A guide to aid the selection of diagnostic tests

Abstract In recent years, a wide range of diagnostic tests has become available for use in resource-constrained settings. Accordingly, a huge number of guidelines, performance evaluations and implementation reports have been produced. However, this wealth of information is unstructured and of uneven quality, which has made it difficult for end-users, such as clinics, laboratories and health ministries, to determine which test would be best for improving clinical care and patient outcomes in a specific context. This paper outlines a six-step guide to the selection and implementation of in vitro diagnostic tests based on Médecins Sans Frontières’ practical experience: (i) define the test’s purpose; (ii) review the market; (iii) ascertain regulatory approval; (iv) determine the test’s diagnostic accuracy under ideal conditions; (v) determine the test’s diagnostic accuracy in clinical practice; and (vi) monitor the test’s performance in routine use. Gaps in the information needed to complete these six steps and gaps in regulatory systems are highlighted. Finally, ways of improving the quality of diagnostic tests are suggested, such as establishing a model list of essential diagnostics, establishing a repository of information on the design of diagnostic studies and improving quality control and postmarketing surveillance.


Introduction
Diagnostic testing has become indispensable for diagnosing and monitoring disease, for providing prognoses and for predicting treatment responses. 1,2 Today, over 40 000 products are available globally for the in vitro diagnostic testing of a wide range of conditions. 3 These include traditional laboratorybased tests, with samples being sent to a central laboratory for analysis, and point-of-care tests, which can be performed near, or at, the point of patient care. Point-of-care testing can help optimize treatment decision-making, avoid referrals, improve the efficiency of care and decrease costs, especially in resourceconstrained settings where laboratory infrastructure is weak. 4 In the early 1990s, the first point-of-care tests for use in resource-constrained settings became commercially available: lateral flow immunoassays (often called rapid diagnostic tests) for the diagnosis of malaria. 5,6 These assays are now well established and have replaced blood film microscopy in many settings. However, as the market for diagnostic tests has increased, the choice has become overwhelming for some diseases: in 2015, the World Health Organization (WHO), the Foundation for Innovative New Diagnostics (FIND) and the Centers for Disease Control and Prevention (CDC) in the United States of America reviewed approximately 250 different diagnostic tests for malaria. 7 We conducted an online market search for screening tests for hepatitis C virus infection and identified more than 50 products, and UNITAID's 2015 report on tuberculosis diagnostics highlighted the increasing complexity of the market, 8 with WHO endorsing (though not prequalifying) several products in recent years. [9][10][11][12] In some cases, a single manufacturer may hold a monopoly, which can lead to high costs. Conversely, diagnostic and monitoring tests for neglected diseases, such as visceral leishmaniasis, human African trypanosomiasis, chikungunya, dengue and brucellosis, remain scarce. 13,14 With the increase in the number of in vitro diagnostic tests has come an increase in the number of guidelines and recommendations, together with countless publications on their performance and implementation. 7,13,[15][16][17] However, this wealth of material covers only a small proportion of commercially available tests. In our experience, the information available on many tests is limited and there is often a lack of independent data on a test's performance and on whether the manufacturing process is reliable enough to ensure consistent quality across multiple lots. Both the quantity and the variable quality of the information available make it difficult for policymakers, laboratories and other end-users to make rational decisions about the selection and use of these tests. 18,19 As a result, tests have been used unnecessarily and incorrectly and results have been misinterpreted. [20][21][22][23] Given these difficulties, the process of selecting one or several tests for use in a diagnostic algorithm can be cumbersome for clinics, countries and nongovernmental organizations providing medical support. 19 The nongovernmental organization, Médecins Sans Frontières, operates or supports health ministry laboratories in more than 40 countries. Currently, the organization has over 15 laboratory advisors at its headquarters and more than 100 staff working in laboratories around the world. The amount of laboratory equipment and the number of in vitro diagnostic tests used by Médecins Sans Frontières itself have almost doubled in the past 10 years and the organization has encountered numerous challenges in selecting and implementing tests for its projects.
In this article, we outline a six-step approach to overcoming the obstacles encountered by Médecins Sans Frontières in selecting and implementing in vitro diagnostic tests. This approach was derived from a review of the diagnostics literature and from our experience with implementing diagnostics programmes. We discuss the challenges involved in each of the six steps and outline current problems with the quality assurance of tests. We hope this simple stepwise guide will help clinics, organizations and health ministries to make rational decisions about the selection of in vitro diagnostic tests and, over the longer term, will contribute to the development of a practical guide for selecting diagnostics.
Abstract In recent years, a wide range of diagnostic tests has become available for use in resource-constrained settings. Accordingly, a huge number of guidelines, performance evaluations and implementation reports have been produced. However, this wealth of information is unstructured and of uneven quality, which has made it difficult for end-users, such as clinics, laboratories and health ministries, to determine which test would be best for improving clinical care and patient outcomes in a specific context. This paper outlines a six-step guide to the selection and implementation of in vitro diagnostic tests based on Médecins Sans Frontières' practical experience: (i) define the test's purpose; (ii) review the market; (iii) ascertain regulatory approval; (iv) determine the test's diagnostic accuracy under ideal conditions; (v) determine the test's diagnostic accuracy in clinical practice; and (vi) monitor the test's performance in routine use. Gaps in the information needed to complete these six steps and gaps in regulatory systems are highlighted. Finally, ways of improving the quality of diagnostic tests are suggested, such as establishing a model list of essential diagnostics, establishing a repository of information on the design of diagnostic studies and improving quality control and postmarketing surveillance.

Selecting a test
The ASSURED (Affordable, Sensitive, Specific, User-friendly, Rapid and robust, Equipment-free and Deliverable to end-users) criteria can be used as a benchmark for identifying the most appropriate diagnostic tests for resource-constrained settings. 24 However, these criteria are generic and need to be adapted to each diagnostic need. In addition, not all test methods can be simplified to fit the ASSURED criteria: for instance, both laboratory infrastructure and equipment are required for diagnosing human immunodeficiency virus (HIV) infections in young infants using dried blood spot collection and for monitoring HIV viral load in infected individuals. In these situations, the test's specification must be broadened during the selection process.
We have identified six steps that must be addressed when selecting an in vitro diagnostic test: (i) define the test's purpose; (ii) review the market and check each product's specification; (iii) review the test's regulatory approval; (iv) obtain data on the diagnostic accuracy of the test under ideal conditions (i.e. in laboratory-based evaluations); (v) obtain data on the diagnostic accuracy of the test in clinical practice; and (vi) monitor the test's performance in routine use (Box 1). These six steps, along with context-specific barriers to use, should all be considered when selecting an in vitro diagnostic test for routine use.

Step 1: Defining the test
Clearly defining the test's purpose is important, because this will influence many of the subsequent steps in the selection process. Factors to be considered include: (i) the disease or condition to be diagnosed; (ii) whether a single test or a diagnostic algorithm is required; and (iii) whether the test should, or can, provide a qualitative or quantitative result. In addition, for optimal clinical utility, consideration should be given to: (i) the site of testing (e.g. in a large laboratory or a small health-care centre); and (ii) the end-user (e.g. a well-trained laboratory technician or a primary health-care worker, such as a nursing assistant). Other important considerations are the clinical use of the test (e.g. whether a screening or confirmatory test is required) and the added value of the test or combination of tests. 25,26 Step 2: Market review The market should be reviewed to identify the tests available for the condition of interest by consulting guidance from international organizations such as WHO and manufacturers' product information. Product specifications may include details of the type of sample required, the test's operating conditions, additional equipment required and shelf life. The veracity of manufacturers' claims should be judged by searching the peer-reviewed medical literature. The manufacturer often overstates a test's diagnostic accuracy in test brochures or instructions for use: for example, its sensitivity and specificity may not have been replicated in independent evaluations. 27 Unfortunately, few prosecutions for inappropriate claims have been pursued by national authorities. The selection process can also be hindered by local market problems with counterfeit tests and by a lack of regulation. 4,[28][29][30] Step 3: Regulatory approval Many countries do not have regulatory procedures in place for assessing the safety, quality or effectiveness of in vitro diagnostic tests, which means that poor-quality tests can be marketed and used. 4,28,30 However, test selection can be assisted by consulting recommendations from international regulatory bodies, for some major conditions at least. For example, WHO has established a prequalification process for in vitro diagnostic tests for diseases with a high individual

Box 1. Steps in selecting a diagnostic test
Step 1: Define the test's purpose -why, what, where, who?
• Decide whether an acute or chronic infection is to be diagnosed • Decide whether the test is to be used for diagnosis, disease monitoring or verifying a cure • Decide whether the test should be quantitative or qualitative • Decide whether test results will be analysed at the point of care or in a central laboratory • Define the test's end-users: trained laboratory technicians or primary health-care workers?
• What is the required performance of the test?
Step 2: Review the market or public health risk -the process has a particular focus on ensuring that tests for HIV, malaria and hepatitis B and C infections are affordable in resourceconstrained settings. 31 The prequalification process has three main components: (i) a review of the product application and dossier; (ii) laboratory evaluation of the product; and (iii) inspection of the manufacturing site. The process thus assesses both the test's performance and manufacturing quality. More than 60 products have been prequalified since the process started in 2010. 13 Recently, tests for glucose-6-phosphate dehydrogenase deficiency, screening for human papillomavirus and emergency assessments in the outbreak of diseases such as Ebola and Zika have been included, but many others have not.
Similarly, the Global Fund to Fight AIDS, Tuberculosis and Malaria has produced a list of in vitro diagnostic tests eligible for procurement, which is based on WHO's recommendations and on products approved by the regulatory authorities of the founding members of the Global Harmonization Task Force: Australia, Canada, the European Union, Japan and the United States. 32 Other authorities and donors may use different lists. For example, the President's Emergency Plan For AIDS Relief (PEPFAR) relies on the United States Agency for International Development's list of approved tests for HIV/AIDS. 33 Although these sources list a broad range of recommended products, guidance on the use of diagnostic tests for conditions other than HIV infection, malaria and tuberculosis is scarce. Moreover, few in vitro diagnostic tests for neglected diseases or products for monitoring the side-effects of drug therapy have been endorsed, or approved for use, by international agencies.

Step 4: Optimal diagnostic accuracy
The performance of a test under ideal conditions (i.e. in phase-II studies) indicates its optimal performance. This information is crucial for enabling users to preselect a test for a trial under reallife conditions. Such evaluations may provide important information not only on a test's diagnostic accuracy but also on its repeatability, reproducibility and ease of use and on variations between production lots. For some infectious diseases, evaluations of diagnostic tests are carried out at regular intervals by in-ternational stakeholders and the results are made publicly available: for example, tests for malaria are routinely evaluated by WHO and FIND and other tests are monitored during WHO's prequalification process. 7 No similar evaluations have been carried out for many other conditions. Nevertheless, guidance on the evaluation of diagnostic tests for several infectious diseases has been provided in several publications. 34,35 Individual organizations may find it difficult to carry out these evaluations themselves because often they can be performed only by reference laboratories with high-quality infrastructure and highly skilled-staff.
Specimen banks can provide material for evaluating in vitro diagnostic tests and can be useful for helping manufacturers develop high-quality diagnostic products. 36 Unfortunately, we only know of specimen banks for tuberculosis and human African trypanosomiasis. In addition, WHO, FIND and the CDC provide malaria specimens for manufacturers evaluating prototypes of rapid diagnostic tests. Access to specimen banks, especially for neglected infectious diseases, would be hugely beneficial for the development and monitoring of in vitro diagnostic tests.

Step 5: Diagnostic accuracy in practice
Both the actual performance of in vitro diagnostic tests and their ease of use should also be considered during the selection process. Evaluations at the end-user level (i.e. phase-III studies) provide information on a test's performance under real-life conditions and can reveal important features that were not revealed in phase-II studies. In practice, a test's performance in the field can be influenced by the user's level of training and by environmental conditions, such as a high temperature or humidity and dust. Evaluations under real-life conditions are essential because often staff carrying out diagnostic testing in resource-constrained settings (e.g. HIV counsellors) will not have had the same training as laboratory workers at reference laboratories. Also, an end-user's perception of a test's ease of use may be quite different from that of staff at a reference laboratory. Real-life evaluations should be conducted in the population in which the test will be used as this will provide clinical accuracy data that are appropriate for the prevalence of the disease locally and for other contextspecific factors that could influence accuracy, such as common comorbidities.
If no phase-III evaluations have been carried out, end-users should consider performing such evaluations themselves. In practice, national regulatory authorities often require phase-III studies to be performed before approving the introduction of an in vitro diagnostic test. Such evaluations can be conducted by the health ministry, test developers or other actors. Some guidance exists: for example, guidelines from WHO, CDC and the Association of Public Health Laboratories on the evaluation of HIV testing technologies in Africa 37 and generic protocols developed by the International Diagnostics Centre. 38 Again, these focus on only a few diseases.
The results of diagnostic accuracy studies can be obtained from public reports or from the peer-reviewed literature. When reviewing publications, it is important to be aware of the stage at which the test was evaluated, which could be: (i) a prototype evaluation (phase I); (ii) an evaluation under ideal conditions (phase II); or (iii) an evaluation under real-life conditions (phase III). Next, the quality of the study data should be assessed. The Standards for Reporting Diagnostic Accuracy (STARD), updated in 2015, provide guidance on improving the quality of reporting of research on diagnostic test accuracy but, unfortunately, much reporting is still incomplete. 39 Nevertheless, these standards can help end-users judge the quality of the study being assessed. It should also be noted that reviewing the literature on diagnostic tests can provide useful information on the challenges faced when introducing and routinely using new tests. Recently, the Cochrane Collaboration has started conducting systematic reviews and meta-analyses of diagnostic test accuracy studies. 40 These reviews are comprehensive, provide information on the quality and reliability of studies, and can greatly help end-users interpret published data.

Step 6: Monitoring performance
Monitoring a test's performance in routine use is important. Quality control, proficiency testing and the supervision of end-users should be carried out regularly and documented. Postmarketing surveillance is another important component of long-term quality assurance.
Choosing the right diagnostic test Cara S Kosack et al.
This is usually performed by the relevant national authorities and is included in WHO's prequalification process, to which end-users can contribute. Postmarketing surveillance is both reactive (e.g. in response to complaints by procurers and end-users) and proactive (e.g. in verifying the quality of production lots both before and after distribution).
In addition, end-users' reports about prequalified tests are collected and investigated during WHO's prequalification process. Unfortunately, neither end-users' complaints nor analyses of these complaints are publicly available. However, WHO does publish field safety notices if complaints are substantiated.

Gaps in guidance
The diagnostic accuracy of in vitro diagnostic tests has been evaluated mainly in diseases subject to major control efforts, such as HIV infection, tuberculosis and malaria. However, even for these diseases, no comprehensive, structured, pragmatic guidelines exist that can be used to help national governments, diagnostic programmes or laboratories with selecting tests. The absence of guidance is an even greater problem when diagnosis requires more than a simple rapid test. Guidance on how to choose, implement or monitor more complex diagnostic methods is also scarce. Furthermore, procurement guides are nonexistent, with two notable exceptions: WHO's procurement guides on laboratory equipment for HIV testing and on rapid diagnostic tests for malaria. 41,42 Finally, most existing guidelines do not fully consider cost. This is particularly important for more complex tests because the cost of transport, storage, remodelling laboratory structures, training and supply chain management, for example, need to be considered together with the cost of the test itself. 43 According to WHO's CHOICE (Choosing Interventions that are Cost-Effective) project, 44 cost-effectiveness studies that use thresholds based on per-capita income provide little guidance because they disregard budgetary constraints. For example, it has been estimated that, although use of the GeneXpert MTB/ RIF test in 15% of suspected tuberculosis cases in India would be cost-effective at 2010 prices, the test would consume the entire budget of the country's tuberculosis programme. 45 However, alternative ways of evaluating cost-effectiveness in low-and middle-income countries have been proposed and could be incorporated into the steps outlined here. 46

The way forward
We hope this stepwise guide will help stakeholders select, implement and monitor in vitro diagnostic tests. Each of the steps outlined should be elaborated into practical guidelines. The comprehensive and accessible online laboratory quality stepwise implementation tool from WHO, 47 which provides medical laboratories with a guide to implementing quality management systems in compliance with ISO (International Organization for Standardization) 15189 standards, may serve as a model.
There have been calls for a model list of essential diagnostics comparable with the Model list of essential medicines maintained by WHO. [48][49][50] The idea was first proposed by WHO in January 2017. 51 Such a list would help in the selection of diagnostic methods and would facilitate improvements in the regulation and affordability of in vitro diagnostic tests and in training in their use. The list should be based on the prevalence, and the relevance to public health, of the diseases considered and not limited to only a few diseases, as are some other regulatory processes. Another challenge is to find the right way of specifying diagnostic tests because several tests might have the same generic name but be very different in terms of quality and performance. Although a model list of essential diagnostics would provide information on diagnostic requirements and test characteristics, it will still be necessary to establish selection criteria for individual tests.
The quality of test evaluations and reports could be improved by establishing a repository of information on, or a central point of assistance for, the design of diagnostic studies. In Africa, the Collaboration for Evidence-Based Health Care in Africa or the African Society for Laboratory Medicine could serve this function. The protocols used for evaluations carried out during WHO's prequalification process should be made publicly available along with information on how other diagnostic test evaluations are taken into account. In addition, peer-reviewed journals should be more rigorous in checking whether STARD criteria have been followed in studies submitted for publication.
Diagnostic tests must be manufactured under strict conditions to ensure their consistent performance. Consistency is crucial and its verification should be taken into account during test selection. Moreover, production lots could be tested independently, as has been proposed for malaria tests by WHO and FIND. 52 Improvements are also needed in the continuous monitoring of test quality and in feedback to end-users. An initial step could be for each laboratory and test centre to have access to internal and external quality controls. Currently, few manufacturers have made positive and negative control materials commercially available for evaluating rapid diagnostic tests and few laboratories or test centres are involved in proficiency testing. Access to information on postmarketing surveillance also needs to be improved; in addition to field safety notices, the results of lot testing and information on complaints about tests should also be made publicly available.
With the guidance presented here and implementation of the improvements suggested, diagnostic testing will have a greater chance of realizing its potential for improving patient care in resource-constrained settings. ■