Overview of reviews in child health: evidence synthesis and the knowledge base for a specific population

Authors

  • Denise Thomson,

    Corresponding author
    1. Cochrane Child Health Field, Department of Pediatrics, University of Alberta, Edmonton, Alberta, Canada
    • Coordinator, Cochrane Child Health Field, Alberta Research Centre for Health Evidence (ARCHE), Department of Pediatrics, University of Alberta, 4-476 Edmonton Clinic Health Academy, 11 405 87 Ave NW, Edmonton AB T6G 1C9, Canada.
    Search for more papers by this author
  • Michelle Foisy,

    1. Cochrane Child Health Field, Department of Pediatrics, University of Alberta, Edmonton, Alberta, Canada
    2. Department of Psychology, University of British Columbia, Vancouver, British Columbia, Canada
    Search for more papers by this author
  • Marta Oleszczuk,

    1. Cochrane Child Health Field, Department of Pediatrics, University of Alberta, Edmonton, Alberta, Canada
    Search for more papers by this author
  • Aireen Wingert,

    1. Cochrane Child Health Field, Department of Pediatrics, University of Alberta, Edmonton, Alberta, Canada
    Search for more papers by this author
  • Annabritt Chisholm,

    1. Alberta Research Centre for Health Evidence, Department of Pediatrics, University of Alberta, Edmonton, Alberta, Canada
    Search for more papers by this author
  • Lisa Hartling

    1. Cochrane Child Health Field, Department of Pediatrics, University of Alberta, Edmonton, Alberta, Canada
    2. Alberta Research Centre for Health Evidence, Department of Pediatrics, University of Alberta, Edmonton, Alberta, Canada
    Search for more papers by this author

Abstract

Background

Overviews of reviews are an evolving form of evidence synthesis. The Cochrane Child Health Field has been producing overviews since 2006, during which time the methods that have been used have changed, both due to the development of guidance within The Cochrane Collaboration and to the decisions made by individual author teams. This paper studies the first 29 overviews published in EBCH.

Objectives

To describe some aspects of the approaches taken in EBCH overviews to producing evidence syntheses relevant to the healthcare needs of children; to highlight the contribution that overviews can make to the knowledge base for treatment for a particular population.

Methods

Data was extracted on: whether the overview included systematic review (SR) data only, or also data from individual trials not present in the included SRs; name(s) of the Cochrane Review Group (CRG) that prepared the included SRs; topics of the overviews as compared to the topics of the included reviews; age-subgroup analyses presented in the overviews.

Results

In 23 overviews, all published in 2012, the authors included trial data as well as SR data; two overviews addressed conditions not explicitly addressed by the included reviews; three overviews included pre-specified age-subgroup analyses.

Discussion

The aim of clinical relevance has been achieved by means such as: drawing from reviews produced by multiple CRGs; using SR evidence to explore clinically relevant topics that may not match exactly with the topics covered by the SRs; ensuring that the evidence in overviews is as up to date as possible by redoing searches and including trials not incorporated in the included SRs; and, where permitted by the data, using age-subgroup analyses to present the data in a way which matches the stages of childhood development.

Conclusion

Overview authors are dependent on the nature of the data and methods reported in the included SRs. This suggests a need for further study about how SRs could be conducted in order to facilitate the conduct of overviews. Copyright © 2013 The Cochrane Collaboration. Published by John Wiley & Sons, Ltd. The Cochrane Collaboration

Background

The rationale for evidence synthesis is that practice and policy decisions should be based on the findings of a body of knowledge rather than a single study1. Systematic reviews (SRs) are a well-established methodology for synthesizing research data1. The Cochrane Collaboration was established in 1993 to support improved decision-making in healthcare through the production and dissemination of high-quality SRs. The Collaboration comprises 53 Cochrane Review Groups (CRGs), each producing reviews addressing a particular disease state (e.g. gynaecological cancer) or a physiological system (e.g. renal). Currently, the Cochrane Database of Systematic Reviews contain 5258 completed SRs2. The Cochrane Child Health Field (the Field) has identified that 1481 of these are relevant to child health as they contain data on those in the 0–18 years age range, either exclusively or in combination with data on adults.

In recent years, the concept of synthesizing data from two or more SRs has gained traction in healthcare research3–5. These syntheses have been given a number of different names, such as ‘umbrella reviews’ or ‘review of systematic reviews’. Within the Cochrane Collaboration they are referred to as ‘overviews of reviews’ or the shortened term, ‘overviews’3–5. The motivation behind the production of overviews within the Cochrane Collaboration is the recognition that clinicians and other decision makers need a ‘friendly front end’ to Cochrane evidence when deciding how to address a particular clinical issue5. Discussion within the Collaboration around the role, scope and format of overviews is ongoing6.

The Child Health Field of The Cochrane Collaboration (the Field) has been producing overviews of reviews since 2006. The ongoing production and dissemination of overviews is part of the Field's mandate of facilitating the child-relevant application of evidence in practice and policy. The overviews have been published in the Field's journal, Evidence-Based Child Health: A Cochrane Review Journal (EBCH). The Field's overviews are intended to collate evidence relevant to children and youth, as the evidence needs of this population are often distinct from those of adults7–9. Table 1 lists the publication dates and topics of all 29 EBCH overviews published at date of writing. Of these, 25 were new overviews and four were updates of previous overviews; updates are carried out when the Cochrane SRs included in the original overviews are updated, and/or new SRs relevant to the overview's topic are published.

Table 1. Publication date and topics of all EBCH overviews*
Issue-YearTopic
  • *

    In 2011, EBCH moved from publishing four to six issues per year. Not every issue of the journal contains an overview of reviews.

1-2006Nocturnal enuresis
2-2006Nephrotic syndrome
3-2006Chronic cough
4-2006Bronchiolitis
1-2007Mother to child transmission of HIV
2-2007Dietary prevention of allergic disease and food hypersensitivity
3-2007Trachoma
4-2007Sickle cell disease
1-2008Autism spectrum disorder
2-2008Procedural pain
3-2008Leukotriene receptor antagonists for children with asthma
4-2008Long-acting beta-agonist treatment for childhood asthma
1-2009Recurrent abdominal pain
2-2009Acute otitis media
3-2009Community acquired pneumonia
4-2009Major depression
2-2010Childhood and adolescent anxiety disorders
3-2010Chronic cough (update of 3–2006 overview)
4-2010Croup
1-2011Bronchiolitis (update of 4–2006 overview)
2-2011Non-pharmacological treatments for Attention Deficit Hyperactivity Disorder
3-2011Sore throat
4-2011Chronic abdominal pain
5-2011Prevention of eczema in infants and children
6-2011Promotion of bicycle helmet use
2-2012Acute otitis media (update of 2–2009 overview)
4-2012Patent ductus arteriosus
5-2012Procedural pain (update of 2–2008 overview)
6-2012Safety of regular long-acting beta2-agonists

A recent analysis of the first 24 overviews published in EBCH found that they include a median of five SRs (range 2–11) and 35 primary studies (range 4–230)3. The EBCH overviews describe the evidence from the included SRs in a qualitative, narrative manner; no further quantitative analysis is carried out3. The process by which they are produced, and the evolution of the methods that have been used, have been described elsewhere4. They have recently been compared with overviews published in other sources3.

The manner in which the evidence is presented in the EBCH overviews has changed considerably over time. Tables 2–4 are drawn from three overviews published in 2006, 2011 and 2012, respectively. Table 2 reproduces an evidence table from the first EBCH overview, on nocturnal enuresis12; here the presentation is organized by outcome, in this case, ‘Failure to achieve 14 consecutive dry nights during treatment’. Table 3 presents an evidence table from the 2011 bronchiolitis update; the authors of this overview chose to present the data in population-based subgroups (outpatients, inpatients and intensive care unit patients). Table 4 reproduces a Summary of Findings (SoF) table from the 2012 acute otitis media update; this was prepared according to the specific guidance for Summary of Findings tables in The Cochrane Handbook13. The change over time in the presentation of evidence is owing, on one level, to the development of methodological guidance within the Cochrane Collaboration for the conduct of overviews, and, at another level, to the decisions made by author teams on individual overviews. To date, no user testing has been done of the various manners of presenting data in overviews; this would be an important avenue of future research.

Table 2. Example of presentation of data in an EBCH overview: data for the outcome, ‘Failure to achieve 14 consecutive dry nights during treatment’, from the 2006 nocturnal enuresis overview12
ComparisonNumber of subjects (studies)Average control group rate*Relative risk (95% CI)% Change in events
  • *

    Average control group rate: baseline risk of enuresis.

  • Relative risk: likelihood of developing enuresis in the intervention group relative to those not receiving treatment or receiving placebo;

  • % change in events: the difference between the average control group rate and the average intervention group rate.

Enuresis alarms versus no treatment576 (13)96%0.38 (0.33, 0.45)62%
Enuresis alarm versus dry bed training108 (3)37%1.33 (0.79, 2.24)− 13%
Enuresis alarm versus desmopressin243 (3)41%0.71 (0.50, 0.99)12%
Enuresis alarm versus imipramine208 (3)80%0.73 (0.61, 0.88)22%
Alarm versus cognitive treatment/psychotherapy/counselling155 (3)63%0.68 (0.52, 0.90)29%
Cognitive treatment/psychotherapy/counselling versus control95 (3)94%0.69 (0.55, 0.85)31%
Dry bed training versus no treatment60 (2)93%0.82 (0.66, 1.02)16%
Dry bed training and alarm versus no treatment143 (4)94%0.17 (0.11, 0.28)83%
20 mcg desmopressin versus placebo288 (5)99%0.84 (0.79, 0.91)18%
40 mcg desmopressin versus placebo463 (6)86%0.81 (0.74, 0.88)16%
60 mcg desmopressin versus placebo165 (2)100%0.94 (0.89, 0.99)6%
Diclofenac versus placebo92 (2)93%0.52 (0.38, 0.70)43%
Imipramine versus placebo813 (11)95%0.77 (0.72, 0.83)20%
Table 3. Example of presentation of data in an EBCH overview: the table ‘Inpatient outcomes’ from the bronchiolitis update10
OutcomeComparisonNumber of subjects (studies)Measure of effect (95% CI)I2Quality of evidence (GRADE)
  • a

    Significantly favours epinephrine;

  • b

    Significantly favours 3% hypertonic saline;

  • c

    Significantly favours chest physiotherapy;

  • d

    I2 not estimable; CI, confidence interval; GRADE, Grading of Recommendations Assessment, Development and Evaluation; MD, mean difference; RR, risk ratio; SMD, standardized mean difference; N/A, not applicable; OR, odds ratio.

Length of stayGlucocorticoid versus placebo633 (8)MD: − 0.18 (−0.39, 0.04)16%High
 Bronchodilator versus placebo349 (6)MD: 0.06 (−0.27, 0.39)0%Moderate
 Epinephrine versus placebo292 (2)MD: − 0.35 (−0.87, 0.17)0%Moderate
 Epinephrine versus bronchodilator261 (4)MD: − 0.28(−0.46, − 0.09)a0%Moderate
 3% hypertonic saline versus 0.9% saline282 (4)MD: − 1.16(−1.55, − 0.77)b0%Moderate
 Chest physiotherapy versus standard care or other drainage/breathing technique172 (3)MD: 0.07 (−0.58, 0.73)0%Low
Re-admissions between two days to four monthsGlucocorticoid versus placebo359 (3)RR: 1.04 (0.12, 8.72)66%Low
 Inhaled corticosteroid versus placebo309 (4)RR: 1.15 (0.60, 2.22)45%Moderate
 Epinephrine versus placebo192 (2)RR: 0.29 (0.05, 1.86)0%Low
Re-admissions between three months and 1 yearInhaled corticosteroid versus placebo358 (5)RR: 1.05 (0.63, 1.75)29%Moderate
Clinical score at 60 minutes3248 (4)SMD: − 0.79(−1.45, − 0.13)a79%Low
 Epinephrine versus placebo232 (2)SMD: − 0.04(−0.49, 0.40)46%Moderate
Clinical score at 120 minutesEpinephrine versus bronchodilator140 (1)SMD: − 0.52(−0.86, − 0.18)aLow
Clinical score at one to three daysGlucocorticoid versus placebo113 (4)SMD: − 0.74(−1.48, 0.01)70%Low
 3% hypertonic saline versus 0.9% saline183 (3)SMD: − 0.84(−1.39, − 0.30)b66%Low
 Chest physiotherapy versus standard care or other drainage/breathing technique87 (1)SMD: − 0.55(−0.98, − 0.12)cLow
Clinical score at 3–10 daysChest physiotherapy versus standard care or other drainage/breathing technique91 (2)SMD: − 0.14(−0.81, 0.53)59%Low
 3% hypertonic saline versus 0.9% saline156 (3)SMD: − 1.08(−2.47, 0.31)93%Moderate
Table 4. Example of presentation of data in an EBCH overview: the table ‘Summary of findings: Antibiotics versus placebo’ from the acute otitis media update11
 Illustrative comparative risks* (95% CI)    
OutcomesAssumed riskCorresponding riskRelative effect (95% CI)I2No of participants (studies)NNT (95% CI)
  • *

    Mean baseline risk in control group was used to calculate assumed risk. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI); CI, confidence interval; NNT, number needed to treat; OR, odds ratio; TM, tympanocentesis.

Pain at 24 h (all studies)40/10036/100OR 0.852%1229 
  (31–41)(0.67–1.07) (5) 
Pain at 24 h (excluding studies with41/10036/100OR 0.820%1145 
TM) (31–42)(0.65–1.04) (4) 
Pain at 2–7 days (all studies)22/10015/100OR 0.610%279114
  (12–18)(0.49–0.75) (10)(10–22)
Pain at 2–7 days (excluding studies22/10014/100OR 0.590%261913
with TM) (12–17)(0.48–0.74) (8)(10–22)
Treatment failure (excluding studies26/10012/100OR 0.4066%28038
with TM) (8–17)(0.27–0.60) (8)(6–12)
Treatment failure (excluding studies47/10018/100OR 0.240%6084
in which < 75% of children had a bulging TM) (13–23)(0.16–0.34) (2)(3–5)
Late recurrence21/10019/100OR 0.920%2153 
  (16–23)(0.74–1.14) (6) 
Adverse events (vomiting, diarrhoea,18/10035/100OR 2.3585%20117
or rash) (25–52)(1.14–4.84) (7)(4–49)

The Cochrane Handbook for Systematic Reviews of Interventions (Cochrane Handbook) advises that the primary aim of Cochrane overviews is to summarise evidence from Cochrane SRs, and therefore encourages authors to include Cochrane SRs only5. With one exception, the authors of EBCH overviews have followed this guidance. The exception was the 2011 overview on the prevention of eczema, for which the authors searched the 2010 United Kingdom National Health Service Evidence Skin Disorders database which returned both Cochrane and non-Cochrane SRs14. This database was built by the Cochrane Skin Group by searching for all available SRs examining all potential interventions for the prevention of eczema; the inclusion criterion for this database was that SRs had to meet the Cochrane Collaboration's definition of an SR (‘reviews of clearly formulated questions that use systemic and explicit methods to identify, select and critically appraise relevant research, and to collect and analyse data from the studies included in the reviews’)14, 15. The availability of this comprehensive database obviated the main practical impediment to searching for non-Cochrane reviews, which would be that searching and appraising the non-Cochrane literature entails much more time than is feasible with the existing resource constraints of EBCH overview production.

This paper presents an analysis of how the approach in EBCH overviews has been developed to analyse and present evidence that is relevant to the needs of a particular population, and to highlight the contribution that these overviews have made to the knowledge base in paediatric care.

Objectives

  • To describe some aspects of the approaches taken in EBCH overviews to producing evidence syntheses relevant to the healthcare needs of children.

  • To highlight the contribution that overviews can make to the knowledge base for treatment for a particular population.

Methods

The following information about each of the EBCH overviews was extracted:

  • Whether the overview incorporated SR data only, or also data from individual trials not present in the included SRs.

  • Name(s) of the CRG(s) that prepared the included SRs.

  • Topics of the overviews as compared to the topics of the included reviews.

  • Age-subgroup analyses presented in the overviews.

This data was extracted by one author and verified by a second author.

Results

Including clinical trials in overviews

From 2006 to 2011, the overview authors did not incorporate any data from clinical trials that were not in the included SRs, even if the SRs were slightly out of date. The only exception to this was a 2009 overview on acute otitis media where the authors of this overview knew of an important trial which had been published since the SRs had been completed. The data from this trial was not included in the analyses, but the trial's findings were incorporated into the discussion. In the 2012 overviews, three author teams chose an approach in which, for any included SR that was more than 2 years old, the search strategies outlined in the original SR were rerun in Medline and subsequent data from trials that met the inclusion/exclusion criteria of the SR were incorporated into the overview. This change to methods affected the acute otitis media update, the patent ductus arteriosus overview and the procedural pain update; these three overviews included 2, 0 and 13 new trials, respectively.

Review groups producing the included SRs

Table 5 shows the list of CRGs that produced the reviews included in the overviews. Five overviews (17.24%) included reviews produced by more than one CRG. Of these, four included reviews from two CRGs and one included reviews from three CRGs (Table 6).

Table 5. CRGs which produced the reviews included in EBCH overviews*
  • *

    The number of CRGs exceeds the number of overviews as some overviews include reviews from more than one CRG.

Acute Respiratory Infections7
Airways5
Developmental, Psychosocial and Learning Problems4
Neonatal4
Pain, Palliative and Supportive Care4
Depression, Anxiety and Neurosis2
Pregnancy and Childbirth2
Cystic Fibrosis and Genetic Disorders1
Eyes and Vision1
Heart1
HIV/AIDS1
Incontinence1
Injuries1
Renal1
Total35
Table 6. Overviews which included reviews from more than one CRG
Overview topicPublication dateCRGs producing the included Cochrane SRs
3 CRGs  
Dietary prevention of allergic disease and food hypersensitivity2 − 2007Airways; Neonatal; Pregnancy and Childbirth
2 CRGs  
Sickle cell disease4-2007Cystic Fibrosis and Genetic Disorders; Pain, Palliative and Supportive Care
Procedural pain2-2008Neonatal; Pain, Palliative and Supportive Care
Sore throat3-2011Acute Respiratory Infections; Heart
Chronic abdominal pain4-2011Developmental, Psychosocial and Learning Problems; Pain, Palliative and Supportive Care
Prevention of eczema in infants and children5-2011Neonatal; Pregnancy and Childbirth

Overview topics differing from the topics of the included reviews

There were two overviews that addressed conditions not explicitly addressed by the included reviews. The overview on treatments for sore throat drew from data in reviews on pharyngitis, laryngitis or tonsillitis, various acute respiratory tract infections, laboratory confirmed group A beta-haemolytic streptococcus infections, infectious mononucleosis and rheumatic fever16. The overview on interventions for chronic abdominal pain incorporated data from reviews on children with abdominal pain, functional gastrointestinal disorders, recurrent abdominal pain and irritable bowel syndrome17. Both overviews incorporated data from reviews produced by two CRGs: the overview on sore throat incorporated reviews from the Acute Respiratory Infections and Heart Groups, while the overview on chronic abdominal pain included reviews from the Developmental, Psychosocial and Learning Problems and Pain, Palliative and Supportive Care Groups.

Age-subgroup analyses

Three of the overviews—the 2008 procedural pain overview and its 2012 update, and the 2011 prevention of eczema overview—included pre-specified age-subgroup analyses. The 2008 procedural pain overview presented results for neonates and older children18. The 2012 update excluded neonates and presented results for older infants (0–48 months) and older children (3 months to 19 years); the overlap in age ranges was due to the nature of the data presented in the included SRs19. While the authors excluded reviews which exclusively focused on preterm infants and neonates, several trials from the included reviews examined a mixed population (neonates and older infants) where it was not possible to distinguish trial level data for specific age groups at an overview level. The overview authors categorized the data into their respective age groups according to how they were pre-defined in the reviews. The prevention of eczema overview analysed the effects of interventions in infants (0–2 years) and children (2–18 years)14.

Discussion

Including clinical trials in overviews

The viability of particular overview topics is limited by several factors, including the availability of up-to-date SRs20. During the 2006–2011 period of producing EBCH overview, topics for which the range of available SRs included ones which were out of date would be discarded. As mentioned above, for three of the overviews completed in 2012, the search strategies from any out-of-date SRs were rerun in Medline. This was a practice undertaken to expand the scope of available topics while ensuring that the completed overviews were up-to-date and clinically meaningful. However, in practice, decisions on inclusion and exclusion sometimes proved to be quite complex and difficult because authors aimed to match the criteria pre-specified in the original SRs. In circumstances where an SR examines a very specific intervention, overview authors may encounter complexity in applying the stringent inclusion/exclusion criteria of the included SR to new trials which examine clinically meaningful comparisons that do not necessarily match those of the review. For example, in one review included in the 2012 procedural pain overview, the authors compared two specific pharmacological treatments (eutectic mixture of local anesthetics (EMLA) versus amethocaine) for two specified types of procedures (venipuncture and intravenous cannulation). Some trials from the updated search contained a similar intervention (e.g. lidocaine), comparison (e.g. EMLA versus music), procedure (e.g. urinary catheterization) or outcome (procedural sedation). None of these trials were included in the overview because they were not a clear match to the original review's inclusion/exclusion criteria. Overview authors intending to conduct searches for trials should consider carefully how straightforward the inclusion/exclusion criteria will be to apply and how they will handle studies that may be relevant to the overview topic but not specific to the original SRs (e.g. incorporate them into the analyses or include them in the discussion section).

Capturing data from reviews produced by more than one CRG

The policies of the Cochrane Collaboration are that only CRGs can register and carry out the overviews that are published on the Cochrane Database of Systematic Reviews; consequently, these overviews draw on reviews from one CRG only. The Child Health Field, whose interest is in the health of a specific population and whose work is not restricted to the scope of a particular disease, is able to conduct overviews which draw from the output of several groups. Five of the 29 overviews included in this analysis included data from more than one CRG, demonstrating that topics selected for relevance to a particular population can cut across condition-specific categorizations, such as those typically used to frame the scope of individual CRGs. This finding has key relevance for the Cochrane Collaboration, as those who wish to address the needs of a population must be able to work across these boundaries. There may be a benefit for the Collaboration, in the interest of publishing the broadest possible range of relevant evidence on The Cochrane Database of Systematic Reviews, to develop procedures for registering overviews that involve more than one CRG.

Overview topics differing from the topic of the included SRs

The choices made by overview authors of topics or approaches to classifying data can enable them to use the overview format to carry out unique syntheses that make a new contribution21. This function of overviews can be described as asking a different question of the evidence than that which was posed by the included SRs4, 20, 22. For example, the topics of sore throat and chronic abdominal pain were not explicitly covered by any Cochrane review; however, both topics are clinically relevant to paediatric practice and were therefore deemed by the respective author teams to be important topics for overviews. In both cases, clinicians often do not know the precise diagnosis of the patient prior to deciding whether and how to treat the presenting symptoms, which makes it very difficult to determine treatment. A systematic approach to assessing included reviews was required in both cases in order to create a clear and concise overview of reviews on a specific clinical symptom independent of the underlying disorder, mimicking the real-life situations that clinicians face on a daily basis.

Both the sore throat and the chronic abdominal pain overviews presented specific complexities. One of these was that the search strategies required to find eligible SRs that were quite broad and complex, as the clinical entities being searched for were present in varying degrees in a number of different SRs but were not captured perfectly by any one Sr. For example, many children and youth present with chronic abdominal pain, for which there can be a wide range of aetiologies. This range was reflected in the fact that the author team on this overview included a general paediatrician, a psychiatrist, an emergency physician and two gastroenterologists17.

The search strategy for the sore throat overview proved particularly challenging, as the authors were searching for data on a specific symptom that was one part of the clinical presentation of multiple heterogeneous illnesses. The authors decided to begin by searching for SRs that explicitly focused on the symptom of sore throat. They then expanded the search to include reviews on disorders (i.e. streptococcal pharyngitis, the common cold, infectious mononucleosis) where sore throat is one of many presenting symptoms. However, this led to the problem of assessing over one hundred potentially relevant reviews for inclusion in the overview. The authors wanted to produce a clinically relevant overview that captured the full spectrum of illnesses related to sore throat, without losing sight of the overall purpose of the overview. In the end, the decision was made to include only those reviews that presented outcome data on the symptom of sore throat. Synthesizing outcome data on sore throat from reviews on multiple different disorders allowed the authors to arrive at evidence-based recommendations for the treatment of the specific symptom of interest. Excluding reviews on relevant disorders that did not provide outcome data on that symptom of interest was equally important, as this prevented the overview from deviating from its main purpose.

Age-subgroup analyses

Subgroup analyses entail splitting the data in a particular trial, SR or overview into smaller groupings, often with the goal of answering specific questions about particular patient groups or types of interventions23. Research on the conduct of SRs relevant to children has led to recommendations that SR authors carry out subgroup analyses by age, when appropriate, to take into account the range of physiological, biological and developmental differences not only between children and adults but also between children of varying ages7, 24.

However, a recent examination of age groups and age-subgroup analyses in both paediatric randomized controlled trials and meta-analyses showed large variability in the included age ranges25. The majority of the paediatric meta-analyses included in this study carried out age-subgroup analyses by combining patients from different studies; any differences found in this manner may have been confounded by differences between the included studies. Furthermore, all except one of the age-subgroup differences found in the included meta-analyses were not adequately powered to give a result that was beyond chance25. The mechanism for addressing these uncertainties is to encourage trialists to use consistent, clearly reported age groups24, 25.

Given the methodological difficulties of establishing appropriate age-subgroup analyses in SRs, one might expect further challenges while synthesizing data from multiple SRs, and these are illustrated by the experiences of the authors of the 2012 procedural pain overview. Some reviews contained data for neonates (part of the exclusion criteria for the overview) as well as older infants and children (part of the inclusion criteria for the overview). Wherever it was possible, the authors extracted only the data for the older infants and children. When there was mixed population within a trial, they included the trial in order to capture data for the population of interest.

Another situation in which doing age-subgroup analyses in overviews can be difficult is when overview authors are also running searches to consider new trials for inclusion. If these new trials have large age ranges, their inclusion can create additional complexity in the analysis and presentation of the data, particularly when considering interventions or populations of interest for which there is a paucity of data. Trial level data has not been subjected to synthesis by an SR; therefore, overview authors who wish to include new trials with mixed populations need to address several factors influencing how that data is categorized into age-subgroups. These may include considerations of whether individual trial authors conduct age-subgroup analyses, whether the data can be grouped according to a pre-defined age-subgroup, or whether it is possible to categorize the population range according to age-appropriate interventions or outcomes. Such complex determinations of age groupings may be conducted on a case-by-case basis tailored to the specific age ranges, interventions and outcomes of interest to the overview authors and its intended audience.

Conclusion

The purpose of overviews is to synthesise a large amount of data in one platform in order to support decision-making and reduce the burden on individual clinicians or policy makers to read and interpret the findings of several SRs. The decision to rerun searches in order to include the most up-to-date evidence possible is part of achieving this goal. In addition to this commonly accepted goal for overviews, we have argued here that they can make an important contribution to the knowledge base about treatment interventions for a particular population. We have discussed some practical ways in which authors of overviews can do this: selecting topics based on clinical relevance rather than the scope of a particular CRG or even, at times, the topics of the included SRs and, where permitted by the available data, presenting age-based variations in the findings.

However, it is evident from our discussion of the difficulties of updating reviews or creating age-based subgroups that overview authors are dependent on the nature of the data and methods reported in the included SRs. This suggests a need for further study about how SRs could be conducted in order to facilitate the conduct of overviews. Considerable work remains to be done on methods for overviews to determine how to address this limitation in order to provide the most relevant and valid information for evidence-based decision-making.

Ancillary