|Home | About | Journals | Submit | Contact Us | Français|
Historically, drugs prescribed for children have not been studied in pediatric populations. Since 1997, however, a 6-month extension of marketing rights is granted if manufacturers conduct Food and Drug Administration (FDA)-defined pediatric trials. In nearly half the drugs studied, there were unexpected results in dosing, safety, or efficacy compared to adult studies, including failure of half of antihypertensive dose-response trials, which are pivotal for deriving dosing recommendations. We sought to define design elements that might have contributed to these trial failures by combining patient-level data from 6 dose-ranging antihypertensive efficacy trials completed for pediatric exclusivity and submitted to the Food and Drug Administration from 1998–2005. We evaluated dosing, primary endpoint, and other components to assess underlying reasons for failure to show efficacy in children. Of 6 trials examined, 3 showed a dose response; 3 did not. Eligibility criteria were similar across studies, as were subject demographics. Successful studies showed large differences in doses, with little or no overlap between low, medium, and high doses; failed trials used narrow dose ranges with considerable overlap. Successful trials also provided pediatric formulations and used reduction in diastolic, not systolic, blood pressure as the primary endpoint. Careful attention to pediatric pharmacology and selection of primary endpoints can improve trial performance. We found poor dose selection, lack of acknowledgment of differences between adult and pediatric populations, and lack of pediatric formulations to be associated with failures. More importantly, our ability to combine data across trials allowed us to evaluate and potentially improve trial design.
Historically, 75% of drug products used in pediatric populations have had insufficient labeling information for pediatric dosing, safety, or efficacy.1 The Food and Drug Administration Modernization Act of 1997 authorized an incentive of 6 months of marketing exclusivity for drug manufacturers who agreed to perform pediatric studies specifically defined in a “written request” issued by the U.S. Food and Drug Administration (FDA). The Pediatric Exclusivity Program has been successful, resulting in more than 130 labeling changes to date.
Since the program’s inception, approximately half of the products studied have been found to have substantive differences in dosing, safety, or efficacy in children when compared with adult populations.2 Twenty-nine of 131 drugs examined were found to be ineffective when studied in children. Several products that did not work (or for which a statistically significant dose response was not observed) were oral antihypertensive agents known to be effective in adults.3-10 These trial failures have significant public health implications because systemic hypertension occurs in nearly 2% of the pediatric population and is rising concomitantly with obesity in children and adolescents.11
The FDA allows for several types of trial design in the written request for an antihypertensive agent. The written request, generally issued by FDA prior to initiation of pediatric exclusivity studies, contains the required elements of the requested studies, including indication, number of studies, sample sizes, trial design, and age ranges.12 The most common antihypertensive trial design (Type C) (please see http://hyper.ahajournals.org for study schematic) was employed for 6 antihypertensive agents. Type C design includes a lead-in period of up to 10 days, an initial randomization phase into ≥ 2 active treatment arms (e.g., low, medium, and high dosage), a second randomization to double-blind withdrawal to placebo, and an open-label “safety” phase. The primary endpoint for these trials was to establish a dose response of sitting blood pressure to the agent.
Of the 6 Type C trials, 3—enalapril, lisinopril, and losartan—were successful, while 3—amlodipine, fosinopril, and irbesartan—did not show a statistically significant dose response. As a pattern of failed pediatric antihypertensive trials emerged, we sought to determine why these trials failed to show dose response in children, and hypothesized that difficulties in dosing might be the cause of trial failure.
Between January 1, 1998 and December 31, 2005, efficacy data from 12 antihypertensive products were submitted to the FDA’s Division of Cardiovascular and Renal Products. One efficacy trial and at least 1 pharmacokinetic trial were completed for each agent. These trials did not indicate differential dosing from adult data and did not suggest differences in absorption, bioavailability, clearance, or differential response by age or development. PK-PD samples were not obtained in these trials, but rather for each product in the PK-PD trials; therefore, we did not include specific PK-PD results in our analysis.
Of the 12 efficacy trials, 6 were of Type C design and included 3 products whose trials demonstrated efficacy, and 3 products whose trials did not demonstrate efficacy, in pediatric populations. The primary endpoint was prespecified by the sponsor as either change from baseline in systolic blood pressure (n=3 trials), or diastolic blood pressure (n=3). The data, protocols, case report forms, and all documents necessary for submission from these 6 trials were submitted electronically to FDA. We documented inclusion and exclusion criteria and other key aspects of study design for each trial (please see http://hyper.ahajournals.org for study eligibility criteria).
We obtained study datasets through FDA’s electronic document room repository. Source data were available in SAS datasets, which we converted to STATA datasets via STAT Transfer. We then combined patient-level data to obtain 1 observation per patient, as defined by the protocol for primary analysis.
From each trial we assembled 30 common variables: antihypertensive product, unique patient identification number, age, sex, race, height, weight, body mass index (BMI), baseline sitting diastolic blood pressure, baseline sitting systolic blood pressure, diastolic and systolic blood pressure at end of dose response phase, and amount of drug in milligrams and per dosing stratum (low, medium, or high). We calculated the z-score for BMI using a formula provided by the Centers for Disease Control and Prevention.13 We used a categorical variable of white/black/other for race because several trials used this format to report race, and more specific information was not available.
Baseline systolic and diastolic blood pressures were the average of 3 sequential values obtained at the beginning of the randomized dose-response phase. If 1 value was missing, the average of 2 observations was used. The blood pressures obtained at the end of the dose-response phase were calculated in similar fashion. If the last observation at the end of the dose-response phase was missing, we used the last observation carried forward as an imputation method. Change from baseline blood pressure was used as a primary endpoint to assess response to antihypertensive therapy. This was calculated by subtracting the end of dose-response phase blood pressures from the baseline values. Body weight dosing (mg/kg) was calculated by dividing the amount of drug administered by body weight.
We extracted each variable listed above into a common dataset in which each patient from each trial was represented with 1 observation in the master dataset. We categorized trials as successful or unsuccessful based on the primary analysis. We replicated the sponsor’s findings with respect to each trial’s primary endpoint—dose-response reduction in sitting blood pressure, either diastolic (n=3) or systolic (n=3).
Three approaches were used to assess dose response in each trial. In each analysis, we used a simple linear regression. For Analysis 1, the dependent variable was change in systolic blood pressure and the primary independent variable was the dosage arm (low, medium, or high) to which the subject was randomly assigned. In this analysis, dosage arm was evaluated as both a categorical variable and as a dummy variable. Analysis 2 was conducted similarly, except that the dependent variable was change in sitting diastolic blood pressure. In Analysis 3, change in sitting diastolic blood pressure was the dependent variable and the primary independent variable was the amount of product (mg/kg) the subject received; the amount of product administered was evaluated as a continuous variable. We used forward selection to add the covariates listed above; covariates were retained if they were associated (P <0.05) with trial failure.
A significant dose response was concluded if the slope of the regression line differed from zero at the 0.05 significance level. We report two-tailed P values and 95% confidence intervals (CI) from the analyses. We presented the project to the Duke University Medical Center Institutional Review Board and received a waiver of review for this analysis, as none of the patient-level data in any of the 6 trials had associated patient identifiers in the datasets.
Inclusion and exclusion criteria among trials were similar. The studies enrolled children 6–16 years of age; 5 of 6 trials accepted either systolic or diastolic blood pressure >95th percentile for age, sex, and height. The trials excluded children with severe hypertension (because of the randomized withdrawal phase), as well as children with low glomerular filtration rates, electrolyte abnormalities, renal disease, and other substantive medical problems.
Subject demographics among trials were also similar (Table 1). The fraction of male children and the age distribution of enrolled subjects were similar among trials. The enalapril and lisinopril trials enrolled a lower proportion of white children: 39% and 44%, respectively (compared to a mean of 65%), because the distribution of race was prespecified by protocol for these 2 trials. The distribution of weight and BMI of subjects varied among trials. The median BMI-z score of subjects enrolled in the enalapril and lisinopril trials was 1.0 and 1.1, respectively; median BMI z-score was 2.0 in the fosinopril trial. Three trials (amlodipine, enalapril, and fosinopril) distributed subjects equally between dosing arms (Table 1).
The range in amount of agent received by subjects randomized to low and high dosage groups was extremely variable among trials (Table 2). In the amlodipine trial, there was a twofold difference between the high dosage and low dosage groups (5 mg/2.5 mg=2). In the fosinopril and irbesartan trials, dosing ranges were also small: 6 and 9-fold, respectively. The enalapril, lisinopril, and losartan trials had considerably higher dosing ranges: 16-fold, 32-fold, and 20-fold, respectively.
Weight-based dosing strategies were inconsistent among trials. The amlodipine trial did not incorporate individual subject weight in dosing (Table 2), but rather gave all subjects in the low dosage arm 2.5 mg of product and all subjects in the high dosage arm 5 mg of product. This dosing strategy resulted in the following paradox: a 100 kg subject randomly assigned to “high” dosage received 0.05 mg/kg, and 20 kg subject randomly assigned to “low” dosage received 0.125 mg/kg. In the low-dosage group, one fourth of subjects received >0.06 mg/kg, and one fourth of the high-dosage group received <0.06 mg/kg (Figure 1). Although blood pressure did not show a dose response to amlodipine as randomized (Figure 2), and despite not being approved by FDA for treatment of hypertension in pediatric populations, increased dosage on an mg/kg basis was associated with decrease in blood pressure.
The fosinopril trial also failed to demonstrate a dose response, even though it incorporated individual subject weight into the dosing. This was likely because the trial limited dosage to a maximum of 40 mg. In this trial, subjects randomly assigned to medium dosage who weighed less than 30 kg received more fosinopril (in mg/kg) than the heaviest subjects randomly assigned to high dosage. Similarly to the amlodipine trial, blood pressure dose response was not associated with product as randomized, but increased dosing on an mg/kg basis was associated with blood pressure reduction (Table 3).
The 3 successful studies were trials of enalapril, lisinopril, and losartan, all of which demonstrated a dose-response reduction in sitting blood pressure. These 3 successful trials used change in diastolic blood pressure as the primary endpoint. The 3 unsuccessful studies (trials of amlodipine, fosinopril, and irbesartan) used change in sitting systolic blood pressure as the primary outcome (Table 2). Sample sizes ranged from 110–318 subjects randomly assigned to the dose-response phase. Larger sample size did not predict success: the 3 successful trials were the smallest. Dosing for 5 of the 6 agents was divided into low, medium, and high dosages (Table 2), but there was considerable variability among trials regarding how much agent a subject in the high-dosage arm received compared to a subject in the low-dosage arm.
The 3 successful trials had the largest differences in dosing ranges: 16-, 20-, and 32-fold differences between low and high dosage groups for the enalapril, losartan, and lisinopril trials, respectively, compared to 2-, 6-, and 9-fold for the amlodipine, fosinopril, and irbesartan trials (P=0.049, rank sum test).
We evaluated the reduction in systolic and diastolic blood pressure related to each agent (Table 3). A reduction in diastolic blood pressure was more closely related to the dosage of agent administered. In the enalapril trial, the dosage was more closely related to a reduction in diastolic blood pressure than systolic blood pressure (coefficient 0.19 P <0.001 vs. coefficient 0.12 P=0.08). We also observed a closer relationship between diastolic blood pressure reduction and dosage in the lisinopril trial (coefficient 0.12, P <0.001 vs. coefficient 0.08, P=0.09). Weight-based (mg/kg) exposure was associated with a reduction in blood pressure.
We were provided access to individual patient data for each of the Type C efficacy trials submitted for pediatric exclusivity from 1998 to 2005, inclusive. The primary endpoint for these trials was a dose-response change in sitting blood pressure between low, medium, and high dosage groups. We found that 3 trials (enalapril, lisinopril, and losartan) were successful. These trials had several design components in common: all used diastolic blood pressure as the primary endpoint, were characterized by a larger range in amount of agent given to low dosage vs. high dosage groups, and used a pediatric formulation in their efficacy trial.
Pharmaceutical companies continue to apply for pediatric exclusivity in the United States for antihypertensive agents, and an analogous program is now in place in the European Union. We believe that these results have several important implications for the design of future pediatric antihypertensive trials.
These data support use of reduction in sitting diastolic blood pressure, rather than systolic blood pressure, as the primary study endpoint. Diastolic blood pressure has less physiologic variability among observations within a subject than does systolic blood pressure in children.14-16 This reduction in variability may have contributed to the success of diastolic blood pressure as the primary endpoint. Systolic hypertension is approximately 3-fold more common than diastolic hypertension,17 and the motivation to use systolic blood pressure as the primary endpoint derives from feasibility, a common problem in conducting pediatric drug trials. However, the underlying causes of systolic and diastolic hypertension may differ (e.g., abnormal aortic compliance versus elevated systemic vascular resistence), and this may become a significant factor depending on the patient population (e.g., systolic hypertension is more common in elderly patients). A primary study endpoint of mean arterial blood pressure that incorporated both systolic and diastolic blood pressure values might prove advantageous, and this possibility should be explored in future trials.
The incentives in place for the Pediatric Exclusivity Program are designed to encourage trial completion; sponsors are given exclusivity if the trials are completed, and the decision to grant exclusivity is not dependent on product safety or efficacy. Feasibility is therefore of far greater importance to sponsors than optimal trial design. Eligibility for exclusivity regardless of outcome is a major advantage of Type C trial design because it is considered interpretable regardless of outcome; avoiding the use of an explicit placebo arm makes this type of trial more appealing to parents of potential subjects and institutional review boards.
Our results indicate that future pediatric antihypertensive trials should incorporate a wide range of doses and use information from adult trials to account for potential pharmacological differences between adult and pediatric populations. For example, the lowest clinical trial dose should be lower than the lowest approved dose in adults, and the highest clinical trial dose should at least be twofold higher than the highest approved dose in adults, unless contraindicated for safety concerns.
None of the failed trials investigated dose ranges higher than the corresponding adult doses. For example, the highest irbesartan dose was 4.5 mg/kg, while adult data indicate that most adults need doses up to 150–300 mg (≈ 2–4 mg/kg for a 75 kg child) for better blood pressure control.18 Data obtained from irbesartan use in adults showed that effects on blood pressure increase at doses up to 600 mg (approximately 8 mg/kg for a 75 kg child), and the maximum irbesartan dose studied in adults was 900 mg.
In contrast, the 3 successful trials provided large differences among low, medium, and high dose strata. All 3 successful trials employed doses much lower (nearly placebo) than the doses approved in adults. For example, the recommended initial lisinopril dose in adults is 10 mg and the usual dose range is 20–40 mg. The lowest dose used in the clinical trial was 0.625 mg, thus providing a wider range for exploring dose response. The selection of wide dose ranges is important for pharmacokinetic reasons, as closely spaced doses yield overlapping exposures among dose groups. If overlap is substantial, the dose response could appear flat and thus fail to demonstrate a significant dose-response relationship.
Further, 3 of these orally administered antihypertensive agents (those used in the failed trials) did not develop a pediatric (e.g., liquid) formulation and thus exhibited a wide range in exposure within each weight stratum. Development of a liquid formulation is often challenging: bioavailability can be unreliable, and dissolving the agent in liquid can require high concentrations of alcohol. Stability and bioequivalence testing of liquid formulations also require additional time and expense. Still, pediatric formulations should be requested in the Pediatric Exclusivity Program whenever possible. Development of these formulations is now more economically feasible because of benefits provided to companies for successfully completing trials requested by FDA as part of this program.
It is possible that the failed trials were unsuccessful because the agents do not work (or do not work well) in children. Each trial had a placebo-withdrawal stage to address this concern. Amlodipine and fosinopril, neither of which showed a dose response, both reduced blood pressure compared to placebo.7,8 The third agent that failed, however (irbesartan), did not show a clinically meaningful reduction in blood pressure in the placebo-withdrawal stage.19
One potential flaw in the conduct of these studies is the overall approach of study sponsors to compliance with FDAMA requirements: these studies are often designed and executed at the end of the product’s period of marketing protection. This problem is not limited to antihypertensive agents, but exists across all products, and can result in trials that fail to provide optimal data for the practicing clinician.
Failure to document dose response has been noted in other FDA-approved study designs and is not confined to Type C trials. One solution might be to encourage lowering of blood pressure compared to placebo (e.g., over a 4-week period) with subsequent follow-up studies to determine long-term safety. This would eliminate reliance on dose-response findings and put greater emphasis on long-term safety exposure.
Our study is limited by the fact that we conducted a post hoc analysis of only 6 trials. However, there are very few pediatric efficacy trials, and if we are to improve trial design, post hoc analyses will need to be completed after relatively few studies.
In conclusion, acccess to protocols, final study reports, and individual patient data for each trial was crucial to our investigation. Our analysis highlights potential improvements in trial design, accurate assessment of product efficacy, and improved public health when access to data across trials is granted to investigators.
Poor dose selection, failure to fully incorporate pediatric pharmacology into trial design, lack of pharmacokinetic information, and use of systolic blood pressure as the primary endpoint likely led to the failure of several antihypertensive pediatric exclusivity trials. Complete access to patient-level data allowed us to fully examine trial results, and may result in better design for future studies. Our data may be applicable to efforts to improve pediatric clinical trial design by government agencies, clinicians, and pharmaceutical sponsors in both North America and Europe. In the future, we recommend that pediatric antihypertensive trials should 1) develop an exposure-response model using adult data and published pediatric data, and employ this model to perform clinical trial simulations of pediatric studies and to explore competing trial designs and analysis options; 2) work with FDA to design pediatric trials by leveraging prior quantitative knowledge; and 3) routinely collect blood samples at informative time points to assess the pharmacokinetics in each subject to ascertain exposure response analysis.
The authors wish to thank Dr. Norman L. Stockbridge of the Division of Cardiovascular and Renal Products Center for Drug Evaluation and Research, FDA, who provided access to study data. The authors also wish to thank Jonathan McCall for editorial assistance with this article.
Sources of Funding Drs. Benjamin and Li received support from National Institute of Child and Human Development Grant 1U10-HD45962-04 and the U.S. Food and Drug Administration. Drs. Benjamin, Li, Smith, and Califf received support from 1UL1RR024128-01. The views expressed are those of the authors. No official endorsement by the U.S. Food and Drug Administration is provided or should be inferred. The Duke Clinical Research Institute was the coordinating center for the Fosinopril and Amlodipine trials.
Disclosure information Daniel K. Benjamin, Jr., MD, PhD, MPH: Dr. Benjamin receives significant support from NICHD Grants 1U10-HD45962-04 and 1UL1RR024128-01. Dr. Benjamin also receives research support from Astrellas, Rockeby, Cape Cod Associates, Pediatrix, and Thrasher Research, and fellowship funding through grants from Astra Zeneca and Johnson and Johnson. All monies are disbursed through Duke University; Dr. Benjamin does not receive salary support or direct monies from these companies, nor does he own stock or have any other financial interest in these companies. P. Brian Smith, MD: Dr. Smith receives significant support from NICHD grant 1UL1RR024128-01. Pravin Jadhav, PhD: none. Jogarao V. Gobburu, PhD: none. M. Dianne Murphy, MD: none. Vic Hasselblad, PhD: none. Carissa Baker-Smith, MD: Dr. Smith receives significant grant support from NICHD 1UL1RR024128-01. Robert M. Califf, MD: Dr. Califf receives significant grant support from NICHD 1UL1RR024128-01. Dr. Califf receives consulting income from the following: nonsignificant: Astra Zeneca, Biogen, Bayer, Brandeis, Bristol Meyers-Squibb/Sanofi, Five Prime, Kowa Research Institute, Sanofi-Aventis, Scios Pharma, Vertex; significant: Heart.org (Conceptis), Merck, Nitrox LLC, Schering Plough. All consulting income is donated to non-profit organizations, with the majority being designated for the Duke Clinical Research Institute’s clinical fellowship fund. Dr. Califf receives salary support from the following: Novartis Pharmaceuticals and Schering Plough (all payments are made to Duke University). Dr. Califf’s educational activities result in revenue from the following for Duke University: nonsignificant: Merck, Sanofi-Aventis, Schering-Plough; significant: Heart.org/Conceptis, Kowa Research Institute, Novartis Pharmaceuticals, Scios Pharma. Dr. Califf holds equity in excess of $10,000 in Nitrox LLC. Jennifer S. Li, MD, MHS: Dr. Li receives significant support from National Institute of Child and Human Development Grant 1U10-HD45962-04.