Child vs Adult Randomized Controlled Trials in Specialist Journals: A Citation Analysis of Trends, 1985-2005

Objective  To compare secular trends in the age representation of published randomized controlled trials (RCTs) in specialty journals during a period of 20 years.

Data Source  A validated electronic search strategy using Ovid MEDLINE was conducted to identify RCTs published in the years 1985 through 2005.

Study Selection  The publications retrieved were subdivided into age-specific groups: adults, children, both adults and children, and studies with no age group identified. Within 31 specialties, we chose up to 5 specialty journals and 5 pediatric specialty journals.

Main Outcome Measure  Number of RCTs targeting children compared with adults over time. Linear trends were identified using regression modeling, and an interaction term was included to compare rates of increase between age groups.

Results  A total of 174 unique journals with 43 326 unique RCTs with age-specific categorization were included. Adult RCTs increased by 90.5 RCTs per year (95% confidence interval [CI], 78-103), which was significantly higher than either pediatric RCTs, which rose by 16.9 RCTs per year (95% CI, 12-22) or RCTs involving both children and adults, which rose by 22.7 RCTs per year (95% CI, 10-35). Twenty four of 31 specialties (77%) demonstrated a greater rise in the number of published RCTs per year involving adults than those enrolling children.

Conclusion  Adult RCT publications are increasing at a faster rate than pediatric RCTs in almost all specialties.

Clinical trials have led to important improvements in pediatric outcomes.¹ However, there are logistical, financial, and ethical barriers to conducting trials involving children.² Despite financial and legislative incentives to promote conducting pediatric trials, there is evidence that the gap between the number of important pediatric and adult trials is widening. The number of adult trials published annually in high-impact general medical journals has doubled during the past 20 years, with virtually no change in pediatric trials.³ However, it is unclear if this trend reflects a paucity of contemporary pediatric randomized controlled trials (RCTs) or a redistribution of published trials into specialist and subspecialist journals. The objective of this study was to compare the secular trends in the publication of age-specific RCTs in high-impact specialty journals.

The unit of study was published RCTs abstracted from selected journals. Journals were sampled from the Institute for Scientific Information's Journal Citation Reports (JCR). Specialties were defined based on JCR categorization (n = 31 specialties). Up to 5 specialty journals (SJs) and pediatric specialty journals (PSJs) with the highest impact factor in 2005 using Garfield's criteria^4-6 per specialty were included. To focus the searches on journals that publish patient-oriented research, only journals with at least 1 RCT (see “Electronic Search Strategy” section) in the last 5 years (2001-2005) were included. The SJs included a few journals that were focused primarily on adult health specialties (eg, Menopause) but mainly journals without an age-specific focus (eg, Circulation). Journals principally aimed at pediatric health specialties were categorized as PSJs. To incorporate “generalist” journals (GJs) with a focus on a particular age group but no particular specialty defined, we also included the 5 highest-impact general internal medicine journals that published an RCT in 2005 from the “Medicine, General and Internal” category of JCR and included the 5 highest–impact factor general pediatrics journals from those in the “Pediatrics” category of JCR without any specialty categorizations that published at least 1 RCT in 2005. General medical journals (eg, New England Journal of Medicine, JAMA, The Lancet, British Medical Journal) were excluded, as these had been reported previously.³

Excluded were journals that failed to span 20 years, unless this was owing to a change in the journal's title. We also included journals that did not meet the time-span criterion if there were fewer than 5 journals in the specialty or pediatric specialty category that met RCT criteria. If 1 or more of the top 5 impact factor journals was excluded or if the journal was categorized into more than 1 specialty (eg, Contact Dermatitis is categorized as both allergy and dermatology), the journal(s) with the next-highest impact factor that met inclusion criteria were selected with the aim to include 5 unique SJs and 5 unique PSJs per category. If fewer than 5 journals in a category were found with JCR, further searches for relevant journals were performed using LocatorPlus (with a limit of “serial of periodical” using the keyword journal search).

An electronic search strategy of Ovid MEDLINE was conducted in 2007 to identify RCTs published from 1985 through 2005. The phase 1 modification of the highly sensitive search strategy for identifying RCTs described by Robinson and Dickersin⁷ was used; we have previously used this strategy to reliably identify RCTs when compared with hand searching.³ All comments, editorials, and letters were excluded. The publications retrieved were subdivided into age-specific groups: adults (aged >18 years), children (aged ≤18 years), both adults and children, and studies with no age group. Age-specific search strategies incorporated age-specific search terms and/or text words⁸ rather than conventional MEDLINE age indexing to maximize sensitivity and specificity. We used the Boolean operator “AND” to identify studies in which both pediatric and adult subjects were included. Table 1 summarizes an example of the search strategy applied for a specific year for 1 journal. All searches were performed by 1 member of the study team (U.S.), and the search was verified using a random sample of 1 unique journal per specialty for SJs and PSJs, and all GJs by another study author (N.M.).

Comparisons of linear trends in the total number of RCTs published over time in studies categorized as adult, pediatric, both adult and pediatric, and no age specified were modeled using linear regression. An interaction term was used to compare the slopes of the lines. Each analysis incorporated all RCTs published in the highest-impact SJs and GPJs. Separate analyses were performed for the all of specialties combined and for individual specialties. A sensitivity analysis was performed, assuming erroneous classification of studies with both adult and pediatric subjects (eg, studies recruiting adults 18 years or older with at least 1 person aged 18 years recruited could potentially be classified as both adult and pediatric). All P values were 2-tailed, and analyses were conducted using SPSS for Windows version 13 (SPSS Inc, Chicago, Illinois) and SAS version 9.1 (SAS Institute Inc, Cary, North Carolina). P ≤ .05 was considered statistically significant.

A total of 174 unique journals were included from 30 specialties plus GJs for a total of 31 unique specialties (Figure). Journal categorizations by specialty can be found in Table 2. Of these journals, 131 (75%) were SJs and 43 (25%) were PSJs. Included among these were 10 GJs, 5 from general internal medicine and 5 from general pediatrics. Of the SJs, there were no eligible journals for the specialty of genetics and heredity. Therefore, this specialty was not included in our analysis. In PSJs, there were no eligible journals for multiple specialties: genetics and heredity, microbiology, and toxicology. A total of 17 journals were categorized into more than 1 specialty, of which 8 were SJs and 9 were PSJs.

The search strategy of all journal categories resulted in 47 232 unique RCTs, of which 43 326 (92%) had an age-specific categorization; of these, 28 989 (67%) had only adults enrolled; 5898 (14%), only children, and the remaining 8439 (19%), both adults and children (Table 3). Overall, adult RCTs increased by an average of 90.5 RCTs per year (95% confidence interval [CI], 78-103), which was significantly higher (P < .001) than pediatric RCTs, which rose by 16.9 RCTs per year (95% CI, 12-22) or RCTs involving both children and adults, which rose by 22.7 RCTs per year (95% CI, 10-35) (eFigure). The difference in slopes between adult and pediatric RCTs persisted (P < .001) after sensitivity analyses were performed that assumed that RCTs involving both children and adults were erroneously classified and were truly adult RCTs.

Table 4 summarizes RCT distribution and secular trends over time by specialty and age group. The number of RCTs varied across the specialties, ranging from 38 (microbiology) to 4210 (GJs). Twenty-four of 31 specialties (77%) demonstrated a greater rise (P < .05 for all specialties) in the number of published RCTs per year involving adults than those enrolling children. A greater rise (P = .001) in RCTs per year in children compared with adults was found in 2 specialties, immunology and tropical medicine. No statistically significant difference in secular trends between adult and childhood age groups was found for allergy, behavioral sciences, GJs, microbiology, or toxicology. Twenty of 23 specialties (87%) with more than 1000 RCTs published in the last 20 years demonstrated a greater rise (all P < .005) in RCTs per year involving adults compared with children; of these specialties, only allergy and GJs had a greater rise in the number of pediatric RCTs than adult RCTs, although these difference were not statistically significant (P = .29 and P = .63, respectively).

To our knowledge, this is the largest study to investigate the secular trends in RCT publications between pediatric and adult age groups and the first to systematically investigate these trends in individual specialties. Our study found that RCTs published in specialty journals that involve adults are increasing at approximately 4 times the rate of those that involve children during the past 20 years. Only 2 specialties with few published RCTs, immunology and tropical medicine, showed a statistically significantly faster increase in pediatric RCTs than adult trials. Almost all specialties with frequent publications of RCTs, with the notable exceptions of allergy and GJs, also demonstrated a faster increase in adult RCTs compared with pediatric RCTs.

Other authors have lamented the paucity in the number of high-quality trials that involve children in general^1,9,10 and in specific subpopulations of children such as those with cystic fibrosis¹¹ or rheumatologic conditions.¹² A recent review of selected GJs during a 3-month period found that studies recruiting adults were almost 3 times more likely to be RCTs than those recruiting children; studies using cross-sectional designs were less than half as likely to involve adults than children.¹³ In contrast, we found no statistically significant difference in the rate of increase in RCTs involving children and those involving adults in the GJs selected in this review. The difference is likely owing to a longer time horizon, a wider number of journals searched, and a focus solely on RCTs from general pediatrics and general internal medicine, as opposed to general medical journals in our analysis. It is possible that there has been a distributional change over time whereby pediatric RCTs involving children of interest to a generalist audience are more likely to be published in general pediatric journals (eg, Pediatrics) rather than general medical journals (eg, New England Journal of Medicine), and adult RCTs are now less likely to be published in general internal medicine journals (eg, Archives of Internal Medicine) and more likely to be published in general medical journals. This is supported by our recent study that demonstrated a widening of the gap between RCTs involving adults and children published in general medical journals.³

If RCTs are considered the preferred study design for assessing interventions, our findings have implications for the quality of evidence-based care delivered to children. It seems that the gap in available evidence informing child and adult health is widening in just about every subspecialty; hence, more funding and incentives need to be provided for clinicians and the industry to conduct trials specifically in children. This thinking lends support to legislative efforts in the United States¹⁴ and the European Union¹⁵ designed to promote drug trials involving children by extending market exclusivity to pharmaceutical companies that conduct pediatric trials. Unfortunately, despite some success with these initiatives,¹⁶ market forces may have led to a disproportionate study of drugs with large adult market sales (eg, lipid-lowering agents) and have therefore underrepresented certain age groups (eg, neonates) and rare diseases.¹⁷ Another strategy has been to initiate consolidated research consortiums like the Medicines for Children Research Network in the United Kingdom¹⁸ to facilitate pediatric trials. Our study did not assess trials that have been launched but not yet published. It will likely take a few years to assess the effect, if any, of these policy-level interventions on the publication of pediatric RCTs.

An alternative interpretation of the widening gap is that, despite accepted wisdom that the RCT is the criterion standard of research design for assessing the safety and efficacy of interventions, there are important child-relevant outcomes at the level of the child, family, community, and society that are best captured using other epidemiological methods (eg, observational studies), and/or qualitative study designs.¹³ An important limitation of our study is that we did not investigate secular trends in these other relevant study designs or in the quality of the studies captured in our electronic search. Other authors have described inadequate quality of reported case-control studies published in pediatric journals.¹⁹ Another limitation is that by using the JCR classification of specialties, we may have missed important secular trends in specialties not captured using this taxonomy where there has likely been an important increase in the number of trials published (eg, neonatology), or in journals not indexed in MEDLINE.

In conclusion, the gap between pediatric and adult RCTs seems to be widening in almost every major clinical specialty. The logistical, financial, and ethical barriers to conducting trials in children have been well reported.²⁰ Future studies focusing on the effects of recent legislation to promote pediatric RCTs, observational study designs, and the effects of these trends in terms of the quality of care provided to children will be helpful.

Correspondence: Eyal Cohen, MD, MSc, FRCPC, The Hospital for Sick Children, 555 University Ave, Toronto, ON, Canada M5G 1X8 (eyal.cohen@sickkids.ca).

Accepted for Publication: June 16, 2009.

Author Contributions: Dr Cohen had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Cohen, Goldman, Uleryk, Atenafu, and Parkin. Acquisition of data: Cohen, Ragone, Uleryk, Siddiqui, and Mahmoud. Analysis and interpretation of data: Cohen, Ragone, Atenafu, Mahmoud, and Parkin. Drafting of the manuscript: Cohen, Ragone, Uleryk, Siddiqui, and Mahmoud. Critical revision of the manuscript for important intellectual content: Cohen, Goldman, Uleryk, Atenafu, Mahmoud, and Parkin. Statistical analysis: Cohen, Atenafu, and Parkin. Obtained funding: Cohen. Administrative, technical, and material support: Cohen, Goldman, Ragone, Uleryk, and Mahmoud. Study supervision: Goldman and Parkin.

Financial Disclosure: None reported.

Funding/Support: The Pediatric Outcomes Research Team is supported by a grant from the Hospital for Sick Children Foundation.

Role of the Sponsors: The Hospital for Sick Children Foundation had no role in design or conduct of the study; collection, analysis, or interpretation of the data; or preparation, review, or approval of the manuscript.

Additional Contributions: The authors wish to thank Denise Thomson, MA, MBA, from the Cochrane Child Health Field for her helpful comments on this manuscript and to Mark Inman for assistance with data collection.