Setting priorities for a research agenda to combat drug-resistant tuberculosis in children

Abstract

Setting: Numerous knowledge gaps hamper the prevention and treatment of childhood drug-resistant tuberculosis (TB). Identifying research priorities is vital to inform and develop strategies to address this neglected problem.

Objective: To systematically identify and rank research priorities in childhood drug-resistant TB.

Design: Adapting the Child Health and Nutrition Research Initiative (CHNRI) methodology, we compiled 53 research questions in four research areas, then classified the questions into three research types. We invited experts in childhood drug-resistant TB to score these questions through an online survey.

Results: A total of 81 respondents participated in the survey. The top-ranked research question was to identify the best combination of existing diagnostic tools for early diagnosis. Highly ranked treatment-related questions centred on the reasons for and interventions to improve treatment outcomes, adverse effects of drugs and optimal treatment duration. The prevalence of drug-resistant TB was the highest-ranked question in the epidemiology area. The development type questions that ranked highest focused on interventions for optimal diagnosis, treatment and modalities for treatment delivery.

Conclusion: This is the first effort to identify and rank research priorities for childhood drug-resistant TB. The result is a resource to guide research to improve prevention and treatment of drug-resistant TB in children.

The burden of childhood tuberculosis (TB) reflects ongoing TB transmission in a community.¹ About one million children fall sick with TB every year.² The pattern of drug resistance in children in a community generally mirrors that of the adult population.³ Among adults, 3.7% of new cases and 20% of previously treated cases were estimated to have multi-drug-resistant TB (MDR-TB) worldwide in 2012.⁴ In 2010, an estimated one million children developed TB disease, among whom 32 000 had MDR-TB.² Difficulties in diagnosis due to insensitive tools for microbiological confirmation in children hamper the estimation of the burden of drug-resistant TB in this population. Thus, accurate information on mortality and morbidity due to drug-resistant TB in children is lacking. This is only one among the multiple research and knowledge gaps that negatively affect children suffering from drug-resistant TB.

The goal of achieving a reduction in the global burden of TB by 2015, in line with the Millennium Development Goals, encompasses the protection of vulnerable populations, including children, from TB, TB and human immunodeficiency virus (HIV) infection and multidrug-resistant TB among its objectives.⁵ Enabling and promoting research by making the best use of currently available tools and developing new tools for TB diagnosis, treatment and prevention is an important step towards achieving the goal of eliminating TB by 2050.⁵ Evidence-based research priority setting is essential for effective use of funds for research and development. In this context, identification of research priorities for drug-resistant TB in children is important to address the knowledge gaps and eventually to reduce TB deaths and disease in children. A global partnership of researchers, care givers and advocates who share a vision of a world where no child dies from this curable and preventable disease, called the Sentinel Project on Pediatric Drug-Resistant Tuberculosis, has begun collaborating toward this end (www.senti-nel-project.org).

We sought to conduct a priority-setting exercise among a group of paediatricians, experts and advocates to identify and rank priority research questions related to epidemiology, diagnostics, treatment and prevention of drug-resistant TB in children.

METHODS

We adapted the methodology developed by the Child Health and Nutrition Research Initiative (CHNRI) (http://www.chnri.org/publications.php) to define the context of the problem and categorise the research priorities.^6–8 CHNRI methodology has been used to set research priorities in child health in several areas, including childhood pneumonia, diarrhoea, neonatal infections, zinc supplementation and asphyxia.^8–12

Categorisation of drug-resistant tuberculosis research priorities in children

As a first step, a list of potential research priorities was prepared by reviewing published studies, including review articles and systematic reviews.^13–21 This included questions that would need to be answered to understand the burden of drug-resistant TB in children and its determinants, to improve the performance of existing strategies and to develop new strategies for diagnosis and management. The initial list of 79 questions was compressed to 53 questions in broad consultation with TB experts and paediatricians.

A total of 10 experts, including paediatricians, identified based on their experience in childhood TB, drug-resistant TB or both, was consulted to finalise the list of questions. The finalised questions were pilot-tested among 10 other clinicians to achieve better clarity in framing and wording. The two groups of respondents included in the formal ranking process were not involved in finalising the questions, nor were they involved in pilot testing the questions.

The questions were then organised using the CHNRI framework for listing research priorities. The final list used for scoring consisted of questions of three different research types: descriptive, development and discovery. These were systematically listed among four research areas, namely epidemiology, diagnostics, treatment and prevention (Table 1).

TABLE 1.

Methodology used for development and categorisation of research

Scoring of research questions

In July 2013, the final list of research questions was e-mailed to two groups of respondents who were asked to score the questions via an online survey (Survey Monkey, Palo Alto, CA, USA). The first group consisted of the 304 individual members of the Sentinel Project on Pediatric Drug-Resistant Tuberculosis; the second group was a convenience sample of 48 medical doctors from all over India who participated in a workshop on childhood drug-resistant TB held at the National Institute for Research in Tuberculosis (Chennai, India) in June 2013. The resulting group of respondents had expertise in the field of TB, drug-resistant and/or childhood TB. Online survey responses were accepted until 30 September 2013.

The respondents were asked to assess the research questions based on five criteria: answerability, feasibility, effectiveness, deliverability and equitability (Table 2). These criteria were selected from a list of all possible criteria that could be used for setting priorities in health research investments.⁴ The priority-setting criteria were modified to reflect the context and subject of drug-resistant TB in children. For each criterion, the respondents were requested to tick any one of the following responses: yes, probably yes, probably no, no, do not know (Appendix Figure A). The respondents answered the questions independently of each other and were sent two reminders before being classified as non-responders.

TABLE 2.

Criteria used to assess the research questions

Computing the research priority score and average expert agreement

The following scores were given to each of the responses at the time of analysis: yes (1), probably yes (0.75), probably no (0.25), no (0), and do not know (blank). The research priority score (RPS) was used to rank the research priorities. RPS is the overall value of each research question when all the criteria are taken into account. The overall RPS was computed as the mean of the scores for the five criteria using the following formula:

The average expert agreement (AEA) score refers to the proportion of respondents who gave the same answer most frequently, and helped to achieve discrimination between levels of agreement among respondents on the prioritisation of the research questions. The AEA was calculated for each research question as the average proportion of respondents who gave the most common answer on the five questions asked, using the following formula:

where q is the criteria question that respondents used to evaluate competing research options ranging from 1 to 5.

The research questions were then ranked based on the RPS.

RESULTS

Of the 81 individuals who responded to the survey, the vast majority were clinicians (n = 63, 78%) (Appendix Table A.1). Of the 63 clinicians, 52 (83%) had expertise in TB or multidrug-resistant (MDR) TB and 26 (41%) were paediatricians. Of the 18 non-clinicians, 11 (61%) had expertise in TB or MDR-TB. Of the total 81 respondents, 63 (78%) had expertise in TB or MDR-TB. Of the 79 who reported their location, 30 (38%) were based in Asia, 29 (37%) in the Americas and 13 (17%) in Africa.

Among the 53 research questions, each research area—epidemiology, diagnosis, treatment and prevention—contained respectively 17, 9, 21 and 6 questions. The number of questions according to the research type—descriptive, development and discovery—was respectively 25, 16 and 12. The ranking of the 53 research questions with their RPS and AEA scores is provided in Appendix Table A.2. The scores ranged from 46 to 88 (potential 0–100), with scores of >80 for about 40% of the questions.

Table 3 shows the 10 highest-ranked questions based on the RPS and AEA. Of the 10 questions, 6 were related to treatment, 3 to epidemiology and 1 to diagnosis. The RPS ranged from 82 to 88 for the top 10 ranked questions and the AEA varied from 0.55 to 0.68. The top-ranked research question referred to the best combinations of existing tools for the early diagnosis of drug-resistant TB in children, which had an AEA of 0.62. The highest AEA, of 0.68, was for the question about the adverse effects of anti-tuberculosis drugs, which had an RPS of 87 and ranked second. Questions pertaining to treatment covered the reasons for poor treatment outcomes, optimal treatment duration and interventions for improving treatment outcomes. The prevalence of drug-resistant TB was the highest-ranked question in the epidemiology grouping, with an AEA of 0.58.

TABLE 3.

The 10 top-ranked research questions with RPS and AEA

The 10 lowest-ranked questions had an RPS ranging from 46 to 66 and an AEA varying from 0.30 to 0.44 (Table 4). There were 6 questions pertaining to epidemiology, 3 to treatment and 1 to diagnosis. In terms of question type, 7 questions were the discovery type, 2 the descriptive type and 1 the development type. The questions related to discovery were on basic research for understanding the disease pathogenesis, transmission and biomarkers. Management of drug-resistant TB in the private sector featured among one of the lowest ranked questions, with an AEA of 0.41.

TABLE 4.

The 10 lowest ranked research questions with RPS and AEA

The highest-ranked questions according to research area and research type are shown in Tables 5 and 6. The epidemiology questions were related to estimating the burden, risk and determinants of disease, while the diagnostic questions evaluated existing and new diagnostic tools. Optimal screening modalities and preventive therapy were the highest-ranked questions for prevention, while adverse effects of second-line drugs and reasons for poor treatment outcomes were predominant in the treatment research area. The highly ranked questions of the descriptive type referred to treatment outcomes, whereas the development type questions were primarily about interventions for optimal diagnosis, treatment and modalities for treatment delivery. New drug evaluation and models for preventive therapy and for preventing new infections were predominant among the discovery type questions.

TABLE 5.

The top three research questions in each research area and their rank

TABLE 6.

The top five questions in each research type and their rank

DISCUSSION

We report on the first effort to identify and rank research priorities specifically for preventing childhood disease and death due to drug-resistant TB in children. This is important because it can support more informed conversations among practitioners, investigators and funders who share an interest in pursuing collaborative work in this area.

Early diagnosis and prompt treatment initiation reduce morbidity and mortality due to TB disease. The highest ranked research priority in this survey was to identify the best combinations of existing diagnostic tools to enhance early diagnosis of drug-resistant TB in children. The utility of newer diagnostic tools and non-invasive specimen collection methods for diagnosing drug resistance were identified as other high-priority diagnostic research questions. For example, the use of more sensitive tools for earlier diagnosis should be explored in children, such as the Xpert^® MTB/RIF Ultra assay (Cepheid, Sunnyvale, CA, USA), as well as earlier identification by whole genome sequencing of drug susceptibility profiles. The difficulties of establishing a diagnosis of drug-resistant TB in children are well known, and our findings reiterate the need for more research that can produce child-appropriate tests.

In the epidemiology research area, the prevalence of drug-resistant TB among children in different settings was a priority question. Conducting TB prevalence studies in children, although challenging, is essential, as this information is important to estimate the disease burden and assess the impact of interventions. Research involving the identification of children at high risk for drug-resistant TB, including household contacts and those with HIV infection, was ranked as high priority. Understanding the disease burden through monitoring high-risk groups is important for early identification of disease and for planning prophylactic strategies to prevent development of disease.

The treatment-related questions that were considered high priority were mainly descriptive and were related to reasons for poor treatment outcomes and adverse effects of anti-tuberculosis drugs. This type of information is essential for planning appropriate interventions that can increase cure rates and reduce mortality due to the disease. The development-type questions related to treatment that were highly prioritised were identifying the optimal treatment duration, child-friendly formulations of second-line anti-tuberculosis drugs and newer interventions for treatment delivery for favourable treatment outcomes. Treatment in children in terms of dosages of drugs, their combination and duration are derived mainly from studies done in adults, from which data are extrapolated to children. This exercise has prioritised the need to study optimal treatment and effective treatment delivery strategies for drug-resistant TB in children, which is important to reduce TB-related mortality and disease sequelae.

The six research questions pertaining to prevention of disease ranked between 26 and 37. The reasons for this could be the paucity of data even among adults on the optimal preventive therapy regimen and its duration. Effective screening algorithms to safely initiate preventive therapy and the optimal preventive therapy regimens ranked higher in the preventive area of research.

Descriptive and development types of research questions received a high RPS of >80%, highlighting the need for research in evaluating and improving existing strategies, while discovery research questions (5/10) in terms of understanding the host susceptibility to disease and identification of biomarkers ranked lower. The reason for this could be that the translational benefits of discovery in basic research take longer to have an impact on reductions in mortality and morbidity. However, discovery research involving new TB drugs and models for prevention of disease scored high, emphasising that respondents consider that discovery research is imperative to reduce the burden of drug-resistant disease in children.

Understanding the current management of childhood drug-resistant TB in the private sector featured among the 10 lowest ranked questions. This is an interesting finding given that TB is managed equally in both the private and public sector in many countries, including India. The lower priority ranking may reflect a misperception that childhood TB is infrequently managed in the private sector,²² or it may reflect the respondents' assessment that producing evidence in other areas (i.e., how best to diagnose or treat childhood drug-resistant TB) will have more impact on improving care for children. Certainly, poor prescription practices have been documented among private practitioners,^23,24 and in some countries any sustainable impact on TB burden must necessarily involve the private sector.

Limitations of this survey include the relatively small pool of survey respondents, whose overall profile is that of clinicians or researchers interested in drug-resistant TB. However, this characteristic is also an advantage, as the group of respondents includes individuals with experience in programmes aiming to treat and prevent drug-resistant TB specifically in children. The fact that 44 of the 53 questions had a score of >70% showed that most of the research questions were already deemed important by the experts. Further, the survey used a modified criterion assessment scale to suit the context and operational feasibility.

Another potential limitation to the generalisability of the ranked question list is that 23 of the 79 (29%) respondents who reported their country were based in a single country (India). This was a result of including the convenience sample afforded by the workshop participants from across India, which is home to 25–30% of global childhood drug-resistant TB cases,^2,25,26 but where very few data are available on MDR-TB in children.²⁷ It is important to note that our survey was not designed to represent countries according to their relative disease burden. Our survey was designed to synthesise the opinions of individuals with expertise or interest specifically in childhood drug-resistant TB—many of whom have worked in multiple countries—who were willing to rank research gaps that presently impact children with drug-resistant TB. Future priority-setting exercises that aim to update this ranking may be designed differently; for example, they may purposely recruit respondents based in specific countries or according to other types of geographic or institutional settings.

In summary, the problem of drug-resistant TB in children is poorly understood, and the optimal approach is unknown for reducing the toll of this disease specifically in children. Diagnosis of TB in children is challenging due to difficulties in collecting samples, while the paucibacillary nature of the disease reduces the sensitivity of testing. The treatment options for children are based on expert opinion, case series and cohort studies, with small sample sizes.²⁸ Treatment is expensive and remains associated with poor prognosis and a high risk of toxic effects.²⁹ Thus, prevention of progression to MDR-TB disease in children could be a practical and cost-effective solution from the public health perspective.³⁰

Our results suggest that, for the time being, priority should be given to more descriptive and development types of research studies related to the diagnosis and treatment of drug-resistant TB in children. The prioritisation of funding for TB research projects will ultimately be determined by each organisation's investment context, i.e., the balance between those results needed within a short term and those that are more long-term. This is the first effort to establish research priorities in this area of child health through a systematic survey of expert stakeholders based in multiple countries. The results may serve as a framework that can be updated periodically and as a shared resource to guide new research to improve the prevention and treatment of drug-resistant TB in children.

APPENDIX

FIGURE A.

Screenshot of the Survey Monkey questionnaire.

TABLE A.1.

Qualification, expertise, and country of survey respondents

TABLE A.1.

(continued)

TABLE A.2.

Final ranking of all 53 research questions with RPS and AEA

TABLE A.2.

(continued)

TABLE A.2.

(continued)

TABLE A.2.

Final ranking of all 53 research questions with RPS and AEA