journals, in particular, Journal of Abnormal Psychology, Schizophrenia Bulletin, American Journal of Psychiatry, Archives of General Psychiatry, Journal of Consulting and Clinical Psychology, and Journal of Nervous and Mental Disease. We searched through on-line databases of Medline and Mental Health Abstracts (from 1967-September 1999) and Psych-Info (which includes dissertation abstracts) (from 1899 through September 1999). We used the following keyword phrases: psychotherapy and schizophrenia, and individual and psychotherapy and schizophrenia or schizophrenic and results. In addition, we ran other searches by varying the keywords and substituting individual and psychotherapy with words describing theoretical approaches such as psychodynamic psychotherapy, psychoanalytic psychotherapy, intensive psychotherapy, cognitive therapy, behavioral therapy, and cognitive-behavioral therapy. These words were selected on the basis of phrases that appear in the clinical and research literature on psychotherapy with people diagnosed with schizophrenia and those suggested by White (1994). These searches yielded 42 reports that satisfied our inclusion criteria. Five reports did not yield effect sizes and were consequently excluded, leaving 37 reports in our meta-analysis.

Fail-Safe N

Incomplete literature searches can bias meta-analytic results, especially if there are many unpublished reports with null results. However, null results are less likely to be published, which means such reports are difficult to obtain. If studies with null results were retrieved and included in a meta-analysis they might reduce the mean effect size of the meta-analysis to a negligible magnitude or cause the results to reverse direction. Rosenthal (1979) has developed a method of determining the number of additional reports that would have to be retrieved to state that no relationship exists between the independent and dependent variables being investigated in a meta-analysis. That number is called the Fail-Safe N, and to calculate it a person multiplies the total sample of retrieved studies by five and adds10. In the present case the Fail-Safe N would be 185; (5 x 37) + 10, which means that we would have to find 185 additional reports with null results to confidently say that no relationship exists between individual psychotherapy and improvement in schizophrenic symptoms. Hedges and Olkin (1985, p. 306) offer an alternative formula to calculate the Fail-Safe N: "k₀ = k( - r_c)/r_c_." The number of studies needed to reduce the mean effect size (in a fixed-effects model) to a negligible magnitude is k₀_;the mean effect size is , r_c is the proposed effect size that would be of negligible magnitude, which we set at r = .05, and k is the number of studies in the meta-analysis. In the current meta-analysis k₀ = 37(.31 - .05)/.05; k₀ = 192.4. Therefore, an additional 192 studies with null results would be required to reduce the current effect size to a negligible level of r = .05.

Meta-Analytic Procedures

Our meta-analytic methods are based on the methods of Rosenthal (1991), Hedges and Olkin, (1985), and Hunter and Schmidt’s (1990; 2000) approach. The latter authors focus on the correction of sampling error and the attenuation of effect sizes caused by study artifacts like measurement error.

Calculation Of Effect Sizes

We employed the Pearson Product Moment correlation r coefficient as the effect size index, and calculated most effect sizes with the aid of the DSTAT software (Johnson, 1995). The DSTAT software or probit transformations (Glass, McGaw, Smith, 1981) were used to calculate effect size estimates for categorical outcomes, and the most conservative effect size estimate was always chosen. These effect size estimates were then converted into (r) coefficients. We also used the program Comprehensive Meta-Analysis (Borenstein & Rothstein, 1999) to calculate the mean effect sizes, confidence intervals, and standard errors. Effect sizes were weighted by the inverse of their variance (Hedges & Olkin, 1985; Hunter & Schmidt, 1990) so that studies with relatively large samples would weigh more than studies with relatively small samples. We employed a random-effects model and a Fisher Z transformation in our calculation of the mean effect sizes. We used random-effects models because they provide the most accurate estimates of confidence intervals in meta-analysis and have, therefore, been recommended over fixed-effects models by the National Research Council (Hunter & Schmidt, 2000).

The advantage to reporting effect sizes in the form of the Pearson r is that it is an easily understood measure of effect because it describes the magnitude and direction of the relationship between two variables of interest (Hunter & Schmidt, 1990; Rosenthal, 1991). Furthermore, it can be interpreted with the Binomial Effect Size Display (BESD), which permits r to be a measure of the percentage of people who were likely to have shown improvement from a given treatment (Rosenthal & Rubin, 1982; Rosenthal, 1991). As an example, if an effect size of r = .50 were calculated, then the mean overall treatment effect size of r = .50 is equal to a 50% improvement rate. A 50% improvement rate means that the number of patients who are likely to benefit from treatment went from approximately 25% to 75%. This means that if 25% percent of patients were to improve without treatment, then 75% would be likely to improve with treatment. To calculate the change in improvement rate, outcome must be considered in dichotomous terms, even though in reality it might be continuous. Treatment outcome must be dichotomized into "improved" and "not improved" categories. The correlation coefficient is divided by two and .50 (which represents the 50% probability level) is added to the dividend. The formula is .50 +/- r/2. It is important to realize that the BESD is not an indication of the degree to which patients lose symptoms. Rather, it indicates how many people are likely to improve in functioning. It is possible that all patients at the end of treatment are still symptomatic, but that they are less symptomatic (i.e., they hallucinate less often, or have less cognitive slippage). The BESD works best with truly dichotomous outcomes (lived or died, improved or not improved). When outcomes occur on a continuum, like they do in schizophrenia, then the BESD reflects the number of people who improved. It will not reflect whether patients improved to a point where they can function without treatment.

One of the important problems that meta-analysts encounter is how to treat reports statistically insignificant results. We took a conservative approach and treated such results as if they had an effect size of zero (Cooper, 1989). If, on the other hand, we encountered reports that simply stated that a test was significant at the p <.05 or p <.01 level etc., we calculated the effect size by using the next highest probability coefficient (e.g., p < .05 was reported we calculated the effect size based on a probability equal to .049).

As stated above we included reports that contained within-groups and between-groups data in this review. This decision allowed us to use some reports more than once when they contained within-groups and between-groups data, which allowed us to capture the different ways in which individual psychotherapy for schizophrenia has been researched. However, each report contributed only one effect size to the meta-analysis. Some reports contained one outcome measure that yielded one effect size; whereas other reports contained multiple outcome measures and yielded multiple effect sizes. In such cases we averaged the multiple effect sizes. In this way, each study, regardless of how many effect sizes it yielded, contributed a single effect size per meta-analysis in order to maintain independence of effect sizes.

After correcting for sampling error, we sought to correct the mean effect sizes for attenuation due to other artifacts (Hunter & Schmidt, 1990; Schmidt & Hunter, 2001). Correction for attenuation due to study artifacts provides a more accurate estimate of the mean population effect size. We only found information on dependent variable measures in our sample of reports, and thus, controlled for attenuation due to measurement error in the dependent variables. Published reliability coefficients were gathered for ten outcome measures, which are listed in Table 1. These measures were chosen because their coefficients were provided in the reports we reviewed. The mean of these coefficients provides an estimate of the mean reliability of all of the outcome or dependent variable measures that had been employed in the various studies. We refer to this estimate as the artifact attenuation factor (Hunter & Schmidt, 1990) and calculated it as A = .85. To correct for measurement error in the dependent variable, we divided the mean effect size by the artifact attenuation factor to yield an estimate of the population effect size.

We decided to look for moderator variables only if the standard deviations of the mean corrected effect sizes were larger than zero (Schmidt & Hunter, 2001). This rule of thumb allows for detection of moderators that have theoretical or practical importance (Hedges & Olkin, 1985; Hunter & Schmidt, 1990) and does not rely on statistical methods for moderator detection, which tend to have low power (Schmidt & Hunter, 2001). The only moderators that provided enough information to investigate were (1) randomization, (2) source of data (between-groups or within-groups), (3) type of individual psychotherapy, (4) use of conjoint antipsychotic medication, (5) chronicity of the disorder, (6) treatment context, and (7) diagnostic criteria.

Coding Reliability

The first author (William H. Gottdiener) coded all of the reports, and a clinical psychology graduate student coded ten randomly chosen reports as a reliability check. The variables included were study design, the existence of a control group, sample size, effect size, and treatment type. We chose these variables because the results of the meta-analyses revolved around them. The kappa coefficient (k), the contingency coefficient (c), and the Pearson Product Moment Correlation coefficient (r) were used as measures of reliability. Table 2 shows that each variable was coded with a high degree of reliability.

Results

Our reason for conducting this meta-analysis was to learn if individual psychotherapy actually benefits people diagnosed with schizophrenia. We answered this question by reviewing 37 studies that were published between 1954 and 1999. We calculated 232 effect sizes. These reports are based on the treatment of 2642 patients with a mean age of 31.1 years. Treatment lasted for an average of 20.2 months with a mean of 1.4 sessions per week.

In order to determine the overall efficacy of individual psychotherapy for people diagnosed with schizophrenia we averaged all of the effect sizes from all 37 studies. The results in Table 3 show that individual psychotherapy was associated with improvement in overall functioning. The table presents the grand mean effect size, the corrected effect size (this is the grand mean effect size divided by the artifact attenuation factor), and improvement rates determined by the BESD. The grand mean effect size was r = .31 (95% CI + .22 to .41). The corrected effect size was r = .36. The BESD results showed that improvement rate increased from 35% to 66%, which means that 66% of patients are better off after treatment compared with 35% before treatment. What does this mean? Rosenthal, Rosnow, and Rubin (2000) write: " r of .32 (or a r² of .10) will amount to a difference between rates of improvement of 34% and 66% if half the population received psychotherapy and half did not, and if half the population improved and half did not" (p. 17). In this meta-analysis the BESD means that 65% percent of the population that received psychotherapy improved compared with only 34% of the population that did not receive psychotherapy.

The grand mean effect size shows the general efficacy of individual psychotherapy for people diagnosed with schizophrenia. Treatment efficacy was, however, moderated by a number of factors. As noted above, we examined only variables for which there was abundant information throughout the database. We present these below . The distribution of moderator variables in our sample of 37 reports is presented in Table 4.

The Effects of Random Assignment

Studies that employ random assignment are thought to minimize threats to internal validity and increase the accuracy of effect size estimates compared with studies that do not employ random assignment (Kazdin, 1998; Shadish & Ragsdale, 1996). Shadish and Ragsdale found that psychotherapy outcome studies that employed random assignment reported larger effect size estimates than those that did not employ random assignment, which led them to conclude that failure to randomize leads to potentially less accurate effect size estimates. Because we reviewed studies that used random assignment and studies that did not we were able to see if the studies differed.

Nineteen studies employed random assignment. The mean effect size for these studies was r = .31 (95% CI + .17 to .43). Another 18 studies either did not employ random assignment or did not report if it was used. The mean effect size for these studies was r = .34 (95% CI + .20 to .46). Thus, in this database there was essentially no difference in outcome between studies that employed random assignment and those that did not. This suggests that failure for primary studies to employ random assignment did not pose meaningful threats to the internal validity of this meta-analysis.

Source of Data

It has been stated that within-groups data tends to inflate effect sizes compared with between-groups data (Morris & DeShon, 2002). The advantage to combining both types of data in one meta-analysis would be for the between-groups data to correct for effect size inflation that could be caused by within-groups data. Whether or not within-groups data inflates effect sizes is open to debate. Rather than inflating effect sizes, within-groups data has been thought to simply address different questions than between-groups data. Within-groups data examines change in one group over time whereas between-groups data examines change in one group relative to another group over time. Because of this it has been questioned whether or not it is appropriate to combine within-groups and between-groups data. Morris and DeShon (2002) suggest that it is appropriate to combine such data when effect size estimates are similar, or when between-groups and within-groups data are conceptually addressing the same fundamental issues. We present both forms of data combined, as shown in Table 3, and separately as follows. For between-groups data, the mean effect size was r = .10 (95% CI + .02 to .18). We found that for within-groups data, the mean effect size was r = .58 (95% CI + .43 to .70).

The between-groups results suggest that individual psychotherapy provides a similar advantage to other treatments. In this meta-analysis there were seven comparisons in which the comparison treatment was not exclusively antipsychotic medication. The comparison consisted of psychotherapy plus medication vs. another form of psychosocial treatment plus medication. The specific results of the between-groups comparisons are as follows: (a) individual psychotherapy alone compared with antipsychotic medication was r = -.01 (95% CI + -.21 to .19); (b) individual psychotherapy plus antipsychotic medication compared with antipsychotic medication was r = .19 (95% CI + .07 to .31); (c) individual psychotherapy plus medication compared with other psychosocial treatments plus medication was r = .08 (95% CI + -.16 to .31). As can be seen, the only comparison that showed a distinct advantage over another treatment was the comparison of individual psychotherapy plus medication compared with antipsychotic medication.

The within-groups data from above suggests that there is a dramatic change that occurs during treatment and that most patients are much better off than they were before treatment. The findings show that the efficacy of treatment between individual psychotherapy and alternative treatments (e.g., antipsychotic medication) was approximately equal. It is, therefore, plausible to think that a similar amount of change occurs from pre-test to posttest in patients treated with individual psychotherapy and with antipsychotic medication. Because of this we thought it was justifiable to combine all of the effects irrespective of whether the data came from a within-groups or between-groups design. Although both types of data address different issues, they also address the same fundamental issue of whether or not individual psychotherapy is an effective treatment for schizophrenia.

Type of Individual Psychotherapy

As noted above, there have been critiques leveled against the use of psychodynamic psychotherapy for schizophrenia (Drake & Sederer, 1986; Lehman & Steinwachs, 1998; Mueser & Berenbaum, 1990). Although Cormac et al. (2002) found equivocal results for cognitive-behavioral therapies, there has also been much praise for their use as well as the use of non-psychodynamic supportive therapies for schizophrenia (Bustillo, Lauriello, Horan, & Keith, 2001; Fowler et al., 1995; Liberman, 1994).

We found similar effect sizes for each form of treatment. The grand mean effect size for psychodynamic psychotherapy was r = .33 (95% CI + .21 to .44). The corrected effect size was r = .39. The BESD results showed that improvement rate increased from 34% to 67%.

The grand mean effect size for cognitive-behavioral therapy was r = .35 (95% CI + .08 to .58). The corrected effect size was r = .41. The BESD results showed that improvement rate increased from 33% to 68%.

The grand mean effect of non-psychodynamic supportive therapy was r = .23 (95% CI + .00 to .44). The corrected effect size was r = .27. The BESD results showed that improvement rate increased from 38% to 62%.

These results contrast with some of the past literature reviews mentioned above and show that all three treatments are beneficial. Furthermore, psychodynamic and cognitive-behavioral therapies produce a similar degree of therapeutic benefit, which is more than that produced by non-psychodynamic supportive therapies.

Use of Conjoint Medication

There is a long-standing controversy about the use of conjoint medication in the psychotherapy of people with schizophrenia. For many years psychotherapists were reluctant to employ antipsychotic medication as an adjunct to psychotherapy because they thought it would disturb the therapeutic process. However, since at least the 1960s most therapists that treat people with schizophrenia have used antipsychotic medication in conjunction with psychotherapy. Most therapists think it is indispensable. However, between 40% and 75% of patients do not take their medication (Perkins, 1999) and there are many for whom medications fail to work (see above Hegarty et al., 1994). For these patients and for therapists who choose to offer treatment with little or no adjunctive medication, it would be important to know if such treatments work.

When antipsychotic medication was used with individual psychotherapy the mean effect size was r = .31 (95% CI + .19 to .42). When antipsychotic medications were not administered with individual psychotherapy the mean effect size was also r = .31 (95% CI + .12 to .48). The corrected effect size and BESD results for psychotherapy with medication and without medication was the same. The corrected effect size was r = .36 and the BESD results showed that improvement rate increased from 35% to 66%.

Chronicity of the Disorder

People diagnosed with acute schizophrenia have long been thought to have better prognoses than people diagnosed with chronic schizophrenia. For people diagnosed with acute schizophrenia the mean effect size was r = .37 (95% CI + .10 to .59). The corrected effect size was r = .44. The BESD results showed that improvement rate increased from 32% to 69%. Some studies included people diagnosed with both acute and chronic schizophrenia and in others the diagnosis was between acute and chronic. For people in this group of studies the mean effect size was r = .12 (95% CI + .01 to .22). The corrected effect size was r = .14. The BESD results showed that improvement rate increased from 44% to 56%. For those people diagnosed with chronic schizophrenia the mean effect size was r = .36 (95% CI + .21 to .49). The corrected effect size was r = .42. The BESD results showed that improvement rate increased from 32% to 68%. Finally, there were studies in which the chronicity was not reported. The mean effect size of those studies was r = .26 (95% CI + .10 to .41). The corrected effect size was r = .31. The BESD results showed that improvement rate increased from 37% to 63%.

The results were similar for people diagnosed with acute and chronic schizophrenia, but it is difficult to know why. It could be a function of the way we defined acute and chronic schizophrenia. We considered schizophrenia to be chronic when it lasted at least two years. This was an arbitrary distinction on our part and others researchers and clinicians might have defined chronicity differently.

Treatment Context

Clinical lore suggests that treatment context has a strong influence on treatment outcome. We could only examine this factor by looking at which studies were conducted on inpatients or outpatients. For inpatients the mean effect size was r = .29 (95% CI + .14 to .42). The corrected effect size was r = .34. The BESD results showed that improvement rate increased from 35% to 65%. For those that were treated in both inpatient and outpatient facilities during the study, the mean effect size was r = .30 (95% CI + .03 to .53). The corrected effect size was r = .35. The BESD results showed that improvement rate increased from 35% to 65%. Finally, for outpatients the mean effect size was r = .37 (95% CI + .19 to .52). The corrected effect size was r = .44. The BESD results showed that improvement rate increased from 31% to 69%. The results show that outpatients improved at a higher rate than inpatients, which supports the prediction made based on clinical lore.

Diagnostic Criteria

All of the participants in each treatment were diagnosed with schizophrenia, but reports vary according to the specificity of the diagnostic information provided. Table 4 shows the diagnostic criteria that were used for each of the reports included in this meta-analysis. One study used DSM-II criteria (APA, 1968), 4 used DSM-III (APA, 1980), 3 used DSM-III-R (APA, 1987), 2 employed Research Diagnostic Criteria (RDC, Spitzer, Endicott, & Robins, 1975), 4 used multiple criteria including either a DSM criteria or RDC criteria, another 4 used other diagnostic criteria, 10 used interviewer agreement, and 9 did not report the method they used.

It is widely known that the diagnostic criteria for schizophrenia were broader before DSM-III (APA, 1980) was published. Before the publication of DSM-III, people with diagnoses of schizoaffective disorder and schizophreniform disorder were often diagnosed with schizophrenia. People with these latter two disorders tend to have better prognoses than people diagnosed with schizophrenia. Given this, patients in pre-1980 studies might have improved more than those in later studies, which would be reflected in larger effect sizes for those studies. We examined the effect sizes for studies conducted prior to DSM-III's publication even if those studies were published after it. The mean effect size for studies conducted after the publication of DSM-III was r = .39 (95% CI + .19 to .56). Whereas the mean effect size for studies conducted before its publication was r = .28 (95% CI + .16 to .38). It is difficult to know why results published before DSM-III would produce smaller effects, because this is contrary to what would be predicted.

This meta-analysis addresses a variety of clinically relevant questions. Most importantly, the meta-analytic findings suggest that individual psychotherapy, whether combined with antipsychotic medication or not, is an effective treatment for schizophrenia. Individual psychotherapy without medication appears to have a roughly similar therapeutic efficacy to antipsychotic medication and each treatment seems to have a roughly similar incremental benefit when combined with the other. Studies that employed random assignment did not have larger effects than those that did not. People diagnosed with acute and chronic schizophrenia obtained similar success rates. Outpatient treatment was associated with larger incremental benefit than inpatient treatment. Effect sizes were noticeably larger for studies published after the publication of DSM-III than before it.

Discussion

The goals of this meta-analytic review were to determine the general efficacy of individual psychotherapy for people diagnosed with schizophrenia and the effects of the following potential moderator variables: (1) randomization, (2) source of data (between-groups or within-groups), (3) type of individual psychotherapy, (4) use of conjoint antipsychotic medication, (5) chronicity of the disorder, (6) treatment context, and (7) diagnostic criteria. To accomplish this we retrieved 37 reports published between 1954 and 1999 that met our inclusion criteria and as noted above, none were dissertations.

Our findings indicate that individual psychotherapy is associated with improved functioning in the majority of patients diagnosed with schizophrenia who receive it. We found that all forms of individual psychotherapy (psychodynamic, cognitive-behavioral, and non-psychodynamic supportive) were associated with an improvement in functioning for people diagnosed with schizophrenia, but that the largest improvement rates were associated with psychodynamic and cognitive-behavioral therapies.

It is surprising that the proportion of patients that were likely to improve without conjoint medication, is similar to the proportion of patients that were likely to improve with a combination of individual psychotherapy and antipsychotic medication. This finding is contrary to most therapists’ clinical expectations. The finding that individual psychotherapy can be effective without medication is not new (see Karon & VandenBos, 1981). However, it is important because it suggests that individual psychotherapy alone might be a viable treatment option for some patients who do not improve from treatment with antipsychotic medications, for some patients who refuse to take medications, or for patients who are treated by therapists that choose to use little or no adjunctive medication.

Effect sizes from within-groups data were considerably larger than those from between-groups data. There were slightly larger effects for people diagnosed with chronic schizophrenia than for those diagnosed with acute schizophrenia. Psychotherapy with inpatients was, however, associated with lower effect sizes than with outpatients. In addition, we found that effect sizes were smaller for reports published before than after the publication of DSM-III.

The results of our meta-analytic review might not at first appear to give a coherent account of the efficacy of individual psychotherapy for schizophrenic patients. But, closer inspection reveals one clear trend: Individual psychotherapy is associated with improvement in functioning in people diagnosed with schizophrenia in the overall meta-analysis and in each moderator analysis of the data.

These findings are not an aberration. They are consistent with those of the clinical literature, the findings of Smith et al. (1980) and of Mojtabai et al. (1998), and the practice guidelines of the APA (1997). The findings of our review, however, contradict those of Malmberg and Fenton (2001) and Cormac et al. (2002), of most previous qualitative reviews of the literature, and the PORT guidelines (Lehman & Steinwachs, 1998). The findings are especially contrary to suggestions that individual psychodynamic psychotherapy is contraindicated for people diagnosed with schizophrenia.

Limitations

Although this meta-analytic review is the broadest one conducted to date, it has certain limitations, which need to be addressed by future meta-analyses. First, there were a small number of studies to review. At the time of this writing, fewer than 60 published empirical reports existed on the efficacy of individual psychotherapy for people diagnosed with schizophrenia. A larger sample of studies will enable more accurate estimates of effect sizes, and analysis of more potentially important moderator variables, such as estimation of changes in effect sizes over time during treatment and through several follow-up periods. The small number of extant studies also limited the amount of information available to conduct analyses of potential clinically important moderator variables such as therapist experience or training. The small number of reports also limited the amount of information available to correct for attenuation due to study artifacts such as, the unreliability of the independent variable of individual psychotherapy. However, we would probably need a database of several hundred reports to minimize these problems.

Second, about half of the studies we reviewed did not assign participants randomly. The purpose of random assignment is to minimize potential threats to a study's internal validity (Kazdin, 1998). Random assignment specifically reduces selection biases, which are

systematic differences in groups on the basis of the selection or assignment of subjects. Obviously, the effects of an independent variable among groups can be unambiguously inferred only if there is some assurance that groups do not systematically differ before the independent variable was applied (Kazdin, 1998, p. 20).

For random assignment to be useful the researcher must presume that there is no interaction between the participants and the independent variable (Hunter & Schmidt, 1990).

As already noted, we chose to include all studies that would yield an effect size because we believe that throwing away data is treating it as if it does not exist. Furthermore, it is simply not true that failure to randomize is equal to absolute invalidation of a study's results and, as we showed, the effects of randomization can be tested empirically. Most supporters of evidence based medicine believe that the RCT is the best method to determine causality and provides the best estimates of population effect sizes (see Hunter & Schmidt, 1990). Observational or naturalistic designs are believed to be inferior because they are thought to inflate effect size estimates.

These long-held assumptions about RCTs have been challenged recently (Benson & Hartz, 2000; Concato, Shah, & Horowitz, 2000; Pawson & Tilley, 1997; Shrout, 1998). In two separate meta-analyses of medical treatments, no differences in effect sizes and confidence intervals were found to exist between RCTs and observational studies (Benson & Hartz, 2000; Concato et al., 2000). This means that in many observational studies there is neither selection bias nor an interaction between the participants and the independent variable. Statistical techniques like structural equation modeling can also be used to help make strong causal inferences in observational and naturalistic studies (Shrout, 1998). Random assignment is, therefore, not necessary to draw causal inferences or to obtain accurate estimates of population effect sizes. It might be the clearest way to do so, but not the only way to do so.

In addition, we agree with Hunter and Schmidt (1990) that a notorious problem to deal with in social science research is low power caused by sampling error. Most studies are underpowered and the consequences often result in conflicting research findings, which disappear once sampling error is corrected via meta-analysis (Hunter & Schmidt, 1990). Therefore, we believed it inappropriate to exclude studies not employing randomization. Despite the findings that randomization does not necessarily invalidate results, we believe that randomization is still important to use whenever it is feasible to do so.

A third limitation to this meta-analysis is that we did not attempt to retrieve unpublished data. It is likely that a small number of unpublished treatment studies exist. We did not identify any through Dissertation Abstracts. As more dissertations are conducted, more of them will examine the efficacy of individual psychotherapy for schizophrenia and will then be available for review. We also did not write to experts in the field for unpublished studies, and we did not search all possible electronic databases (see Malmberg & Fenton, 2001 for an example of a very broad search). However, we searched the sources likely to bring the highest yield. We are confident that the limitations of our search strategy did not deleteriously affect our findings because of the results of the checks we employed (e.g., Fail-Safe N).

Fourth, we did not report outcome variables, the effect sizes for each outcome variable, nor did we report effect sizes separately for dichotomous and continuous data. Fifteen effect sizes were reported as dichotomous outcomes, the rest were from continuous data. In addition, we calculated and reported all dichotomous outcomes using a Pearson Product Moment Correlation coefficient, rather than an odds ratio or relative risk. We did this to maintain uniformity in outcomes across dichotomous and continuous effect sizes. Because of the relatively small sample of existing studies, we thought that obtaining a picture of the general relationship between treatment and outcome would be more informative than reporting specific outcomes by measurement or symptom. We realize, however, that an impressionistic picture like the one we have painted limits the ability to make finer-grained critiques of this meta-analysis.

Fifth, we did not perform intent-to-treat analysis. Such analysis views treatment drop-outs as treatment failures, which is not necessarily a correct interpretation. Patients drop out of treatment for many reasons and rarely is it known why. However, such analysis might be worthwhile when more studies exist and if they have tracked reasons for attrition.

Sixth, we were unable to conduct time-series analysis and estimate how much change occurred over time at different time periods, such as at six months or two years. All effect sizes were averaged regardless of how long treatment lasted, whether the effect was from follow-up data or from the immediate end of treatment. We did this because reporting effects for different time periods would produce the spurious impression of a treatment by time interaction that might not exist or that might differ among treatment types. Rather than create a false impression that a treatment by time interaction exists--something we would be unable to test--we thought it best to report one single treatment effect.

Seventh, we were unable to use the Jacobson-Truax statistic (Jacobson & Truax, 1991) to determine how close toward being in a non-clinical population patients were at the end of their treatments. We could not use the statistic because it requires the same outcome measures to be used in all studies.

Eighth, our inclusion of within-group, pretest-posttest effect sizes from change scores can face interpretative problems, given that change scores have the potential to be influenced strongly by regression to the mean (Campbell & Kenny, 1999). Despite these potential difficulties, we thought that inclusion of effect sizes from within-group designs was essential to deriving an overall sense of how much change could possibly occur from pretest to posttest.

Future Directions

Additional studies would increase our understanding of the benefits of individual psychotherapy for people diagnosed with schizophrenia. They would allow reviewers to assess which therapeutic interventions benefit which patients and when. A number of authors have pointed out that the goals of treatment, and the interventions used will inevitably vary as a function of the type of schizophrenia being treated, the duration of the disorder, the personality of the patient, and the context of treatment (Eissler, 1951; Fenton, 2000; McGlashan & Keats, 1989). In addition, a number of therapists have developed psychotherapeutic approaches divided into stages that represent steps of progression that patients pass through as they recover (e.g., Arieti, 1974; Pao, 1979). Much of the clinical literature also suggests that the most important issue in the use of individual psychotherapy for people diagnosed with schizophrenia is for therapists to maintain flexibility, and to use supportive and insight-oriented interventions as needed (see Fenton, 2000). Although none of the reports we reviewed reported outcomes specifically as a function of the type of interventions used, it is possible to infer from our review that both supportive and insight-oriented interventions are beneficial. Cognitive-behavioral and non-psychodynamic supportive therapies primarily consist of supportive interventions (see Hogarty et al., 1995; Kingdon & Turkington, 1994), but not exclusively. Some cognitive-behavioral treatments extensively use insight-oriented interventions (see Perris, 1989). Although psychodynamic therapies are generally thought to consist of insight-oriented interventions (see Robbins, 1993), they traditionally consist of supportive interventions in the early stages of treatment with severely disturbed patients, and for some patients supportive interventions will be the core of their treatment (see Rockland, 1989). What would be most useful to draw from future research is to learn which patients could benefit from which types of interventions at which stage of treatment. It would also be important to be able to employ the Jacobson-Truax (Jacobson & Truax, 1991) statistic to find out how much clinically meaningful change actually takes place for which patients under which conditions.

A final aspect that future research should address is the cost-benefit ratio of individual psychotherapy. One argument against using individual psychotherapy, even by those who believe in its efficacy, is that its expense prohibits its use on a large scale: It is simply too costly to offer individual psychotherapy to the majority of patients compared with medication management or other psychosocial treatments, such as group therapy, so the argument goes. Some would also argue that because individual psychotherapy does not seem to be considerably more efficacious than medication, it is simply not worth the added expense. Thus, the costs are presumed to outweigh the benefits.

There is some evidence, however, that suggests that the overall cost of treating severely mentally ill patients decreases significantly with the use of psychotherapy (Gabbard, Lazar, Hornberger, & Spiegel, 1997; Karon & VandenBos, 1981). Compared with patients that are not in psychotherapy, patients in psychotherapy tend to use less inpatient treatment; they function at a higher level, and their psychiatric problems become less likely to interfere with obtaining and maintaining employment (Gabbard et al., 1997). The current meta-analytic review shows that dramatic improvement rates are associated with the use of individual psychotherapy, which suggests that over the course of a person’s lifetime, the use of individual psychotherapy could actually help to reduce treatment costs. However, this is an empirical issue that requires further study.

Related to the issue of cost-effectiveness of individual psychotherapy is the related issue of costs due to the side-effects or adverse reactions to antipsychotic medications. Seventy-five percent of patients diagnosed with schizophrenia stop taking their antipsychotic medication within two years, and problematic side-effects are the primary reason (Perkins, 1999). Antipsychotic medication is the primary treatment for schizophrenia, but if the compliance rate is so poor, then many people will ultimately forgo treatment. It is thus plausible to speculate that the adverse effects of antipsychotic medications might contribute to the long-term high costs of treatment, lower capacity to be employed, and diminished psychosocial functioning. This hypothesis also needs to be empirically investigated.

Final Remarks

The findings of this meta-analysis clearly show that there is a relationship between the use of individual psychotherapy and improvement in overall functioning in people diagnosed with schizophrenia, when used with medication and without medications. Furthermore, all major forms of individual psychotherapy appear to be effective. However, primarily because of limitations of the available data on this topic, a more fine-grained understanding of the role of individual psychotherapy for the treatment of schizophrenia remains to be elucidated by future primary research and by future meta-analyses. It is clear that much remains to be learned about the utility of individual psychotherapy for the treatment of schizophrenia, and we have mentioned some of the important topics that need further exploration. Nevertheless, this review shows that the pessimism and skepticism that has long surrounded the utility of individual psychotherapy for people diagnosed with schizophrenia is unwarranted. It is now time to roll up the sleeves and learn more about how this treatment works, and how it can be made as effective and as available as possible.

References

References marked with an asterisk indicate studies included in the meta-analysis.

*Alanen, Y. O., Räkköläinen, V., Rasimus, R., Laakso, J., & Kaljonen, A. (1985). Psychotherapeutically oriented treatment of schizophrenia: Results of a 5-year-follow-up. Acta Psychiatrica Scandinavica, 71(Supplementum No. 319), 31-49.

*Alanen, Y. O., Ugelstad, E., Armelius, B.-Å., Lehtinen, K., Rosenbaum, B., & Sjöström, R. (Eds.). (1994). Early treatment for schizophrenic patients. Oslo: Scandinavian University Press.

Arieti, S. (1974). Interpretation of schizophrenia. New York: Jason Aronson.

American Psychiatric Association (1997). Practice guideline for the treatment of patients with schizophrenia (1^st ed.). Washington, DC: American Psychiatric Association.

American Psychiatric Association (1968). Diagnostic and statistical manual of mental disorders (DSM-II) (2^nd ed.). Washington, DC: American Psychiatric Association.

American Psychiatric Association (1980). Diagnostic and statistical manual of mental disorders (DSM-III) (3^rd ed.). Washington, DC: American Psychiatric Association.

American Psychiatric Association (1987). Diagnostic and statistical manual of mental disorders (DSM-III-R) (3^rd ed. revised). Washington, DC: American Psychiatric Association.

Benson, K., & Hartz, A. J. (2000). A comparison of observational studies and randomized, controlled trials. The New England Journal of Medicine, 342 (25), 1878-1886.

*Bookhammer, R. S., Meyers, R. W., Schober, C. C., & Piotrowski, Z. A. (1966). A five-year clinical follow-up study of schizophrenics treated by Rosen’s ‘Direct Analysis’ compared with controls. American Journal of Psychiatry, 123(5), 602-604.

Borenstein, M., & Rothstein, H. (1999). Comprehensive Meta-Analysis: A computer program for research synthesis. (Version 1.0.12) [Computer Software]. Englewood, NJ: Biostat.

Boyer, L. B., & Giovacchini, P. L. (1980). Psychoanalytic treatment of schizophrenic, borderline, and characterological disorders (2^nd ed.). New York: Jason Aronson.

*Buchkremer, G., Klingberg, S., Holle, R., Schulze, M. H., & Hornung, W. P. (1997). Psychoeducational psychotherapy for schizophrenic patients and their key relatives or care-givers: Results of a 2-year follow-up. Acta Psychiatrica Scandinavica, 96, 483-491.

*Bullard, J., Dexter M., Hoffman, B. R., & Havens, L. L. (1960). The relative value of tranquilizing drugs and social and psychological therapies in chronic schizophrenia. Psychiatric Quarterly, 34, 293-306.

Bustillo, J., Lauriello, J., Horan, W., Keith, S. (2001). The psychosocial treatment of schizophrenia: An update. American Journal of Psychiatry, 158 (2), 163-175.

Campbell, D. T., & Kenny, D. A. (1999). A primer on regression artifacts. New York: Guilford.

*Carpenter, W. T., McGlashan, T. H., & Strauss, J. S. (1977). The treatment of acute schizophrenia without drugs: An investigation of some current assumptions. American Journal of Psychiatry, 134(1), 14-20.

Cohen, J. (1988). Statistical power analysis for the behavioral sciences (Second ed.). Hillsdale, NJ: Lawrence Erlbaum Associates.

Cooper, H. M. (1989). Integrating research: A guide for literature reviews (2^nd ed.). London: Sage.

Concato, J., Shah, N., Horwitz, R. I. (2000). Randomized, controlled trials, observational studies, and the hierarchy of research designs. The New England Journal of Medicine, 342 (25), 1887-1892.

Cormac, I., Jones, C., Campbell, C. (2002). Cognitive behaviour therapy for schizophrenia (Cochrane Review). Cochrane Database of Systematic Reviews (1) [Electronic database]. The Cochrane Library.

*Coursey, R. D., Keller, A. B., & Farrell, E. W. (1995). Individual psychotherapy and persons with serious mental illness: The clients’ perspective. Schizophrenia Bulletin, 21(2), 283-301.

Drake, R. E., & Sederer, L. I. (1986). The adverse effects of intensive treatment of chronic schizophrenia. Comprehensive Psychiatry, 27(4), 313-326.

Eissler, K. R. (1951). Remarks on the psychoanalysis of schizophrenia. International Journal of Psychoanalysis, 32, 139-156.

*Falloon, I. R. H., Boyd, J. L., McGill, C. W., Williamson, M., Razani, J., Moss, H. B., Gilderman, A. M., & Simpson, G. M. (1985). Family management in the prevention of morbidity of schizophrenia. Archives of General Psychiatry, 42, 887-896.

Federn, P. (1952). Ego psychology and the psychoses. New York: Basic Books.

Fenton, W. S. (2000). Evolving perspectives on individual psychotherapy for schizophrenia. Schizophrenia Bulletin, 26(1), 47-72.

Fowler, D., Garety, P., & Kuipers, E. (1995). Cognitive behaviour therapy for psychosis: Theory and practice. New York: Wiley.

*Fowler, D., & Morley, S. (1989). The cognitive-behavioural treatment of hallucinations and delusions: A preliminary study. Behavioural Psychopathology, 17, 267-282.

Freud, S. (1904). On psychotherapy (Vol. 7, pp. 255-268). London: Hogarth.

Fromm-Reichmann, F. (1960). Principles of intensive psychotherapy. Chicago: University of Chicago Press.

Gabbard, G. O., Lazar, S. G., Hornberger, J ., & Spiegel, D. (1997). The economic impact of psychotherapy: A review. American Journal of Psychiatry, 154(2), 147-155.

*Garety, P. A., Kuipers, L., Fowler, D., Chamberlain, F., & Dunn, G. (1994). Cognitive behavioural therapy for drug-resistant psychosis. British Journal of Medical Psychology, 67, 259-271.

*Glass, L. L., Katz, H. M., Schnitzer, R. D., Knapp, P. H., Frank, A. F., & Gunderson, J. G. (1989). Psychotherapy of schizophrenia: An empirical investigation of the relationship of process to outcome. American Journal of Psychiatry, 146(5), 603-608.

Glass, G. V., McGaw, B., & Smith, M. L. (1981). Meta-Analysis in social research. Beverly Hills: Sage.

Gomes-Schwartz, B. (1984). Individual psychotherapy of schizophrenia. In A. Bellack (Ed.), Schizophrenia: Treatment, management and rehabilitation (pp. 307-335). New York: Grune & Stratton.

*Gottlieb, J. S., & Huston, P. E. (1951). Treatment of schizophrenia: A comparison of three methods: Brief psychotherapy, insulin coma and electric shock. Journal of Nervous and Mental Disease, 113, 237-246.

*Grinspoon, L., Ewalt, J. R., & Shader, R. (1968). Psychotherapy and pharmacotherapy in chronic schizophrenia. American Journal of Psychiatry, 124(12), 1645-1652.

*Gunderson, J. G., Frank, A. F., Katz, H. M., Vannicelli, M. L., Frosch, J. P., & Knapp, P. H. (1984). Effects of psychotherapy in schizophrenia II: Comparative outcome of two forms of treatment. Schizophrenia Bulletin, 10 (4), 564-598.

Haddock, G., & Slade, P. D. (Eds.). (1996). Cognitive-behavioural interventions with psychotic disorders. London: Routledge.

*Hamill, W. T., & Fontana, A. F. (1975). The immediate effects of Chlorpromazine in newly admitted schizophrenic patients. American Journal of Psychiatry, 132(10), 1023-1026.

Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta-analysis. New York: Academic Press.

Hegarty, J. D., Baldessarini, R. J., Tohen, M., Waternaux, C., & Oepen, G. (1994). One hundred years of schizophrenia: A meta-analysis of the outcome literature. American Journal of Psychiatry, 151(10), 1409-1416.

Heinrichs, D. W., & Carpenter, W. T. (1981). The efficacy of individual psychotherapy: A perspective and review emphasizing controlled outcome studies. In S. Arieti & H. K. Brodie (Eds.), American Handbook of Psychiatry (pp. 586-613). New York: Basic Books.

Hogarty, G. E., Kornblith, S. J., Deborah, G., DiBarry, A. L., Cooley, S., Flesher, S., Reiss, D., Carter, M., & Ulrich, R. (1995). Personal therapy: A disorder-relevant psychotherapy for schizophrenia. Schizophrenia Bulletin, 21(3), 379-393.

*Hogarty, G. E., Goldberg, S. C., Schooler, N. R., Ulrich, R. F., & Group, T. C. S. (1974). Drug and sociotherapy in the aftercare of schizophrenic patients: II. Two-year relapse rates. Archives of General Psychiatry, 31, 603-608.

*Hogarty, G. E., Greenwald, D., Ulrich, R. F., Kornblith, S. J., DiBarry, A. L., Cooley, S., Carter, M., & Flesher, S. (1997). Three-year trials of personal therapy among schizophrenic patients living with or independent of family, II: Effects on adjustment of patients. American Journal of Psychiatry, 154(11), 1514-1524.

*Hogarty, G. E., Kornblith, S. J., Greenwald, D., DiBarry, A. L., Cooley, S., Ulrich, R. F., Carter, M., & Flesher, S. (1997). Three-year trials of personal therapy among schizophrenic patients living with or independent of Family, I: Description of study and effects on relapse rates. American Journal of Psychiatry, 154(11), 1504-1513.

*Hogarty, G. E., Schooler, N. R., Ulrich, R., Mussare, F., Ferro, P., & Herron, E. (1979). Fluphenazine and social therapy in the aftercare of schizophrenic patients. Archives of General Psychiatry, 36, 1283-1294.

Hunter, J. E., & Schmidt, F. L. (1990). Methods of meta-analysis: Correcting error and bias in research findings. Newbury Park, CA: Sage.

Hunter, J. E., & Schmidt, F. L. (2000). Fixed vs. random effects meta-analysis models: Implications for cumulative research knowledge. International Journal of Selection and Assessment,8, 275 - 292.

Jacobson, N.S., & Truax, P. (1991). Clinical significance: A statistical approach to defining meaningful change in psychotherapy research. Journal of Consulting and Clinical Psychology, 59 (1), 12-19.

Johnson, B. T. (1995). DSTAT: Software for the Meta-Analytic Review of Research Literatures (Version 1.11) [DOS] [Computer Software]. Hillsdale, NJ: Lawrence Erlbaum Associates.

Karon, B. P., & O’Grady, P. (1969). Intellectual test changes in schizophrenic patients in the first six months of treatment. Psychotherapy: Theory, Research, and Practice, 6, 88-96.

Karon, B. P., & O’Grady, P. (1970). Quantified judgements of mental health from the Rorschach, TAT, and clinical status interview by means of a scaling technique. Journal of Consulting and Clinical Psychology, 34, 229-235.

Karon, B. P., & VandenBos, G. R. (1970). Experience, medication, and the effectiveness of the psychotherapy with schizophrenics: A note on Drs. May and Tuma’s conclusions. British Journal of Psychiatry, 116, 427-428.

Karon, B. P., & VandenBos, G. R. (1972). The consequences of psychotherapy for schizophrenic patients. Psychotherapy: Theory, Research, and Practice, 9, 111-120.

Karon, B. P., & VandenBos, G. R. (1975). Treatment costs of psychotherapy versus medication for schizophrenics. Professional Psychology, 6, 293-298.

*Karon, B. P., & VandenBos, G. R. (1981). Psychotherapy of schizophrenia: The treatment of choice. Northvale, NJ: Jason Aronson, Inc.

Kazdin, A. E. (1998). Research Design in Clinical Psychology (3^rd ed.). Boston: Allyn and Bacon.

Kingdon, D. G., & Turkington, D. (1994). Cognitive-behavioral therapy of schizophrenia. New York: Guilford.

Lehman, A. F., Steinwachs, D. M., & project, a. t. C.-i. o. t. P. (1998). At issue: Translating research into practice: The schizophrenia patient outcomes research team (PORT) treatment recommendations. Schizophrenia Bulletin, 24(1), 1-10.

LeLorier, J., Gregoire, G., Benhaddad, A., Lapierre, J., Derderian, F. (1997). Discrepancies between meta-analyses and subsequent large randomized, controlled trials. The New England Journal of Medicine, 337 (8), 536-542.

Levene, H. I., Patterson, V., Murphey, B. G., Overbeck, A. L., & Veach, T. L. (1970). The aftercare of schizophrenics: An evaluation of group and individual approaches. The Psychiatric Quarterly, 44 (2), 296-304.

Liberman, R. P. (1994). Psychosocial treatments for schizophrenia. Psychiatry, 57, 104-113.

*Linszen, D. H., Dingemans, P. M. A. J., Scholte, W. F., Lenior, M. E., & Goldstein, M. (1998). Early recognition, intensive intervention and other protective and risk factors for psychotic relapse in patients with first psychotic episodes in schizophrenia. International Clinical Psychopharmacology, 13(suppl 1), S7-S12.

Lotterman, A. (1996). Specific techniques for the psychotherapy of schizophrenic patients. Madison, CT: International Universities Press.

Malmberg, L., Fenton, M. (2001). Individual psychodynamic psychotherapy and psychoanalysis for schizophrenia and severe mental illness (Cochrane Review). Cochrane Database of Systematic Reviews (1) [Electronic database]. The Cochrane Library. Retrieved from www.cochrane.org.

Marks, J., Sonoda, B., & Schalock, R. (1968). Reinforcement versus relationship therapy for schizophrenics. Journal of Abnormal Psychology, 73 (4), 397-402.

Matthews, J. G., & Burkhart, B. R. (1977). A-B therapist status, patient's diagnosis, and psychotherapy outcome in a psychiatric outpatient population. Journal of Consulting and Clinical Psychology, 45 (3), 475-482.

*May, P. R. A. (1968). Treatment of schizophrenia: A comparative study of five treatment methods. New York: Science House.

*May, P. R. A., Tuma, A. H., Dixon, W. J., Yale, C., Thiele, D. A., & Kraude, W. H. (1981). Schizophrenia: A follow-up study of the results of five forms of treatment. Archives of General Psychiatry, 38, 776-784.

*McGlashan, T. H. (1984). The Chestnut Lodge follow-up study I.: Follow-up methodology and study sample. Archives of General Psychiatry, 41, 573-585.

*McGlashan, T. H. (1984). The Chestnut Lodge follow-up study II.: Long-term outcome of schizophrenia and the affective disorders. Archives of General Psychiatry, 41, 586-601.

McGlashan, T. H., & Keats, C. J. (1989). Schizophrenia: Treatment process and outcome. Washington, D.C.: American Psychiatric Press.

*Messier, M., Finnerty, R., Botvin, C. S., & Grinspoon, L. (1969). A follow-up study of intensively treated chronic schizophrenic patients. American Journal of Psychiatry, 125(8), 1123-1127.

Mojtabai, R., Nicholson, R. A., & Carpenter, B. N. (1998). Role of psychosocial treatments in management of schizophrenia: A meta-analytic review of controlled outcome studies. Schizophrenia Bulletin, 24(4), 569-587.

Morris, S. B., & DeShon, R. P. (2002). Combining effect size estimates in meta-analysis with repeated measures and independent-groups designs. Psychological Methods, 7(1), 105-125.

Mosher, L. R., & Keith, S. J. (1980). Psychosocial treatment: Individual, group, family, and community support approaches. Schizophrenia Bulletin, 6(1), 10-41.

Mueser, K. T., & Berenbaum, H. (1990). Psychodynamic treatment for schizophrenia: Is there a future? Psychological Medicine, 20, 253-262.

*O’Brien, C. P., Hamm, K. B., Ray, B. A., Pierce, J. F., Luborsky, L., & Mintz, J. (1972). Group vs individual psychotherapy with schizophrenics: A controlled outcome study. Archives of General Psychiatry, 27, 474-478.

Pao, P.-N. (1979). Schizophrenic disorders: Theory and treatment from a psychodynamic point of view. New York: International Universities Press.

Pawson, R., & Tilley, N. (1997). Realistic Evaluation. London: Sage.

Perkins, D. O. (1999). Adherence to antipsychotic medications. Journal of Clinical Psychiatry, 60 (suppl), 21-25.

Perris, C. (1989). Cognitive therapy with schizophrenic patients. New York: Guilford.

Robbin s, M. (1993). Experiences of schizophrenia. New York: Guilford.

Rockland, L. (1989). Supportive therapy: A psychodynamic approach. New York: Basic Books.

Rosenthal, R. (1979). The "file drawer problem" and tolerance for null results. Psychological Bulletin, 85, 638-641.

Rosenthal, R. (1991). Meta-Analytic procedures for social research. (Revised ed.). (Vol. Applied social research methods series volume 6). Newbury Park, CA: Sage.

Rosenthal, R., & Rubin, D. B. (1982). A simple general purpose display of magnitude of experimental effect. Journal of Educational Psychology, 74, 166-169.

Rosenthal, R., Rosnow, R. L., & Rubin, D. B. (2000). Contrasts and effect sizes in behavioral research: A correlational approach. Cambridge: Cambridge.

*Rubins, J. L. (1976). Five-year results of psychoanalytic therapy and day care for acute schizophrenic patients. The American Journal of Psychoanalysis, 36, 3-26.

Schmidt, F. L., & Hunter, J. E. (2001). Meta-Analysis. Unpublished manuscript.

Scott, J. E., & Dixon, L. B. (1995). Psychological interventions for schizophrenia. Schizophrenia Bulletin, 21(4), 621-630.

Searles, H. F. (1965). Collected papers on schizophrenia and related subjects. New York: International Universities Press.

Shadish, W. R., & Ragsdale, K. (1996). Random versus nonrandom assignment in controlled experiments: Do you get the same answer? Journal of Consulting and Clinical Psychology, 64 (6), 1290-1305.

Shrout, P. (1998). Causal modeling of epidemiological data on psychiatric disorders. Social Psychiatry and Psychiatric Epidemiology, 33, 400-404.

*Sjöström, R. (1985). Effects of psychotherapy in schizophrenia: A retrospective study. Acta Psychiatrica Scandinavica, 71, 513-522.

Smith, M. L., Glass, G. V., & Miller, T. I. (1980). The benefits of psychotherapy. Baltimore: The Johns Hopkins University Press.

Spitzer, R. L., Endicott, J., & Robins, E. (1975). Research Diagnostic Criteria (RDC) for a selected group of functional disorders. New York: New York State Psychiatric Institute.

*Stanton, A. H., Gunderson, J. G., Knapp, P. H., Frank, A. F., Vannicelli, M. L., Schnitzer, R., & Rosenthal, R. (1984). Effects of psychotherapy in schizophrenia: I. Design and implementation of a controlled study. Schizophrenia Bulletin, 10(4), 520-563.

*Stone, M. H. (1986). Exploratory psychotherapy in schizophrenia-spectrum patients: A reevaluation in the light of long-term follow-up of schizophrenic and borderline patients. Bulletin of the Menninger Clinic, 50, 287-306.

Sullivan, H. S. (1962). Schizophrenia as a human process. New York: W. W. Norton.

Svere, V. (1991). A retrospective follow-up investigation of a group of schizophrenic patients treated in a psychotherapeutic unit: The Kastanjebakken study. Psychopathology, 24, 336-344.

*Tarrier, N., Beckett, R., Harwood, S., Baker, A., Yusupoff, L., & Ugarteburu, I. (1993). A trial of two cognitive-behavioural methods of treating drug-resistant residual psychotic symptoms in schizophrenic patients: I. Outcome. British Journal of Psychiatry, 162, 524-532.

*Tarrier, N., Yusupoff, L., Kinney, C., McCarthy, E., Gledhill, A., Haddock, G., & Morris, J. (1998). Randomised controlled trial of intensive cognitive behaviour therapy for patients with chronic schizophrenia. British Medical Journal, 317, 303-307.

*Walker, R. G., & Kelley, F. E. (1960). Shortterm psychotherapy with hospitalized schizophrenic patients. Acta Psychiatrica et Neurologica Scandinavica, 35, 34-56.

Warner, R. (1994). Recovery from schizophrenia: Psychiatry and political economy. (2^nd ed.). London: Routledge.

White, H. D. (1994). Scientific communication and literature retrieval. In H. Cooper, & L. V. Hedges (Eds.), The Handbook of Research Synthesis (pp. 41-55). New York: Russell Sage Foundation.

*Whitehorn, J. C., & Betz, B. J. (1954). A study of psychotherapeutic relationships between physicians and schizophrenic patients. The American Journal of Psychiatry, 3, 321-331.

*Whitehorn, J. C., & Betz, B. J. (1957). A comparison of psychotherapeutic relationships between physicians and schizophrenic patients when insulin is combined with psychotherapy and when psychotherapy is used alone. American Journal of Psychiatry, 113, 901-910.

Wilson, G. T., & Rachman, S. J. (1983). Meta-analysis and the evaluation of psychotherapy outcome: Limitations and liabilities. Journal of Consulting and Clinical Psychology, 51 (1), 54-64.

Author Note

William H. Gottdiener, The New School University, Department of Psychology, (now at Beth Israel Medical Center in New York City); Nick Haslam, Department of Psychology, The University of Melbourne.

The title of this work was inspired by Smith et al. (1980).

We would like to thank Tina Tan, the members of the Behavioral Sciences Training program at Medical and Health Research Association of New York City, Inc. and National Development and Research Institutes, Inc., and several anonymous reviewers for their constructive feedback on earlier drafts of this article. In addition, this article was written in part while the first author was a post-doctoral fellow in the Behavioral Sciences Training in Drug Abuse Research Program sponsored by Medical and Health Research Association of New York City, Inc. and National Development and Research Institutes, Inc. with funding from the National Institute on Drug Abuse (5 T32 DA07233-09). This paper is based on the first author’s dissertation research, which was conducted under the guidance of the second author at The New School University. Opinions in this paper do not necessarily represent the official positions of the United States Government, Medical and Health Research Association of New York City, Inc., National Development and Research Institutes, Inc., Beth Israel Medical Center, and the University of Melbourne.

Requests for reprints and correspondence concerning this article should be addressed to William H. Gottdiener, Ph.D., Beth Israel Medical Center, Milton and Carroll Petrie Division, First Avenue at 16^th Street, Rm. 6K42, New York, NY 10003. (Email: wgottdiener@chpnet.org).
Table 1

Reliability of Outcome Tests Used To Calculate The Artifact Attenuation Factor

______________________________________________________________________________

Outcome Measure

Reliability Coefficient

______________________________________________________________________________

Psychiatric Status Schedule	.91
Menninger Health Sickness Rating Scale	.92
Camarillo Dynamic Assessment Scale	.88
Rorschach (Holt’s Scoring Method)	.93
Psychotherapy Outcome Interview	.59
Golschalk Social Attenuation/Personal-Disorganization Scale	.92
Behavioral Disturbance Index	.94
Strauss-Carpenter Outcome Scale	.92
Jenkin’s Symptom Rating Scale	.77
Brief Psychiatric Rating Scale	.77
Mean Reliability	.85

______________________________________________________________________________

Table 2

Coding Reliability

______________________________________________________________________________

Variables

Reliability Coefficient

______________________________________________________________________________

Study Design	c = .77
Control group used (yes/no)	k = .74
Sample size	r = .98
Effect size	r = .93
Treatment type	r = 1.00

______________________________________________________________________________

Table 3

Meta-Analysis Results: All Effect Sizes Combined

______________________________________________________________________________

Citation

Effect

Lower

Upper

Ntotal

Pvalue

StndError

______________________________________________________________________________

1. Alanen (1985)	0.47	0.234	0.654	58	0	0.139
2. Alanen (1994)	0.70	0.500	0.829	41	0	0.162
3. Bookhammer (1966)	0.31	-0.545	0.849	47	.50	0.475
4. Buchkremer (1997)	0.77	0.551	0.889	27	0	0.204
5. Bullard (1960)	0.64	0.429	0.78	46	0	0.152
6. Carpenter (1977)	0.13	-0.016	0.279	244	0.08	0.077
7. Coursey (1995)	0.75	0.619	0.840	65	0	0.127
8. Falloon (1985)	-0.00	-0.733	0.729	54	0.993	0.475
9. Fowler (1989)	0.00	-0.882	0.882	5	0	0.707
10. Garety (1994)	0.40	0.034	0.665	32	0.033	0.196
11. Glass (1989)	0.07	-0.352	0.468	23	0	0.223
12. Gottlieb (1951)	0.11	-0.029	0.245	198	0	0.007
13. Grinspoon (1968)	0.6	0.200	0.828	19	0	0.250
14. Gunderson (1984)	0.058	-0.150	0.261	97	0.585	0.107
15. Hamill (1975)	0	-0.343	0.343	33	0	0.182
16. Hogarty (1974)	0.098	-0.010	0.204	686	0.076	0.056
17. Hogarty (1979)	0.312	-0.436	0.805	132	.424	0.403
18. Hogarty (1997)	0.23	0.072	0.375	151	0	0.008
19. Karon (1981)	0.31	0.065	0.52	65	0.014	0.130
20. Levene (1970)	0.22	-0.329	0.658	15	0	0.288
21. Linszen (1998)	0.85	0.772	0.902	76	0	0.117
22. Marks (1968)	0.03	-0.443	0.490	18	0	0.258
23. Mathews (1977)	0.22	-0.275	0.622	18	0	0.258
24. May (1968)	0.512	-0.185	0.866	363	0.141	0.384
25. May (1981)	-0.058	-0.373	0.268	276	0.730	0.170
26. McGlashan (1984)	0.47	0.340	0.581	163	0	0.008
27. Messier (1969)	0.275	-0.043	0.543	143	0.089	0.166
28. O'Brien (1972)	0.66	0.457	0.797	46	0	0.152
29. Rubins (1976)	0.45	0.176	0.658	44	0	0.156
30. Sjöström (1985)	0.39	-0.002	0.685	24	0	0.218
31. Stone (1986)	0.69	0.545	0.794	72	0	0.120
32. Sverre (1991)	0.46	0.008	0.719	26	0	0.208
33. Tarrier (1993)	0.23	-0.156	0.555	28	0	0.200
34. Tarrier (1998)	0.22	-0.004	0.452	58	0	0.134
35. Walker (1960)	-0.14	-0.346	0.008	82	0	0.112
36. Whitehorn (1954)	0.51	0.348	0.642	100	0	0.101
37. Whitehorn (1957)	-0.04	-0.226	0.149	109	0	0.008
Totals	0.31	0.220	0.410	3684*	0	0.054
Corrected Effect	0.36
		BESD Success rate Increased
		BESD Success rate Increased
	ES r	From	To	r²
	0.31	0.35	0.66	0.10
		35%	66%

______________________________________________________________________________

* The Ntotal is the total sample used to calculate all of the effect sizes from a given study. If two effect sizes were calculated in a study with 10 participants, then the Ntotal for that study would be 20.