This systematic review has been funded by the Canadian Institutes for Health Research (Funding Reference Number KA1 – 119795). The protocol has been registered in the PROSPERO prospective register of systematic reviews (CRD42012003194).
The review methodology was designed to be consistent with reporting guidelines described in the Meta-analysis of Observational Studies in Epidemiology statement [33
] and the Preferred Reporting Items for Systematic Reviews and Meta-analyses statement [34
]. Key questions for this systematic review are based on the USPSTF analytic framework [8
]. The USPSTF logic model describes a process for determining whether screening benefits likely outweigh harms, including the specific populations, screening procedure, and health outcomes to be considered [8
]. The model assumes that patients with undetected depression undergo screening, which classifies patients as cases or non-cases of MDD, and that the evaluation of the accuracy of screening tests is a key component. Treatment studies are evaluated to determine the degree to which they improve symptoms of depression. Trials of actual screening programs, in which health outcomes are evaluated for screened versus unscreened children and adolescents are, if they exist, the final standard by which the value of screening can be tested in terms of its ability to improve depression outcomes and produce benefits to children and adolescents that exceed potential harms from screening.
Four distinct sets of searches will be conducted, one for Key Question #1 (accuracy of screening tools), one for Key Question #2 (effects of treatment), one for Key Question #3 (effects of screening), and one for Key Question #4 (harms). To identify studies relevant to Key Question #1 (accuracy of screening tools), the MEDLINE, EMBASE, PsycINFO, HAPI, and LILACS databases will be searched. For Key Questions #2 (effects of treatment), #3 (effects of screening), and #4 (harms of screening), the MEDLINE, EMBASE, PsycINFO, Cochrane CENTRAL, and LILACS databases will be searched. All search strategies have been peer-reviewed. There will be no language restrictions. See Additional file 2
for specific search terms that will be used.
Searches will be run from January 2006 to the present because an AHRQ systematic review on depression screening in children and adolescents [1
] searched to May 2006. Eligible studies from that review will be included in the present review. For the effects of depression treatment, the previous AHRQ review [1
] included only studies on the effects of selective serotonin reuptake inhibitors (SSRIs) and psychotherapy. However, serotonin-norepinephrine reuptake inhibitors (SNRIs), other non-SSRIs (bupropion, mirtazapine, trazodone), and exercise are also treatments for depression that may be commonly used with children or adolescents in general pediatric settings or primary-care and family medicine settings. Thus, for Key Question #2 (effects of treatment), an additional search with no date restriction will be run for studies on the effects of SNRIs, a select group of non-SSRIs (bupropion, mirtazapine, trazodone), and exercise.
In addition, manual searching will be done for the period of 3 months prior to the final database search on the references of eligible original articles, relevant systematic reviews and meta-analyses, and selected journals, including journals in psychiatry, psychology, and pediatrics. To identify unpublished or ongoing trials, we will use methods that include reviewing trial registries and results databases (for example, ClinicalTrials.gov), pharmaceutical industry trial registries and results databases, regulatory agency online databases, regulatory agency submissions, litigation documents, and conferences abstracts, as well as contacting trialists and sponsors for unpublished or ongoing trials [37
Identification of eligible studies
Eligible studies will include children and adolescents aged 6 to 18 years. We will include studies conducted in general medicine clinics, schools, and community settings. Studies including medically ill children are eligible. Studies of college and university populations will be excluded, as college and university students have a different pathway to mental health treatment than adolescents under their parents' care.
Key Question#1(accuracy of screening tools): Studies on the accuracy of screening tools will be included if they compared a screening instrument with a valid criterion standard, defined as a DSM diagnosis of MDD or an ICD diagnosis of depressive episode based on a validated diagnostic interview procedure, and if they reported data allowing determination of sensitivity and specificity, positive predictive value, and negative predictive value. Examples of validated psychiatric interviews that have been used in assessments of children or adolescents include, but are not limited to, the Composite International Diagnostic Interview, the Diagnostic Interview for Children and Adolescents, the Revised Diagnostic Interview Schedule for Children, and the Schedule for Affective Disorders and Schizophrenia for School-Aged Children. Studies that assess broader diagnostic categories, such as any depressive disorder or dysthymia, will be included only if they report data for MDD separately. Studies must administer the screening tool and the diagnostic assessment within 2 weeks of each other to be eligible. Studies in which only parent- or teacher-completed measures, but no child or adolescent self-report measures, are compared to a diagnosis of MDD will be excluded. Studies will be excluded if a cutoff score above a threshold on a depression screening tool is used as an eligibility criterion to receive assessment for MDD with a validated structured interview. Studies conducted in high-risk populations where many or most children and adolescents may have a psychiatric disorder, such as in psychiatric or youth-protection settings, will be excluded.
Key Question#2(effects of treatment): Studies on treatment will include RCTs with placebo or usual care controls that evaluated pharmacological, psychotherapeutic, or other interventions that would typically be available to children or adolescents in general pediatric settings or primary-care and family medicine settings as treatment for depression as diagnosed with a validated psychiatric interview and DSM criteria for MDD or ICD criteria for a depressive episode. Specifically, eligible interventions will include SSRIs (fluoxetine, sertraline, paroxetine, citalopram, escitalopram, and fluvoxamine), SNRIs (venlafaxine, duloxetine, desvenlafaxine), other non-SSRIs prescribed for children or adolescents with depression (bupropion, mirtazapine, trazodone), psychotherapy, educational interventions, and exercise. We will require, as an eligibility criterion, a DSM diagnosis of MDD or ICD diagnosis of depressive episode based on a valid diagnostic interview because unassisted clinician diagnoses have poor reliability and because a large proportion of patients scoring above cutoffs on self-report questionnaires do not have MDD. Head-to-head trials of different interventions without a comparison to usual care or placebo are not eligible.
Key Question#3(effects of screening): Eligible studies will be RCTs that compared depression outcomes between children or adolescents who underwent depression screening and those who did not. Studies in which comprehensive depression care programs were provided to children or adolescents with depression as part of the screening program will be included if children and adolescents in the unscreened group could also access these treatment programs if identified as depressed by means other than screening. Otherwise, it will not be possible to disaggregate effects of screening from effects of providing comprehensive depression management. Changes in rates of depression recognition and treatment will be noted, but not included as depression outcomes. This is because increased treatment is not a benefit and, if improved depression outcomes are not obtained, exposes children and adolescents to costs and potential harms of treatment without benefit. Screening is defined per the United Kingdom National Screening Committee’s definition. Thus, eligible screening trials had to include a case identification strategy based on an a priori defined cutoff score on a depression screening tool to make decisions regarding further assessment or treatment.
Key Question#4(harms): Studies that document harms from depression treatment or screening will include RCTs, as well as observational studies.
Two investigators will independently review titles and abstracts for eligibility with full-text review of articles identified as potentially eligible by one or both. Disagreements after full-text review will be resolved by consensus. Chance-corrected agreement will be assessed with Cohen’s κ. See Additional file 3
for the coding manual.
Data extraction and outcomes
Two reviewers will independently extract relevant data from each eligible article and enter the data directly into formatted Excel spreadsheets. Entries will be compared for accuracy and any discrepancies will be resolved by consensus. Authors of original studies will be contacted if necessary to clarify inconsistencies in reported results. See Additional file 4
for variables included in data extraction templates.
Key Question#1(accuracy of screening tools): Outcomes reported will include sensitivity, specificity, positive predictive value, and negative predictive value, based on child or adolescent self-report measures.
Key Questions#2(effects of treatment)and#3(effects of screening): The primary outcome will be standardized mean difference effect size based on a continuous measure of depressive symptoms. When multiple depression outcomes are reported, primary outcomes as identified in each study will be given highest priority. Then observer-rated scales will be prioritized over self-report measures. Outcomes will be reported as intent-to-treat analyses where possible and will be prioritized over those presented for completers only.
Key Question#4(harms): Outcome reporting will be based on harms identified from the review, but we expect the main potential harm to be suicidal ideation.
Appraisal of risk of bias
For Key Question #1 (accuracy of screening tools), we will use the Quality Assessment of Diagnostic Accuracy Studies-2 tool [38
]. This tool incorporates assessments of risk of bias across four core domains: patient selection, the index test, the reference standard, and the flow and timing of assessments. For Key Questions #2 (effects of treatment) and #3 (effects of screening), we will use the Cochrane Risk of Bias tool [39
], which includes assessments of six possible sources of bias related to randomization sequence generation; allocation concealment; blinding of participants, personnel, and outcome assessors; incomplete outcome data; selective outcome data; and other sources of bias. In addition to these domains, based on recent recommendations regarding potential bias due to financial conflicts of interest [40
], we will code two further domains, for pharmaceutical industry funding, and author-industry financial ties or employment by industry.
Synthesis of included studies
The decision as to whether data can be synthesized using meta-analysis will be determined based on search results. Based on the studies included in the 2009 United States AHRQ systematic review [1
] and a rough preliminary search of MEDLINE conducted in October 2011, we expect that there will be substantial heterogeneity between diagnostic accuracy studies with respect to assessment timing, criterion standards, screening instruments, and scoring thresholds. Most studies have used population-specific cutoff thresholds based on receiver operator characteristic curves, which yields overly optimistic estimates of screening accuracy that do not replicate consistently [42
]. Similarly, for studies of MDD treatment, there appears to be substantial heterogeneity in the nature of interventions, outcome measures, and length of follow-up. Thus, due to clinical heterogeneity, it appears that data pooling will likely not be appropriate and that a systematic review, rather than a meta-analysis, will be conducted. If, following the full literature review, we find that meta-analytic data pooling is feasible, then we will use appropriate models that have been used previously in meta-analyses of diagnostic accuracy [43
], screening effectiveness [44
], and treatment outcomes from depression care interventions [45