Each year, 38.2% of the population of the European Union, i.e., 164.8 million persons, suffer from a mental disorder [Reference Wittchen, Jacobi, Rehm, Gustavsson, Svensson and Jönsson120]. This is associated with huge societal and individual burden [34,119]. Prevention has therefore become an integral part of European and international health care policies in order to reduce the prevalence and burden of mental disorders across the lifespan [14,20,36].
1.1. Functional disability in psychotic disorders
Schizophrenia is among the seven leading causes of years lost to disability (YLDs) in adults in Europe [Reference Wittchen, Jacobi, Rehm, Gustavsson, Svensson and Jönsson120]. This is mainly due to the fact that functional recovery rates have not changed substantially over the past 25 years, despite advances in pharmacological and psychological treatments [44,106]. As a consequence, prevention of schizophrenia and psychotic disorders in general has attracted special interest [Reference Solis108].
Functional impairments are present before [Reference Addington, Penn, Woods, Addington and Perkins1] and often worsen until the onset of psychosis [Reference Häfner and an der Heiden35]. Furthermore, they are one of the main predictors of poor clinical outcome including conversion to psychosis [Reference Granö, Kallionpää, Karjalainen, Edlund, Saari and Itkonen28]. This emphasizes the need to intervene as early as possible to avoid or at least diminish these burdens and thereby to prevent transition to full blown psychosis.
1.2. Prevention in clinical high risk states of psychoses
In psychosis research, an indicated prevention approach has been adopted that targets help-seeking persons who experience early signs of emerging psychosis, but do not meet diagnostic criteria. The ultimate goal of this approach is to prevent this condition from converting to psychosis [61,65,67]. Thus, indicated prevention faces two challenges: the accurate identification of the target population and their effective treatment. For the purpose of early detection, two complementary sets of clinical high risk (CHR) criteria have been developed: the ultra-high risk (UHR) [73,126] and the basic symptom approach [41,105]. The UHR criteria were originally developed to detect individuals with an inherent risk for a first episode of psychosis within the next 12 months, and comprise the attenuated psychotic symptom (APS) criterion, the brief limited intermittent psychotic symptom (BLIPS) criterion, and the genetic risk and functional decline criterion [Reference Yung, Yuen, McGorry, Phillips, Kelly and Dell’Olio126]. The basic symptom approach was developed to detect the risk for psychosis as early as possible as defined by the presence of the cognitive-perceptive basic symptoms (COPER) and the cognitive disturbances (COGDIS) criterion [41,105]. Recommendations for the use of the UHR and the basic symptom approach and a review of the underpinning evidence are given in the accompanying European guidance on early detection (see Schultze-Lutter et al., this issue). Notably, fulfilling these criteria only indicates an increased risk for developing psychosis, which is always associated with an error probability resulting in false-positive predictions [Reference Ruhrmann, Schultze-Lutter and Klosterkötter96]. This has fueled ongoing debates about the risk of negative effects associated with the identification and treatment of CHR states of psychoses (e.g., stigmatization, side effects of medication, financial loss) [24,51,83].
1.3. Need for treatment in clinical high risk states of psychoses
In addition to the markedly increased risk for developing psychosis (Schultze-Lutter et al., this issue), the most important argument in favor of an intervention in CHR patients is the reported distress and stigmatization caused by their mental problems, which are already present at the time of referral to an early detection and intervention service [50,99,98,110]. This is demonstrated by their high levels of hopelessness, depression, anxiety, and poor quality of life, in comparison to other help-seeking patients and patients with first-episode psychosis [30,31,95]. In fact, they often fulfill the diagnostic criteria for other mental disorders that require treatment, in particular depression, anxiety, and substance abuse or dependence [25,42,100]. Furthermore, CHR patients seem to exhibit poor coping skills, low self-efficacy, and excessive external attributions that resemble profiles of depressive patients [Reference Schmidt, Grunert, Schimmelmann, Schultze-Lutter and Michel104].
CHR patients also demonstrate abnormalities in neuro- and social cognition that are usually intermediate between those of healthy controls and schizophrenia patients [12,22,27,112], and are associated with markedly impaired functional outcome and negative symptoms [15,56,71,101,102,107]. These neuro- and social-cognitive abnormalities are also accompanied by various abnormalities in functional and structural imaging [11,52,107], electrophysiological measures [10,85,118], and neurochemistry [19,33,55]. In summary, CHR patients are in need for treatment – independent of any potential risk to develop psychosis in the future [Reference Ruhrmann, Schultze-Lutter and Klosterkötter96].
1.4. Requirements for early intervention approaches
In accordance with this obvious need for treatment, an increasing number of interventions have been evaluated in CHR samples in recent years [97,89,109]. With the primary goal of preventing conversion to psychosis, these studies have drawn upon well-established interventions for adult schizophrenia patients and used conversion to psychosis as their primary outcome [63,68,75]. Other more recently developed interventions take into account that CHR patients not only suffer from risk symptoms but also from several other mental problems and have targeted a broader array of outcomes in various settings with various intervention techniques (e.g., intensive case management, multi-family psycho-education) [29,32,59]. Yet most “new generation” intervention studies have an uncontrolled single-group design, and therefore lack methodological rigor, and were not included in recent meta-analyses on the efficacy of interventions in CHR states using randomized controlled trials (RCTs) [23,43,57,92,109,117].
Current guidelines have not yet considered these “new generation” interventions [9,18,82]. Moreover, no sufficient evidence-based differential indication for the available interventions exists as this requires the evaluation of the type of intervention as a potential moderator variable in meta-analyses and/or in direct multi-head-to-head comparisons in large RCTs. Age should also be studied as a potential moderator because CHR samples commonly include adolescents and young adults who differ in their social, emotional, and cognitive developmental state.
The aim of this guidance paper on early intervention in CHR states was to evaluate the efficacy of interventions that aim at preventing the conversion to psychosis and/or a deterioration of functional outcome. We also considered the potential moderating effects of age and intervention type. From this analysis, we have derived evidence-based recommendations on early intervention in CHR states assessed according to the recommendations provided by the EPA guidance on early detection (see Schultze-Lutter et al., this issue).
2.1. Study selection
2.1.1. Literature search
We conducted a systematic literature search in June 2014 in PubMed (no time limit), PsycInfo (no time limit), Scopus (no time limit) that covers all journals included in Embase, and in the Cochrane Collaboration Controlled Trials Register using the following search terms and syntax: ((prevention) OR (early intervention) OR (treatment) OR (therapy)) AND ((risk) OR (prodrome) AND ((psychosis) OR (schizophrenia)). Furthermore, we inspected reference lists of all identified reviews and meta-analyses.
2.1.2. Selection criteria
We included all studies in our meta-analysis that: evaluated a psychological and/or pharmacological intervention including interventions with nutritional supplements and other substances, included a majority (>50%) of CHR patients as defined by the UHR and/or the basic symptom criteria, reported conversion rates and/or functional outcome as intervention outcome, and were published in English. Exclusion criteria were: studies with samples that were also sub-sets of other studies included in our analyses with a larger sample size and/or a longer follow-up period, studies that were only published as an abstract or trial protocol, case-reports, and studies that used an observational naturalistic design.
2.1.3. Selection process and quality assessment
Fig. 1 illustrates the flowchart of the study selection process following the PRISMA guidelines [Reference Moher, Liberati, Tetzlaff and Altman74]. Two authors (S.J.S., S.R.) independently assessed trial quality using the methodology checklist of the Scottish Intercollegiate Guideline Network (SIGN) for randomized controlled trials. In cases of disagreement, discrepancies were discussed among the raters or supplementary information was requested from the authors until agreement was reached.
2.2. Literature analysis
2.2.1. Data extraction
The following variables were extracted from the included studies:
study characteristics: country, sample sizes, design, inclusion/exclusion criteria, sample description, experimental and control condition, follow-up (defined as the time-period after baseline assessments), and drop-out rate;
prevalence of psychosis at all available follow-up assessment points (conversion rates). Following an intention-to-treat (ITT) approach, the total number of persons randomized to each group was chosen as the point of reference in order to avoid higher conversions rates due to an increase of drop-outs during the course of the trial;
means and standard deviations of functional outcome measures of the experimental (EG) and the control group (CG) at baseline and at each follow-up assessment.
2.2.2. Meta-analytic procedure
For binary data (i.e. conversion rates), we estimated relative risks (RR) with their 95% confidence intervals (CIs) [Reference Sullivan and Feinn111]. Additionally, we calculated the number-needed-to-treat statistics (NNTs) with their 95% CIs for the combined studies [Reference Leucht, Corves, Arbter, Engel, Li and Davis54]. Notably, RRs and NNTs cannot be calculated for uncontrolled intervention designs for lack of a normative conversion rate.
For continuous data (i.e. functional outcome), the effect size Hedges’ g was estimated as the standardized mean difference. For between-group comparisons, i.e. intervention studies with a control group, Hedges’ gb was calculated as the difference of the mean scores of the treatment and the control group at the respective follow-up divided by their pooled standard deviation [21,38]. For within-group comparisons, i.e. controlled and uncontrolled intervention studies with pre- and post-assessments for each condition, Hedges’ gw was calculated by the difference of the respective mean scores at pre- and post-assessment divided by their pooled standard deviation [Reference Lakens53]. To test if the within-group effect sizes differ between EG and CG, the Q b-statistic was calculated [Reference Borenstein, Borenstein, Hedges, Higgins and Rothstein13].
For both binary and continuous data, heterogeneity between effect sizes was determined with the Q w-statistic [Reference Hedge and Olkin37]. In addition, I 2 served as an estimate of the relative size of homogeneity. Values of 25, 50, and 75% are regarded as signifying low, moderate, and high heterogeneity [Reference Higgins, Thompson, Deeks and Altman39]. If no substantial heterogeneity was detected, a fixed-effects model was applied. Otherwise, results of random-effects models are reported [Reference Borenstein, Borenstein, Hedges, Higgins and Rothstein13].
The meta-analytic calculations for this paper were conducted with the Review Manager (version 5.3) of the Cochrane Collaboration.
2.2.3. Potential moderator variables
Additionally, we investigated the impact of two potential moderator variables on treatment effects: intervention type and age. For this purpose, the meta-analytic calculations for between- (RR, Hedges’ gb) and within-group comparisons (Hedges’ gw) described above were performed separately for the respective sub-groups. Intervention type was distinguished by psychological interventions (PSY) and pharmacological interventions with antipsychotics or other substances (MED; Table 1). Age was rated in two categories: ≥50% minors (YOUTH), when the mean age ≤18 with an upper standard deviation still spanning patients ≤18 years, and samples with a proportion of minors of <50%, when the mean age >18 years (ADULT; Table 1). To test if the within-group effect sizes differ between the respective sub-groups, the Q b-statistic was calculated as already described above.
APS: attenuated psychotic symptoms; AUS: Australia; AUT: Austria; BLIPS: brief limited intermittent psychotic symptoms; BS: basic symptom approach; CA: Canada; EG: experimental group; CG: control group; GER: Germany; JPN: Japan; NL: Netherlands; RCT: randomized controlled trial; UK; United Kingdom; UHR: ultra-high risk approach; USA: United States of America.
2.3. Development of recommendations
In line with the EPA's methodological approach within the guidance project [Reference Gaebel and Möller26], the consensus process was restricted to the experts, i.e. authors. General consensus on recommendations was achieved by circulating results of the literature search and manuscript drafts prepared by the main authors (S.J.S., S.R.) to all co-authors for feedback and discussion after the following steps: compilation of studies to be included in meta-analyses, including their grade of evidence rating, conducting of analyses and first drafting of the manuscript, and recurrently adaptating after each feedback-related until full agreement among authors was reached on the manuscript's submission version. This step was also performed once more after receiving external review.
Furthermore, during the process of guidance development, the manuscript's submission version underwent review by the EPA Guidance Committee and EPA Board (see Acknowledgements) to guarantee that authors had adhered to the consented methodology. Only upon its approval by both committees, the manuscript was submitted for external review.
3.1. Literature search
The process of the literature search is detailed in Fig. 1. The initial search produced 10,206 papers. After the exclusion of duplicate reports and screening of 8452 titles, 77 abstract were possibly relevant. They were further screened for inclusion and exclusion criteria, which reduced them to 61 full-text papers. These papers were especially checked for redundancies, i.e., significant overlap of samples, and presence of additional follow-up data in already included samples. Finally, 15 independent studies covered by 25 papers were included in the meta-analysis.
3.2. Evidence-base of the European Guidance
3.2.1. Sample and study characteristics of the included studies
Trial characteristics of the 15 studies that cover a total of 1394 participants are detailed in Table 1. Sample sizes at baseline ranged from 11 to 288 (mean = 87.12, S.D. = 73.82). Samples consisted mainly of young adults (mean = 20.44 years, S.D. = 3.29, range = 15.70–26.80). Nine samples (60.0%) included mainly adults, whereas six (40.0%) had a majority of minors. In eight studies (53.3%), more than half of the participants were male (range = 33.0–83.0%). Six studies (40.0%) provided information on co-morbidities consistently reporting the highest prevalence for affective (26.0–41.0%) and anxiety disorders (27.0–46.0%).
Ten studies (66.6%) were carried out in or at least involved well-established early detection and intervention services. Nine studies (60.0%) assessed conversion to psychosis. Two (13.3%) used the Positive and Negative Syndrome Scale (PANSS [Reference Kay, Fiszbein and Opfer46]), three (20.0%) the Structured Interview for Psychosis-Risk Syndromes (SIPS [Reference McGlashan, Walsh and Woods64]), one (0.7%) the Early Recognition Inventory (ERIraos [Reference Maurer, Hörrmann, Trendler, Schmidt and Haefner60]), and three (20.0%) the Comprehensive Assessment of At-Risk Mental States (CAARMS [Reference Yung, Phillips, Simmons, Ward, Thompson and French127]) in its original [Reference Morrison, French, Stewart, Birchwood, Fowler and Gumley78] and new version in 2006 [69,116]. Fourteen studies (93.3%) assessed functional outcome; most of them (n = 11, 78.6%) with the Global Assessment of Functioning Scale (GAF ). The follow-up duration ranged from 2 to 48 months (mean = 16.33 months, S.D. = 13.19). Twelve studies (80.0%) were RCTs, and three studies (20.0%) used an uncontrolled single-group design. Levels of evidence for RCTs according to SIGN varied between GE 1− (i.e., RCTs with a high risk of bias) and GE 1++ (i.e., RCTs with a very low risk of bias). Studies with no control group were generally rated on GE 2− (i.e., case-control studies with a high risk of confounding, bias, or chance, and a significant risk that the relationship is not causal; Table 1).
3.2.2. Treatment characteristics of the included studies
184.108.40.206. Psychological interventions
Nine studies (60.0%) reported on the efficacy of psychological interventions with a mean therapy duration of 6.87 months (S.D. = 3.7, range = 2–12), a mean follow-up period of 16.67 months (S.D. = 10.40, range = 2–36), and a drop-out rate between 15.0 and 45.0% (Table 1). Five interventions used cognitive-behavioral therapy (CBT) techniques such as normalization, cognitive and behavioral experiments, and cognitive restructuring to improve stress- and symptom-management. These compared CBT with monitoring [75,78], supportive therapy [Reference Addington, Epstein, Liu, French, Boydell and Zipursky2], supportive therapy with placebo [Reference McGorry, Nelson, Phillips, Yuen, Francey and Thampi69], and other evidence-based interventions for the disorders patients sought help [Reference van der Gaag, Nieman, Rietdijk, Dragt, Ising and Klaassen116]. One uncontrolled study evaluated cognitive remediation therapy (CRT) in CHR patients [Reference Hooker, Carol, Eisenstein, Yin, Lincoln and Tully40]. Moreover, a multi-family psycho-educational group program was evaluated first in one uncontrolled study [Reference O’Brien, Zinberg, Bearden, Daley, Niendam and Kopelowicz87], and next in a RCT with enhanced care as control condition [Reference Miklowitz, O’Brien, Schlosser, Addington, Candan and Marshall72]. One of the included RCTs [Reference Bechdolf, Wagner, Ruhrmann, Harrigan, Putzfeld and Pukrop7] combined all of the aforementioned approaches with social skills training and compared this integrated psychological intervention with supportive therapy.
220.127.116.11. Pharmacological studies with antipsychotics or nutritional supplements
Six pharmacological studies (40.0%) have been published, two uncontrolled studies (33.3%) and four RCTs (66.7%). The mean therapy duration was 6.83 months (S.D. = 4.31, range = 2–12), the mean follow-up period was 15.29 months (S.D. = 16.23, range = 2–48), and the drop-out rate ranged between 13.0 and 55.0%. These trials investigated the efficacy of aripiprazole [Reference Woods, Tully, Walsh, Hawkins, Callahan and Cohen123] and perospirone [Reference Tsujino, Nemoto, Morita, Katagiri, Ito and Mizuno113] using an uncontrolled design, and, as RCTs, olanzapine versus placebo [Reference McGlashan, Zipursky, Perkins, Addington, Miller and Woods63], ‘risperidone plus CBT’ versus need-based intervention (NBI) [Reference McGorry, Yung, Phillips, Yuen, Francey and Cosgrave68], ‘amisulpride plus NBI’ versus NBI [Reference Ruhrmann, Bechdolf, Kühn, Wagner, Schultze-Lutter and Janssen94], and ‘risperidone plus CBT’ versus ‘placebo plus supportive therapy’ [Reference McGorry, Nelson, Phillips, Yuen, Francey and Thampi69]. Only one included pharmacological study did not use antipsychotic medication but a neuroprotective approach, and investigated the effect of omega-3 polyunsaturated fatty acids (PUFAS) in CHR patients compared to placebo in a RCT [Reference Amminger, Schäfer, Papageorgiou, Klier, Cotton and Harrigan4]. Side effects of each pharmacological trial are listed in Table 1.
3.2.3. Excluded studies and ongoing trials
Studies that had to be excluded because their trials are still underway or their results have not been published yet are briefly summarized below, as they may entail important information on current developments in early intervention in CHR states.
Recent psychological trials paid special attention, not only to risk symptoms, but also to other common features or stressors of CHR samples such as deficits in neuro- and social cognition and in coping. Studies in adolescents that were excluded due to their naturalistic design emphasized the importance of good and frequent net-working of multi-professional teams in achieving not only symptomatic and functional gains but also reductions in depression, hopelessness, and anxiety [29,32]. One additional trial investigated the potential beneficial effects of CBT based on the intervention of Morrison et al. [Reference Morrison, French, Walford, Lewis, Kilcommons and Green75] in CHR samples [Reference Kim, Lee, Kang, Kim, Choi and Park48] with further trials being underway using CRT [70,114,115].
Recent ongoing RCTs have been designed to disentangle the effects of pharmacological and psychological interventions [Reference Bechdolf, Müller, Stützer, Wagner, Maier and Lautenschlager6] and to replicate the neuroprotective effect of PUFAS [66,84]. Other studies used lithium as a neuroprotective agent [Reference Berger, Wood, Ross, Hamer, Wellard and Pell8] or followed the glutamate hypothesis of psychosis and investigated glycine [122,124] or – still ongoing – d-serine . Results for lithium and glycine are promising but no effects on conversion rates or functional outcome were reported. Other pharmacological trials of antidepressants [16,86] and aspirin  used a naturalistic design and/or are still ongoing.
3.3. Meta-analytic results
3.3.1. Conversion rates
All seven RCTs [2,7,68,72,75,78,116] (n = 916) that investigated psychological and/or pharmacological interventions reported on the conversion rates at 6-month follow-up (Fig. 2a). As there was no relevant heterogeneity among the study effects, the results of the fixed-effects model are reported. On average, 10.4% of the control and 3.6% of the experimental group converted to psychosis within six months after baseline assessment. The experimental condition significantly reduced the risk for conversion to first episode psychosis by 64.0% relative to the control condition as indicated by an overall pooled RR of 0.36 (95% CIs = 0.21, 0.60; P < 0.001). NNT based on the pooled risk difference (RD) of −0.07 (95% CIs = −0.10, −0.04; P < 0.001) was 15 (95% CIs = 10, 25).
At 12-month follow-up, nine RCTs [2,4,7,63,68,69,75,78,116] (n = 1071) provided data for 10 comparisons (Fig. 2b). Again, study effects were homogeneous, and the experimental condition significantly reduced the conversion risk on average by 56.0%. While 17.8% of the control group converted to psychosis, only 8.1% of the experimental group did. This corresponds to a significant pooled RR of 0.44 (95% CIs = 0.31, 0.61; P < 0.001) and a NNT of 10 (95% CIs = 8, 17; RD = −0.1, 95% CIs = −0.14, −0.06; P < 0.001).
All four RCTs [2,7,78,116] (n = 668) that investigated conversion rates at 18-month follow-up had offered CBT for up to 12 months and assessed study subjects at least six months after the end of treatment (Table 1). The conversion rates were 13.7% in the control and 5.4% in the EG group. The heterogeneity between the study effects increased compared to 6- and 12-month follow-up but was still moderate and insignificant. Using the fixed-effects model resulted in a pooled RR of 0.41 (95% CIs = 0.25, 0.69; P < 0.001), an average risk reduction of 59.0%, and a NNT of 13 (95% CIs = 8, 25; RD = −0.08, 95% CIs = −0.13, −0.04; P = 0.01; Fig. 2c).
We pooled the conversion rates of five studies (n = 595) that investigated a follow-up period of more than 18 months: three studies reported conversion rates at 24 months [7,63,78], one study [Reference Morrison, French, Walford, Lewis, Kilcommons and Green75] at 36 months, and one study [Reference McGorry, Yung, Phillips, Yuen, Francey and Cosgrave68] at 36 to 48 months (Table 1). No significant heterogeneity was present. Again, the pooled RR of 0.58 demonstrates a significant risk reduction by the experimental condition (95% CIs = 0.40, 0.85; Fig. 2d). NNT was 13 (95% CIs = 8, 34; RD = −0.08, 95% CIs = −0.14, −0.03; P < 0.01). Conversion rates of 19.0% in the CG and of 11.8% in the EG were the highest during the follow-up period.
3.3.2. Functional outcome
Seven RCTs [2,63,69,72,78,94,116] (n = 869) assessed the functional outcome after a short-term follow-up of between two and six months (Table 1). Due to significant heterogeneity a random-effects model was applied that revealed no significant overall between-group effect for the experimental condition compared to the control condition (Hedges’ gb = 0.05, 95% CIs = −0.22, 0.32; P = 0.70; Fig. 3a). However, I 2 = 73% indicated high heterogeneity which was mainly introduced by one study [Reference McGorry, Nelson, Phillips, Yuen, Francey and Thampi69] with a slightly negative effect of the experimental condition (Hedges’ gb = −0.22, 95% CIs = −0.74, 0.30; P = 0.40) and could be reduced to a moderate degree (Q w(6) = 9.88, P =0.13; I 2 = 39.0%) by its exclusion. This resulted in a significant improvement in functional outcome favoring the experimental condition (Hedges’ gb = 0.22, 95% CIs = 0.02, 0.42; P = 0.03).
At 12-month follow-up, eight RCTs [2,4,7,63,68,69,78,116] (n = 798) assessed functional outcome (Table 1). Again, significant heterogeneity between effect sizes was present. No significant treatment effect on functional outcome was detected (Hedges’ gb = 0.43, 95% CIs = −0.12, 0.97; P = 0.13; Fig. 3b). Testing successively the effect of excluding each single study revealed that high levels of heterogeneity (I 2 = 92%) between effect sizes were mainly due to the large positive effect (Hedges’ gb = 4.85, 95%s CI = 3.97, 5.73; P < 0.001) in the study of Amminger et al. [Reference Amminger, Schäfer, Papageorgiou, Klier, Cotton and Harrigan4]. When this study was excluded from the analysis, heterogeneity decreased substantially (Q w(8) = 4.71, P = 0.79; I 2 = 0.0%) but the overall effect on functional outcome remained insignificant (Hedges’ gb = −0.02, 95% CIs = −0.13, 0.17; P = 0.77).
Only four RCTs assessed functional outcome at least 18 months after baseline: two at 18 months [2,116], one at 24 months [Reference Morrison, French, Stewart, Birchwood, Fowler and Gumley78], and one between 36 to 48 months [Reference McGorry, Yung, Phillips, Yuen, Francey and Cosgrave68]. Therefore, we summarized the effect of these four studies. No significant heterogeneity among the four RCTs (n = 344) was detected, and again no significant effect on functional outcome was revealed (Hedges’ gb = 0.13, 95% CIs = −0.09, 0.34; P = 0.25; Fig. 3c).
Within-group effect sizes (Hedges’ gw) were calculated for the experimental and the control group separately. Random-effects model were applied to all sub-group comparisons because significant heterogeneity was present at all follow-up assessments (Table 2). Notably, both experimental and control group demonstrated significant functional improvements at short-term follow-up (2–6 months; Hedges’ gw ≥ 0.62, P ≤ 0.002) and at between 9- to 12-month follow-up (Hedges’ gw ≥ 0.84, P < 0.001). These within-group effects on the control group were larger than those on the experimental group but these differences were not significant at both follow-ups (Q b(1) ≤ 1.12, P ≥ 0.29). The control group still demonstrated significant functional gains at between 18 to 48 months (Hedges’ gw = 0.64, 95% CIs = 0.12, 1.17; P = 0.02) but not the experimental group. Yet again these between-group differences were not significant (Q b(1) = 0.01, P = 0.91; Table 2).
CG: control group; EG: experimental group; gw: standardized mean difference for pre-post improvements in the respective group; MED: pharmacological intervention; PSY: psychological intervention;
effect sizes are presented in a way that positive values indicate an improvement in functional outcome.
a and b refer to two different tests that were used in a study to assess functional outcome.
3.4. Moderator analysis
3.4.1. Intervention approach
At 6- and 12-month follow-up, both conditions significantly reduced the conversion rates, and no differential treatment effect between both groups could be detected (Table 3). No comparative data was available for the 18-month follow-up, as all available studies administered PSY (Table 1). Between the 24- to 48-month follow-up, only PSY significantly reduced the risk for conversion to psychosis. However, the between-group effect was again insignificant.
MED: pharmacological intervention; PSY: psychological intervention; RR: risk ratio;
effect sizes are presented in a way that positive values indicate a reduction in conversion rates.
With regard to functional outcome, the comparison of between-group effects (Hedges’ gb) of RCTs revealed that PSY and MED did not differ in their effects on functional outcome at any follow-up assessment (Q b(1) ≤ 1.96, P ≥ 0.16). This negative finding was likely caused by the general lack of significant effects in functional improvement between the experimental and control conditions (see above). When within-group effects (Hedges’ gw) were compared between PSY and MED, both PSY and MED showed significant improvements in the experimental condition that did not differ significantly from each other (Table 2). Notably, at medium-term follow-up (18–48 months), only PSY revealed a significant within-group effect on functional outcome in the experimental condition but not MED. This between-sub-group difference was significant, but MED comprised only one study [Reference McGorry, Yung, Phillips, Yuen, Francey and Cosgrave68]. With regard to the control condition, only PSY achieved significant improvements in functional outcome.
3.4.2. Age group
ADULT demonstrated a significant reduction of conversion rate at 6-, 12-, 18-month, and medium-term follow-up (24–48 months; Table 3). YOUTH generally had a lower risk reduction which was significantly larger at 12 months but not at 6- and 24 to 48-month follow-ups. However, all other differences between sub-groups were insignificant.
With regard to functional outcome, potential age effects could only be investigated after 6 and 12 months because studies with longer follow-up periods included only ADULT samples (Table 1). Both ADULT and YOUTH in the experimental condition did not increase their functional outcome after 6 and 12 months relative to the control condition (Q b(1) ≤ 2.10, P ≥ 0.15). With regard to changes in functional outcome within each age group, ADULT benefited from both the experimental condition and the control condition, and improved their functional outcome significantly at 6 and 12 months independent of the treatment condition. YOUTH however only demonstrated functional gains in the experimental condition and only at 12 months but they were significantly larger than those of ADULT. Yet again, all other between-sub-group differences were insignificant (Table 2).
4.1. Literature review and meta-analysis as evidence-base
Based on the described evidence-base, seven recommendations were derived with different degrees of evidence and grades of recommendations according to the methodology checklist of the Scottish Intercollegiate Guideline Network (SIGN) for randomized controlled trials. Grades of recommendations were A (“meta-analysis, systematic review, or RCTs with very low risk for bias”), C (“large body of evidence including well-conducted case-control, or cohort studies with a low risk of bias”), and D (“expert opinion”).
4.2. Proposed recommendations of the European guidance
4.2.1. Recommendation 1 (grade of recommendation: D)
In line with the general EPA guidance on prevention of mental disorders [Reference Campion, Bhui and Bhugra14], the EPA considers that an early intervention in patients presenting with a clinical high risk (CHR) of psychosis should not only aim to prevent the first episode of an affective or non-affective psychotic disorder but also the development or persistence of functional, i.e. social, educational, or vocational deficits.
4.2.2. Recommendation 2 (grade of recommendation: C)
The EPA considers that any psychosis-preventive intervention requires that the CHR status was assessed in full accordance with the EPA guidance on early detection of psychosis (Schultze-Lutter et al., this issue).
4.2.3. Recommendation 3 (grade of recommendation: A)
The EPA considers that psychological, in particular CBT, as well as pharmacological interventions are able to prevent or at least postpone a first psychotic episode in adult CHR patients.
4.2.4. Recommendation 4 (grade of recommendation: D)
The EPA considers that in adult CHR patients a staged intervention model should be applied with the least restrictive service approach, i.e., CBT, being offered as first choice. Where psychological interventions have proved ineffective, they should be complemented by low dose second-generation antipsychotics in adult CHR patients if severe and progressive CHR symptomatology (APS with only minimal or clearly declining insight, or BLIPS in higher or increasing frequency) is present and with the primary aim to achieve a degree of symptomatic stabilization that is required for psychological interventions to be effective. Thus, any long-term antipsychotic treatment with a primarily preventive purpose is not recommended.
4.2.5. Recommendation 5 (grade of recommendation: D)
The EPA considers that any intervention in CHR should also address current individual needs and other mental disorders present (co-morbidities), in particular depression and anxiety, according to their respective treatment guidelines. These disorders should be thoroughly assessed and monitored regularly by a specialist (psychiatrist, clinical psychologist, or equivalent mental health professional).
4.2.6. Recommendation 6 (grade of recommendation: A)
The EPA considers that the current evidence for the psychosis-predictive value of CHR criteria (Schultze-Lutter et al., this issue) and for the efficacy of psychological and pharmacological interventions in children and young adolescents is not sufficient to justify primarily preventive interventions.
4.2.7. Recommendation 7 (grade of recommendation: D)
The EPA considers that in children and adolescents, specific psychological interventions with the aim to improve functioning should be provided as part of an overall treatment plan and complemented by interventions for other psychosocial problems and co-morbid mental disorders according to their treatment guidelines. CHR symptoms should be carefully monitored and assessed for a potential progression over an extended period, and the treatment plan should be adapted according to their course.
We found that recommendations for early intervention in CHR states could be derived from the current evidence-base. In line with the available treatment guidelines [9,18,82], this guidance paper emphasizes the need to target both conversion rates and functional impairments. CBT is regarded as the first choice intervention for the prevention of conversion to psychosis but it might be complemented by pharmacological interventions with low dose second-generation antipsychotics for symptomatic stabilization, if risk symptoms limit the efficacy of CBT. Our recommendations advance existing guidelines in that they provide guidance on a differential application of early interventions in psychoses with respect to developmental aspects and type of intervention, despite limited availability of evidence.
Our results suggest that early interventions can significantly reduce conversion rates in adult CHR patients at short- to medium-term follow-up. However, the effect of these interventions may only be specific for conversion rates but not for functional outcome because the experimental conditions did not achieve larger functional improvements than the control conditions, i.e., monitoring, supportive therapy, and TAU including evidence-based treatment for axis-I and axis-II disorders. This indicates that patients have functionally benefited from control interventions to a similar degree. This may be due to the fact that this particular patient group is quite heterogeneous, for example, with regard to their individual vulnerability, their developmental status, their level of functional impairments, different environmental factors [Reference Kirkbride, Fearon, Morgan, Dazzan, Morgan and Tarrant49], and the prevalence of co-morbid mental health issues. Therefore, it seems likely that a ‘one-fits-all’ approach may not sufficiently address the intra- and inter-individual heterogeneity of CHR patients. Consequently, future early intervention programs should be optimized with regard to their effect on functional outcome by allowing a need- and evidence-based tailored service provision derived, for example, from personalized risk stratification by a prognostic index that does not only stratify risk for psychosis but also for disabling functional impairment [96,97]. Another implication of the lack of a differential treatment effect but significant within-group effect on functional outcome is that control interventions are also effective in improving functional outcome and, from the exclusive point of view of functional improvement, may be an alternative treatment option for current early intervention programs.
The identification of moderator variables such as age is another means of optimizing treatment approaches by identifying sub-groups of patients that can be targeted most successfully with an intervention adjusted to be developmentally appropriate [Reference Schmidt and Schimmelmann103]. Our meta-analysis provides preliminary evidence that early intervention programs are less effective in reducing conversion to psychosis in youth compared to adult patients. This may be due to the lower conversion rates generally found in children and adolescents (Schultze-Lutter et al., this issue) but needs to be interpreted cautiously due to the lack of studies consisting of youth samples, in particular with regard to conversion rates. However, together with the result that youth also achieved lower functional improvements than adults this suggests that current intervention programs do not sufficiently address the special needs and developmental stage of younger CHR patients.
There are additional limitations to our study that need to be considered. Almost all pharmacological interventions also offered some kind of psychosocial and/or psychological intervention, and in psychological trials patients were also allowed to take medication. Therefore, the positive treatment effects of most trials cannot be clearly attributed to one intervention approach. Furthermore, the number of studies included with a follow-up longer than 12 months was small. Therefore, no recommendations with regard to the optimal treatment duration and no firm conclusions about the maintenance of the effects can be derived. However, because of their unfavorable side effect profiles [17,47], pharmacological interventions with antipsychotics should only be applied following thorough cost-benefit considerations and only for a limited time-period with the primary aim to achieve symptomatic stabilization as a starting point for psychological interventions but not with the aim to prevent conversion to psychosis. Pharmacological interventions with PUFAS and antidepressants seem promising so far but require further replication.
Another limitation is the use of the GAF scale to assess functional outcome as it is confounded with the current level of psychiatric symptoms. As a consequence, one cannot conclude if positive within-group effects in functional outcome were mediated by symptomatic improvements, if they are only a spurious effect of these, or if both current early intervention programs and their respective control conditions indeed improved functional outcome directly. Therefore, future studies need to investigate symptomatic and functional improvements separately. Furthermore, at follow-ups several studies reported the mean scores of functional outcome of the remaining sample without converters. This may have led to samples that are biased in that they demonstrate fewer functional deficits than ITT samples and consequently less possibility to achieve functional improvements because converters usually exhibit more pronounced functional deficits than non-converters [Reference Jang, Shin, Shim, Park, Kim and Jang45]. Both limitations may explain why we did not find significant improvements favoring the experimental condition at all follow-ups.
6. Conclusions and perspectives
The accumulating evidence on early intervention in CHR states of psychoses allowed us to derive some first evidence-based recommendations. Yet, our results also demonstrate the need for individualization of treatment and for the development of differential treatment indication. With regard to the still increasing conversion rates beyond 24-month follow-up in observational studies (Schultze-Lutter et al., this issue), interventions should also include longer follow-ups. Moreover, the focus of interventions in CHR patients needs to be broadened with regard to outcomes and intervention approaches (e.g., vocational rehabilitation).
Disclosure of interest
The authors declare that they have no conflicts of interest concerning this article.
Obtaining the required approval of the EPA Guidance Committee and EPA Board was coordinated by the EPA President, Wolfgang Gaebel (Germany). In alphabetical order, EPA Guidance Committee members were: Dinesh Bhugra (UK), Peter Falkai (Germany), Andrea Fiorillo (Italy), Wolfgang Gaebel, Reinhard Heun (UK), Hans-Jürgen Möller (Germany), Michael Musalek (Austria) and Danuta Wasserman (Sweden); EPA Board members were: Sue Bailey (UK), Julian Beezhold (UK), Geert Dom (Belgium), Peter Falkai, Andrea Fiorillo, Wolfgang Gaebel, Silvana Galderisi (Italy), Paz García-Portilla (Spain), Philip Gorwood (France), Cécile Hanon (France), Andreas Heinz (Germany), Marianne Kastrup (Denmark), Levent Küey (Turkey), Tamas Kurimay (Hungary), Michael Musalek, Wulf Rössler (Switzerland), Jerzy Samochowiec (Poland), Rutger J. van der Gaag (Netherlands) and Danuta Wasserman (source: http://www.europsy.net/).
The authors want to thank Jana Roos for her assistance in preparing the manuscript.