BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable...
Transcript of BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable...
![Page 1: BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy (background document 1), it is important](https://reader036.fdocuments.us/reader036/viewer/2022081620/6105b8559ff6c663dd579060/html5/thumbnails/1.jpg)
1
BACKGROUND DOCUMENT 4e
Prevalence surveys of TB disease post-2015 Why, How, Where?
Prepared by: Philippe Glaziou, Irwin Law, Babis Sismanidis,
Ikushi Onozaki, Katherine Floyd
Questions for discussion
1. Do you recommend that national prevalence surveys should be conducted
post-2015?
If yes: 2. Do you agree with the suggested updates to survey methods, in particular:
● Use of Xpert® MTB/RIF (hereafter Xpert) instead of smear microscopy
and culture, with adjustments to survey results until Xpert has equivalent
performance to culture;
● Strengthening of overall governance/oversight mechanisms including
more formal arrangements for survey monitoring and related actions by
implementers and sponsors;
● Ensuring Good Clinical Data Management Practices, including quality
control at each stage of data handling to ensure that all data are reliable
and have been processed correctly;
● Investment of more resources in the work required once results are
finalized, especially to ensure the timely production of survey reports and
effective communication of findings and their implications.
3. Do you have any suggestions for other improvements to survey methods?
4. Do you agree with the proposed criteria for identifying which countries
should consider a national survey post-2015, or would you propose
modifications to these criteria?
5. Should there be any country prioritisation within groups 1 and 2, from a
regional and/or global perspective?
![Page 2: BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy (background document 1), it is important](https://reader036.fdocuments.us/reader036/viewer/2022081620/6105b8559ff6c663dd579060/html5/thumbnails/2.jpg)
2
Introduction
National TB prevalence surveys in 22 global focus countries was one of the three major strategic
areas of work of the Global Task Force on TB Impact Measurement during the period 2008–
2015. An additional 31 countries were identified as eligible to implement a survey, based on
epidemiological criteria agreed by the Task Force in December 2007. In 2010, a major
collaborative effort of Task Force members and other contributors (a total of 50 people from 15
institutions) was used to produce an updated WHO handbook on the design, implementation,
analysis and reporting of prevalence surveys (the “Lime Book”) [1]. The Lime Book built on the
first edition of the handbook (the “Red Book”), which was published in 2007 under the
coordination of WHO’s Western Pacific Regional Office. The Lime Book was able to draw on
experience and lessons learned from a growing number of surveys conducted in Asia as well as a
survey in Eritrea, and a recognized need to expand the content of the book and in some instances
provide more definitive recommendations (e.g. on the screening strategy to be used).
Between 2009 and 2015, 19 countries implemented a survey according to the key methods set out
in the Lime Book. These included 14 of the 22 global focus countries,1 plus 5 other countries.
2 A
further two global focus countries started a survey in 2015 (Bangladesh, Kenya), an additional
three will implement repeat surveys in 2016 (the Philippines) or 2017 (Myanmar, Viet Nam), and
two others will likely commence in 2016 (Mozambique and South Africa). These surveys have
provided a substantial new body of evidence on the burden of TB disease in high TB burden
settings, based on direct measurements (see background documents 4a-c).
Building on the results and lessons learned from surveys implemented 2009–2015 and in the
context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy
(background document 1), it is important to consider the role of prevalence surveys in the post-
2015 era: are they still needed, and if yes, where should they be prioritized and should methods
be updated?
This background paper addresses these three questions. It is structured in three parts:
1. Why are prevalence surveys useful? This section explains the rationale for surveys,
illustrated by results and lessons learned from surveys completed 2009–2015.
2. How could survey methods be improved post-2015? This section starts by
summarizing the major challenges faced in surveys implemented 2009–2015, based on
the more detailed assessment provided in background paper 4d. Suggestions for possible
updates to how surveys are done post-2015, to address these challenges and to take
advantage of advances in diagnostic tests and other technologies, are then discussed. The
focus is on four major topics: use of Xpert (or equivalent rapid molecular test) as a
replacement for smear and culture; strengthening overall governance/oversight
mechanisms, especially survey monitoring and the role of the sponsor (funder); ensuring
Good Clinical Data Management Practices; and investment of more resources in the
work required once results are finalized, in particular to ensure the timely production of
survey reports.
3. Where should prevalence surveys be prioritized post-2015? This section proposes
criteria for identifying countries where prevalence surveys should be considered. This is
done separately for a) countries that implemented a survey 2007–2015 and b) countries
that did not implement a survey 2007–2015.
1 In order of when they were implemented: Myanmar (2009-2010); China (2010); Pakistan, Cambodia, Ethiopia
(2011); Thailand, Tanzania, Rwanda and Nigeria (2012); Malawi, Ghana (2013); Indonesia (2013-2014); Zambia
(2014); Uganda (2015). 2 Lao PDR (2010-2011), Gambia (2012), Sudan (2013), Zimbabwe (2014), Mongolia (2015).
![Page 3: BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy (background document 1), it is important](https://reader036.fdocuments.us/reader036/viewer/2022081620/6105b8559ff6c663dd579060/html5/thumbnails/3.jpg)
3
1. Why are prevalence surveys useful?
Measuring the burden of TB disease and monitoring time trends are critical for planning TB
control interventions, assessing their impact on population health and for evaluation of whether
global targets for reductions in disease burden are achieved.
Post-2015 targets for reductions in TB disease burden have been set as part of the United
Nations’ post-2015 SDGs (which cover the period 2016–2030 and replace the MDGs, 2000–
2015) and in WHO’s End TB Strategy (2016–2035). Under the health-related SDG3, target 3.3 is
defined as: “By 2030, end the epidemics of AIDS, TB, malaria and neglected tropical diseases
and combat hepatitis, water-borne diseases and other communicable diseases.” Incidence is the
key indicator that will be used to monitor progress for TB (see also background document 1). The
End TB Strategy includes targets (for 2030 and 2035) and milestones (for 2020 and 2025) for
reductions in TB deaths and TB incidence. The 2030 targets are an 80% reduction in the TB
incidence rate and a 90% reduction in the number of TB deaths, compared with levels in 2015
(for further details, see background document 1).
Although TB disease prevalence is not an indicator explicitly included within the SDGs and End
TB Strategy, WHO guidance on the operationalization of the strategy at country level emphasizes
that measurement of TB prevalence will continue to be relevant in some countries. Part of the
country adaptation that is an explicit principle of the End TB Strategy will need to include
consideration of whether a national (or subnational) prevalence survey is needed to measure TB
disease burden and trends. The rest of this section explains why prevalence surveys will continue
to be relevant post-2015, illustrating the knowledge and insights that they can provide using the
results and lessons learned from surveys implemented 2009–2015 (see also background
documents 4a-c).
1.1 Variability in TB burden
Ideally, nationwide disease surveillance systems should provide direct measurements of TB
incidence. However, most countries with a high TB burden do not yet have notification systems
that capture all cases (see also background document 2a and 2d). In particular, cases diagnosed in
the private sector may not be reported, and health systems in many countries lack the reach and
quality required to ensure that all (or virtually all) cases are diagnosed. Routine surveillance may
also include patients misdiagnosed as TB cases. Given gaps in routine surveillance, the only
way to obtain an unbiased estimate of the burden of TB disease in many endemic countries,
and to monitor trends, is to conduct population-based national surveys of the prevalence of TB
disease.
Methods currently used by WHO [2] to estimate TB incidence can be grouped into four major
categories:
1. Case notification data combined with expert opinion about case detection gaps (120
countries in 2015);
2. Results from national TB prevalence surveys (19 countries in 2015);
3. Notifications in high-income countries adjusted by a standard factor to account for under-
reporting and under-diagnosis (73 countries in 2015);
4. Capture-recapture modelling (5 countries in 2015).
Findings from national TB prevalence surveys provide much more robust estimates of TB
incidence (method 2) compared with the first method (currently, the most commonly used
method in endemic countries) based on case notifications and expert opinion about the plausible
range of the case detection gap, which results in biases about the best estimate of incidence and
often, an over-optimistic assessment of uncertainty about it. Figure 1 compares recent estimates
of TB incidence derived from prevalence survey findings to estimates obtained from eliciting
expert opinion about gaps in case detection for the same year.
![Page 4: BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy (background document 1), it is important](https://reader036.fdocuments.us/reader036/viewer/2022081620/6105b8559ff6c663dd579060/html5/thumbnails/4.jpg)
4
Figure 1. Estimates of TB incidence obtained indirectly (in blue) based on case notifications
and expert opinion about detection and reporting gaps prior to a recent national prevalence
survey, compared with estimates derived from prevalence using survey results (in red). Estimates correspond to the survey year (2012–2015), and are based on recent WHO methods [2]. Indirect
estimates were generally provided by experts with over-confident uncertainty ranges and are more heavily
biased than estimates derived from national survey results.
Figure 2. Estimates of TB prevalence (all ages, all forms of TB) for 18 countries, before (in
blue) and after (in red) results from national prevalence surveys became available,
corresponding to the year of the survey. Panels are ordered according to the size of the before−after
difference. The wide uncertainty interval of the post-survey estimate for the United Republic of Tanzania is
because laboratory challenges meant that it was only possible to directly estimate the prevalence of smear-
positive (as opposed to bacteriologically confirmed) TB. Results for Mongolia were not final by the end of
March 2016 and thus are not shown.
![Page 5: BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy (background document 1), it is important](https://reader036.fdocuments.us/reader036/viewer/2022081620/6105b8559ff6c663dd579060/html5/thumbnails/5.jpg)
5
Disease prevalence also has the advantage of being a direct measure of illness caused by TB in a
population [1,3]. Compared with incidence, prevalence captures another facet of disease burden
by accounting for illness duration. The prevalence of infectious cases also determines how much
transmission takes place in any population, because the annual risk of infection, denoted λ, equals
the number of infectious contacts made by each case per person per year (β) multiplied by the
prevalence of infectious cases (P): λ=βP. Given its dependence on the duration of illness,
prevalence responds more rapidly than incidence to improved case finding and drug treatment,
since both interventions shorten disease duration [4].
Large variation in levels of TB prevalence has been documented in prevalence surveys (Figure
2). Furthermore, surveys in Cambodia, Lao PDR, Myanmar and Viet Nam demonstrated that the
burden was much higher than previously estimated using findings from tuberculin surveys, while
those in Nigeria and Indonesia demonstrated a burden much higher than previously estimated
from case notifications and expert opinion about plausible detection gaps. In some other countries
such as the Gambia, prevalence was measured at a lower level than previously estimated.
1.2 Time trends
When repeat surveys are implemented with intervals of approximately 10 years, trends in
prevalence over time can be assessed. Surveys in Cambodia, China, Republic of Korea and the
Philippines have shown that TB prevalence can be halved in a decade (Figure 3).
Figure 3. Time trends in TB prevalence
1.3 Distribution of disease by age and sex
Prevalence surveys can be used to better understand TB epidemiology (e.g. the distribution of
disease by age, sex and geographical variation). The distribution of prevalent cases by age in
recent surveys is shown in Figure 4, and by sex in Figure 5. Findings show that men are
consistently at a higher risk of TB disease across countries and that age-specific rates tend to be
highest in the elderly.
![Page 6: BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy (background document 1), it is important](https://reader036.fdocuments.us/reader036/viewer/2022081620/6105b8559ff6c663dd579060/html5/thumbnails/6.jpg)
6
Figure 4. Age-specific bacteriologically confirmed prevalence rates in countries that
implemented a national TB prevalence survey, 2009–2015.
Figure 5. Sex ratio of prevalent cases, surveys implemented 2009–2015. Tanzania is not shown
because the number of bacteriologically-confirmed cases could not be verified.
1.4 Gaps in detection and reporting
When measurements of prevalence are compared with notifications, prevalence surveys can
identify gaps in detection and reporting. Ratios of prevalent to notified cases (for smear-positive
cases) are shown in Figure 6. Cross-country and male:female comparisons of the prevalence to
notification (P:N) ratio show that in several countries, it should be possible to achieve much more
with strategies and technologies for TB care and control that are already available, and to close
reporting and detection gaps for men. While the burden of TB disease is much higher in men,
![Page 7: BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy (background document 1), it is important](https://reader036.fdocuments.us/reader036/viewer/2022081620/6105b8559ff6c663dd579060/html5/thumbnails/7.jpg)
7
P:N ratios indicate that women are probably accessing available diagnostic and treatment services
in primary care more effectively. They also indicate that cases among older age groups tend to be
detected less effectively, at least in some countries (Figure 7). Reasons for detection and
reporting gaps, particularly among men, such as access barriers, quality of clinical and diagnostic
services, need to be investigated.
Figure 6: Prevalence to notification ratios by sex in countries where prevalence surveys
were implemented 2009–2015. Following the publication of WHO’s Definitions and Reporting
Framework for Tuberculosis (2013 revision), data on smear-positive notifications disaggregated by sex
have not been systematically collected at the global level since 2013 [5]. For this reason, the P:N ratio
could not be calculated for Sudan, Uganda, Zambia and Zimbabwe. For Ghana and Malawi, smear-positive
notifications disaggregated by sex were obtained from the NTP.
Figure 7. Prevalence to notification ratios by age group (years) in four selected countries.
In Vietnam [6] and Indonesia, records of cases on treatment among participants in national
prevalence surveys could be linked with the records of newly detected cases from routine TB
surveillance. This allowed measurement of the magnitude of under-reporting of detected cases.
Under-reporting from the private sector was found to be more important (over 50%) in Indonesia
than previously anticipated, explaining in part the very large P:N ratio. More surveys in the future
could include record-linkage studies of cases detected before survey investigations with records
![Page 8: BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy (background document 1), it is important](https://reader036.fdocuments.us/reader036/viewer/2022081620/6105b8559ff6c663dd579060/html5/thumbnails/8.jpg)
8
from the national TB database, to assess the level of under-reporting, although inventory studies
(see background document 2d) may generally be a better approach since cases can be better
documented.3
In recent surveys in Africa, it has been possible to measure the prevalence of HIV infection
among prevalent TB cases.4 Findings confirmed a lower prevalence of HIV among prevalent
cases compared with newly notified cases (Figure 8). HIV-infected cases may have TB detected
faster, particularly if already enrolled in HIV care programmes. In addition, their illness duration
is shortened by higher case fatality ratios, particularly in people who are immunocompromised
but not on antiretroviral therapy. These results help better understand the dynamics of HIV-
positive TB.
Figure 8. Prevalence of HIV among prevalent TB cases compared with newly notified cases. The dashed line shows equality. The area of bubbles is proportional to the standard deviation of the survey
estimate. Horizontal segments in blue indicate the mean values of HIV prevalence among prevalent cases
predicted using mixed-effects modelling, and vertical segments their 95% confidence interval.
1.5 Screening for TB
All prevalence surveys indicate a poor sensitivity of symptom screening for TB. Among
bacteriologically confirmed cases, typically 30–50% reported no symptoms meeting survey
screening criteria (Figure 9). This has implications for routine TB detection policies, since many
countries focus their TB detection efforts on chronic coughers.
Commonly used diagnostics, particularly direct microscopic examination of sputum smear
samples, need to be upgraded with better technology, including WHO-approved rapid diagnostics
that are more sensitive and more specific than sputum microscopy. Screening and diagnostic
3 Cases detected before survey investigations are typically not as well documented as survey cases detected during
investigations, particularly in countries where culture or Xpert are not routinely used. 4 HIV testing during field operations was conducted in four countries: Zambia (all participants offered an HIV test);
and Rwanda, Tanzania and Uganda (HIV test offered only to those who screened positive by symptom-based interview
and/or chest X-ray). In Zimbabwe, all bacteriologically confirmed TB cases were offered HIV testing as part of routine
case management, but not directly as part of the survey itself. Instead of offering HIV testing, in Malawi all
participants were asked if they had ever been tested for HIV and, if willing, to state their status. These questions were
also asked in Rwanda, Tanzania and Zimbabwe, but limited to only those participants who screened positive. In
Uganda and Zambia, participants were not asked about a history of being tested for HIV or knowledge of their HIV
status.
![Page 9: BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy (background document 1), it is important](https://reader036.fdocuments.us/reader036/viewer/2022081620/6105b8559ff6c663dd579060/html5/thumbnails/9.jpg)
9
algorithms used in routine clinical care over the past decade need revisiting, and the role of chest
X-ray should be re-evaluated and likely expanded.
Figure 9. Proportion symptom screening negative but chest X-ray positive where
prevalence surveys were implemented 2009–2015.5
1.6 Health care-seeking behaviour
Standardised survey data from multiple countries on patterns of health care-seeking behaviour
(Figure 10, Figure 11) can help to identify actions that NTPs and/or health services in general
could take to shorten the time to TB diagnosis and to ensure prompt provision of high-quality
care. For instance, a large proportion of symptomatic patients took no action (Figure 10) in most
countries, suggesting barriers to access health services. The observed proportion of cases treated
in the private sector (Figure 11) is a useful measure to assess the need for, and coverage of,
public-private mix approaches. All TB care providers need to be engaged, including for case
notification, which should be a mandatory policy that is actively enforced.
Figure 10. Frequency of medical care sought by symptomatic TB cases before the survey
and distribution of types of services
5 Bacteriologically confirmed cases could not be verified in Tanzania.
![Page 10: BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy (background document 1), it is important](https://reader036.fdocuments.us/reader036/viewer/2022081620/6105b8559ff6c663dd579060/html5/thumbnails/10.jpg)
10
Figure 11. Percentage of participants on TB treatment at the time of the survey who were
being treated in the private sector
1.7 Performance of sputum smear microscopy
Sputum smear microscopy was found to be poorly predictive of TB in several surveys (Figure
12), with a large proportion of false positive results even though the population tested had already
been screened for presumptive TB. This highlights the limitation of smear microscopy when used
without confirmation (using another test based on a different technique). In the context of active
case finding, sputum microscopy should be considered an unreliable diagnostic test for TB,
unless high positive predictive values can be demonstrated in the targeted population group.
Figure 12. Frequency of false-positive examination to diagnose TB in selected surveys using
rapid molecular tests. Results are shown for the surveys in which there was systematic use of rapid
molecular tests. In all except one survey, Xpert was used. The exception was Sudan, where LPA was used.
![Page 11: BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy (background document 1), it is important](https://reader036.fdocuments.us/reader036/viewer/2022081620/6105b8559ff6c663dd579060/html5/thumbnails/11.jpg)
11
2. How could survey methods be improved post-2015?
Between 2009 and 2015, 19 national TB prevalence surveys were completed, including 11 in Africa
and 8 in Asia. This is a historically unprecedented effort and achievement, built on substantial
efforts made by countries and their technical and financial partner agencies and advisors, with
overall global guidance and coordination from WHO and with the Lime Book as the key reference
document for all aspects of surveys from design through to reporting of results. Of the 19 surveys,
12 were first-ever national surveys (Ethiopia, Gambia, Lao PDR, Malawi, Mongolia, Nigeria,
Rwanda, Sudan, Tanzania, Uganda, Zambia, Zimbabwe) a further two (Ghana, Indonesia) were the
first in almost 50 years to use the screening methods recommended in the Lime Book, a further one
(Myanmar) was the first using screening methods recommended in the Lime Book and the survey in
Pakistan was the first since 1987. Only Cambodia, China and Thailand had already conducted
surveys in the period since 1990.
A substantial new body of knowledge has been generated by the 19 surveys completed between
2009 and 2015 (section 1). At the same time, given the scale and technical, financial and logistic
demands of conducting surveys with samples sizes typically ranging from 50 000–100 000 people,
and no previous experience of such surveys in most countries, it is inevitable that various challenges
have been faced, despite efforts to standardize methods and to provide as much guidance and
support as possible. For surveys post-2015, it is important that these challenges are clearly
documented, and solutions proposed where possible. Solutions may include, but are not limited to,
the adoption of new technologies.
This section starts by providing a summary of the main challenges that were faced, based on the
more detailed tabular summary shown in background document 4d. This is followed by discussion
of four suggestions for how to improve survey methods post-2015.
2.1 Challenges faced in prevalence surveys implemented 2009–2015
The major challenges faced in surveys implemented 2009–2015, and the number of times they
occurred, are shown in Table 1. Other challenges that occurred but that were not considered to be
“major” challenges are shown in Table 2.6 Counts are based on the more detailed information
shown in background document 4d, for which the main sources were survey monitoring reports and
personal knowledge of the surveys among WHO staff (based on a mixture of first-hand experience
of a survey, information shared during meetings or workshops as well as directly with WHO, the
update of survey status prepared by GTB/TME each quarter, and draft chapters of a book currently
in preparation on prevalence surveys 2009–2015 that includes country-specific chapters within
which major successes and challenges are summarized).7
Major challenges included:
● The time taken to complete survey preparations. This was more than three years in most
countries, with delays primarily linked to delays in the procurement of X-ray equipment, a
need to strengthen laboratory capacity or the time taken to identify a suitable implementing
agency;
● Low participation rates in urban areas;
● Problems with culture examinations. More than 15% of the sputum smear-positive samples
of TB survey cases failed to grow in more than half of all surveys;
● An incomplete data management plan at the time field operations started, and problems
with data management and data cleaning during and after field operations (often linked to
the initial data management plan);
● Delays in case management;
● Delays in producing the final survey report, sometimes (but not always) related to delays in
producing a final, clean dataset.
6 These challenges are not considered “major” in this document either because a) they only affected a few clusters, and/or
b) they could be resolved in a timely manner, and/or c) did not lead to a major bias in final results and estimates. 7 Prior to any wider distribution of the assessment shown in background document 4d, the contents will be reviewed by
survey investigators and those who provided technical assistance.
![Page 12: BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy (background document 1), it is important](https://reader036.fdocuments.us/reader036/viewer/2022081620/6105b8559ff6c663dd579060/html5/thumbnails/12.jpg)
12
Table 1. Major challenges observed in national prevalence surveys completed 2009–2015
Challenges Number of surveys
affected* Preparations Time taken to complete survey preparations (i.e. ≥3 years from decision to
implement to start of field operations 14/19
Delays due to chest X-ray procurement 7/19
Delays due to upgrading laboratory capacity 3/18
Field and central operations Low participation rates (<80%) 5/19
Delays in case management 7/16
Laboratory Low culture confirmation among smear-positive study cases (<85%) 11/19
Failure of biosafety cabinet 1/19
Laboratory protocol violations 6/19
Data management Inadequate data management plan 6/19
Major data management problems during field operations that were not resolved
during field operations 5/19
Considerable time taken to clean data (1+ year) 5/19
Dissemination of results Long delays (1+ year) before survey results accepted by public health authorities 4/19
Long delays in producing the final report of the survey (1+ year) 8/18 * The total number of countries in the denominator varies because information was not available for all surveys at the
time this document was prepared.
Table 2. Other challenges of note that were observed in national prevalence surveys
completed 2009–2015. These challenges are not considered as “major” either because a) they only
affected a few clusters, and/or b) they could be resolved in a timely manner, and/or c) did not lead to a
major bias in final results and estimates.
Challenges Number of surveys
affected Recruitment of survey coordinator and/or data manager 6/19
Access to census data prior to sampling; inaccurate census data or sampling errors;
restricted sampling frame (>5% of national population in eligible age group) 5/19
Issues related to maintenance of X-ray equipment 11/19
Data management problems impacting time to process data, but resolved 5/19
Delays/interruption of 2-3 months during the survey in disbursement of funding 4/19
A few clusters not visited or postponed due to insecurity or adverse weather 10/19
Inadequate corrective action following recommendations of external experts 5/19
Delays in central chest X-ray reading 4/18* * The total number of countries in the denominator varies because information was not available for all surveys at te
time this document was prepared.
It is notable that there were challenges related to data management in each phase of the surveys
(preparations, field operations, post-field operations). These occurred despite an entire chapter
dedicated to the topic in the Lime Book and later updates in the form of web-based material, as
well as direct technical assistance. Examples of problems included data entry errors (including
wrong personal identifiers) and disjointed data tables captured in different systems with no
enforcement of database relations (e.g. laboratory data captured in a database system that was
separate from the system holding clinical data, resulting in an inability to successfully match all
laboratory records with corresponding clinical records), the absence of an audit trail capturing all
![Page 13: BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy (background document 1), it is important](https://reader036.fdocuments.us/reader036/viewer/2022081620/6105b8559ff6c663dd579060/html5/thumbnails/13.jpg)
13
changes made to case report forms and to computer records, and loss of source documents. In a
few cases, these problems required data re-entry in a newly designed database system, leading to
long delays in the production of a final, clean dataset. In a small number of surveys, it was not
possible to check all records against source documents, either because source documents were
not maintained properly and/or because some were lost (e.g. Tanzania), or because identifiers
were wrongly captured in the first place (e.g. Pakistan).
2.2 Possible updates to how surveys could be undertaken post-2015
Given the challenges faced in surveys 2009–2015, consideration of changes to how surveys are
undertaken post-2015 is warranted. Four possible updates that address the biggest challenges
faced are proposed for discussion. These are: 1. Use of Xpert (or equivalent rapid molecular test) instead of smear and culture, alongside
adjustments to results until Xpert tests are comparable to culture;8
2. Strengthening of overall governance/oversight mechanisms including more formal
arrangements for survey monitoring and related actions by implementers and sponsors;
3. Ensuring Good Clinical Data Management Practices[10], including quality control at
each stage of data handling to ensure that all data are reliable and have been processed
correctly;
4. Investment of more resources in the work required once results are finalized, especially
to ensure the timely production of survey reports and effective communication of
findings and their implications.
2.2.1 Use of Xpert (or equivalent rapid molecular test) instead of smear and culture
The gold standard test for prevalence surveys has been culture of M. tuberculosis from sputum
samples. However, culture suffers from four limitations in the context of a survey:
● Test quality is affected by the time to process a sample. Long transportation times from
remote clusters to the designated laboratory put the natural viability of M. tuberculosis at
risk and also increases the risk of sample contamination. Furthermore, if large numbers
of samples are delivered on top of the normal workload for routine (non-survey) testing,
the average processing time may increase. Contamination can be partly controlled
through sample decontamination, but decontamination may kill M. tuberculosis if it is
applied too harshly. Maintaining high quality standards for culture throughout the survey
period and across designated laboratories was problematic in several surveys
(background document 4d);
● Culture quality is generally lower than pre-survey proficiency standards, as illustrated by
the failure to grow more than 15% of the sputum smear-positive samples of TB survey
cases in more than half of recent surveys. This can generate false-negative cases, but it is
difficult to predict in what number or to adjust for this problem after surveys are
completed;9
● Missing culture results are observed in most surveys, creating problems to ascertain
cases. In more recent surveys, such problems have been addressed using Xpert as a
confirmatory test, combined with a panel review of individual cases (including review of
X-rays and other documentation) with missing results.
Xpert was first recommended for use by WHO in December 2010. The use of this test (or
improved versions of Xpert, or equivalent tests) has various advantages compared with culture. It
is rapid, automated, does not require fresh samples to perform optimally and does not require
stringent laboratory containment. Testing without centrifugation has the added advantage of
8 The current WHO policy update for Xpert MTB/RIF states that “Xpert MTB/RIF may be used rather than
conventional microscopy and culture as the initial diagnostic test in all adults suspected of having TB”. Policy
update: Xpert MTB/RIF assay for the diagnosis of pulmonary and extrapulmonary TB in adults and children
http://www.who.int/tb/laboratory/xpert_launchupdate/en/ 9 An extreme case (the only survey where this happened) was Tanzania, where the total number of cases with positive
cultures was less than the number of cases with positive smear results.
![Page 14: BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy (background document 1), it is important](https://reader036.fdocuments.us/reader036/viewer/2022081620/6105b8559ff6c663dd579060/html5/thumbnails/14.jpg)
14
minimising cross-contamination. Surveys which have used Xpert for all smear-positive samples
include Ghana, Malawi, Indonesia, Uganda, Zambia and Zimbabwe. Current surveys in
Bangladesh, Kenya and the Philippines are now using Xpert alongside culture for all samples.
Two possible designs
There are two possible designs for a survey based on testing with Xpert that maintains the
existing screening algorithm based on symptoms and chest X-ray. These are:
1. Chest X-ray and symptom screening, followed by testing of those meeting screening
criteria with Xpert, with no confirmatory test used for Xpert-positive cases.
2. As for 1, plus confirmatory testing of Xpert-positive cases using a different test (e.g.
culture, or test with Xpert on a second sample).
Design 1 may be particularly useful using the upcoming Xpert® MTB/RIF Ultra test. With the
currently available Xpert tests, Design 2 is preferable to Design 1.
Adjustments required to survey results based on Xpert testing
Until a version of Xpert (or equivalent molecular test) has equivalent performance to culture
under optimum conditions, then if Xpert were to be used in prevalence surveys it would be
necessary to adjust survey results to account for false-positive and false-negative results:
● A test with suboptimal specificity (in this case, compared with culture in optimum
conditions) will generate a greater proportion of false-positive results in a survey of the
general population compared with routine clinical investigations. How to account for this
problem when providing test results to patients requires a decision prior to starting the
survey, based on the expected predicted values for a positive test, and following WHO
recommendations for the clinical use of test results.
● False-positive and false-negative results will generate a bias in the estimate of prevalence
in the tested population. This misclassification bias can be quantified and accounted for
to obtain a bias-corrected prevalence estimate from the observed apparent prevalence,
using a Bayesian modelling approach described immediately below that is simple to
implement .
The adjustments to results that would be needed if either of the two designs that include Xpert
testing for all participants that screen positive (on symptoms or chest X-ray) are illustrated below.
Design 1: Chest X-ray and symptom screening, followed by testing of those meeting
screening criteria with Xpert, with no confirmatory test used for Xpert-positive cases
Suppose that a survey identifies n=5000 survey participants who screen positive using a
combination of chest X-ray signs and reported symptoms based on current WHO
recommendations [1]. Further assume that m=150 among them have an Xpert positive test result.
A recent meta-analysis showed that the pooled sensitivity (se) and the 95% confidence interval
for Xpert was 67% (62% to 71%) while the pooled specificity (sp) was 98% (97% to 99%) [7].
These pooled distributions for sensitivity and specificity values may be updated with data
obtained from recent active case finding activities or from survey results where both Xpert and
culture were used and performed to quality standards. It is suspected that sensitivity may be
decreased to some extent in the context of active case finding, due to a higher expected
proportion of paucibacillary cases.
Let denote apparent prevalence
No sampling design effect is assumed when calculating the 95% confidence interval of (simple
![Page 15: BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy (background document 1), it is important](https://reader036.fdocuments.us/reader036/viewer/2022081620/6105b8559ff6c663dd579060/html5/thumbnails/15.jpg)
15
random sampling is assumed).
Let denote bias-corrected (true) prevalence. is expressed in terms of , se and sp, as follows
(1)
In the above numerical example, the measured (apparent) prevalence is severely biased towards
high values, driven upwards by false-positive results.
Both the numerator and the denominator of the right hand side of equation (1) represent variable
quantities, while se and sp are usually found to be slightly variable between populations and
studies; meta-analyses typically report a 95% confidence interval about these quantities.
Uncertainty about se and sp needs to be propagated as well as uncertainty about that is due to
sampling. Propagating uncertainty can be carried out using a simple Bayesian model
implemented in JAGS [8] or in WinBUGS (see model specification in Appendix 1). To set up the
model, one may assume that se and sp follow a Beta distribution, with parameters obtained using
the method of moments [9]. An uninformative prior is set on . The likelihood function
is obtained by solving equation (1) for
(2)
The conditional probability distribution of 𝜙 is proportional to the product of the likelihood and
the prior,
from which the usual summary statistics are extracted
This example illustrates that the bias correction using equation (1) is not sufficient as it yields an
underestimate of disease prevalence among participants with a positive screening result.
To account for sampling design effects in the measurement of , one may define n’ as the
effective sample size (n’ will usually be smaller than n) and calculate m’ so that the ratio m’/n’ =
m/n. The effective sample size is defined as the number tested divided by the sampling design
effect (deff): n’=n/deff. The design effect equals 1 when simple random sampling is used and is
typically greater than 1 when cluster sampling is used.
Design 2: As for Design 1, plus confirmatory testing of Xpert-positive cases using a different
test (e.g. culture, or test with Xpert on a second sample)
If positive Xpert test results are confirmed with a second test, the modelling approach for Design
1 would be modified as follows:
1. define m as the number of Xpert positive cases and a positive confirmatory test
2. set specificity to 100% (Xpert false-positive results are no longer included in m) and
simplify the likelihood function:
With n=5000 tested individuals, and m=150 confirmed cases,
![Page 16: BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy (background document 1), it is important](https://reader036.fdocuments.us/reader036/viewer/2022081620/6105b8559ff6c663dd579060/html5/thumbnails/16.jpg)
16
To account for sampling design effects in the measurement of , one may define n’ as the
effective sample size and calculate m’ so that the ratio m’/n’ = m/n, as for Design 1.
The adjustments required for a third design in which there is no chest X-ray nor symptom
screening is shown in Appendix 2. This may have relevance in countries in which widespread
difficulties in transporting mobile chest X-ray equipment to cluster sites are anticipated (e.g.
those with poor roads such as in the Democratic Republic of the Congo, Mozambique and
Nepal).
Biases introduced by the use of a test that does not perform as well as the recommended gold
standard can be accounted for in a Bayesian model at the cost of increased uncertainty about the
adjusted prevalence estimate, as shown in the numerical examples above (calculations of sample
size may be adjusted upwards to account for the loss in precision). The benefits are decreased
survey time, simpler laboratory procedures and improved quality of laboratory data, and less
uncertainty generated by multiple imputation of missing laboratory data.
It should be emphasized that countries that have implemented a national TB prevalence survey in
the past may not benefit as much from adopting updated designs, since this would introduce a
loss of comparability with results from a previous survey. However, direct microscopic
examination of sputum samples of Xpert-positive cases could be considered if Design 1 were
used, or in confirmed cases if Design 2 was used, to allow valid comparisons in estimates of the
prevalence of sputum smear-positive pulmonary TB.
2.2.2 Strengthen governance/oversight mechanisms including more formal arrangements
for survey monitoring and related actions by implementers and sponsors
In terms of governance/oversight mechanisms relevant to surveys (and many clinical research
studies in general), the following roles and responsibilities can be distinguished:
1. The sponsor(s) provides the financing for a survey. Examples include external agencies
such as the Global Fund, development agencies, or the national government, and may
include a mixture of these. Sponsors may request regular reports from survey
implementing agencies, and reports may be linked to periodic release of funds.
2. The Principal Investigator represents all survey investigators, is in charge of the
recruitment of competent staff, and leads the writing of the final report and scientific
papers.
3. Investigators are responsible for survey design (including the development of a protocol
and standard operating procedures, and ethics review and approval), implementation of
field operations including quality control, analysis of results and preparation of a survey
report. During field operations, this includes ensuring the accuracy, completeness,
legibility, and timeliness of the data reported in data collection tools. Data that are
derived from source documents should be consistent with the source documents or
discrepancies should be explained. To achieve maximum data quality, a standard set of
quality assurance procedures10
should be in place [10]. These include checking that
batches of newly entered records are consistent with defined standards.
4. Survey Monitors assess the implementation of survey operations, including checking
protocol modifications and checking for protocol violations. They may conduct batch
checks of data. They advise investigators about their findings and provide
recommendations for corrective actions if needed. They also report to an Independent
Data Monitoring Committee, and may assist the Principal Investigator to prepare the
10 Quality assurance is a process of systematic activities designed to ensure, assess and confirm the quality of the data
collected during a survey. Quality assured data are data that are suitable for their intended purpose. This includes
accuracy, timeliness, accessibility, comparability. A dataset is accurate to the extent that it is free of errors. Information
in database records should exactly match the corresponding information found in case report forms and source
documents and an audit trail should be maintained when updating records. Data are timely if available at the time it is
needed. Accessibility is determined by the relative ease or difficulty of use. Data are comparable if they are the same
from one unit to another, whether that unit of comparison is between individuals, interviewers, clusters, or even
national prevalence surveys
![Page 17: BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy (background document 1), it is important](https://reader036.fdocuments.us/reader036/viewer/2022081620/6105b8559ff6c663dd579060/html5/thumbnails/17.jpg)
17
final report. In the general context of Good Clinical Practices [10], study monitors
represent the sponsor.
5. An Independent Data Monitoring Committee may be established by the sponsor to
assess the progress of the survey at regular intervals (based on reports from survey
monitors) and to provide recommendations to the sponsor about whether to continue,
modify, or stop the survey.
In relation to these roles and associated responsibilities, the 19 surveys implemented 2009–2015
can be characterized as follows:
● Sponsors included the Global Fund in 15 countries, the US government (via USAID,
PEPFAR and TB CARE) in two countries and national governments (≥50% contribution)
in three countries. Major contributions were provided from other donors in two countries.
● All surveys had at least one Principal Investigator.
● Monitoring was done in a variety of ways. This included independent review of survey
protocols by at least two external experts, and in-country monitoring by one or several
monitors (though they were not called “monitors”), who were either from external
technical agencies (WHO, US-CDC, KNCV, RIT), were people who had previously held
a leading role in a high-quality survey or were independent consultants with extensive
experience and expertise in the topic that they were asked to monitor. Mid-term survey
reviews in which a team of people visited together were undertaken in some countries.
Monitors provided information including recommendations to survey teams, to WHO
and (in some cases) to the sponsor. In most surveys, there was no permanent monitor in-
country for the duration of the survey. Monitors visited regularly and carefully checked
that all recommendations from previous visits had been implemented.
● A formal Independent Data Monitoring Committee was usually not established.
However, all countries established survey steering committees to provide oversight. The
regularity with which these committees met varied between countries. In at least some
countries, the sponsor(s) actively requested formal and regular reporting of findings from
the monitoring of survey operations (e.g. Bangladesh, Cambodia, Myanmar, Pakistan and
Zambia).
In future surveys, possible ways to strengthen governance and oversight mechanisms include:
1. Systematic establishment of a formal Independent Data Monitoring Committee to report
to the sponsor(s), with action taken by the sponsor if recommendations from monitors are
not implemented by investigators.
2. Systematic posting of a permanent external study monitor, who has been trained in Good
Clinical Practices [10], within the survey team at the central level (such as the place
where the survey data are centrally managed) for the duration of the survey until final
reporting. This person would manage monitoring activities, be in regular contact with
Coordinating Investigator(s), ensure that new investigators enrolled during the survey
receive appropriate training and help the Principal Investigator to prepare the final survey
report. The funding required for a permanent study monitor in-country should be covered
by the sponsor (funding for external monitoring is covered by the sponsor). Surveys that
have used an approach similar to this are Cambodia, Indonesia and Lao PDR.
3. Systematic assessment by the sponsor of whether the prerequisites for implementing a
survey (11 prerequisites are clearly defined in the Lime Book) 11
are met, prior to the go-
ahead being given for survey implementation. Based on recent surveys, particular issues
that require scrutiny include the likelihood of sufficiently high participation in urban
areas, especially capital cities; ensuring that all key members of the survey team have the
required qualifications and experience; the data management plan; and access to census
data.
11 These are defined as: 1) strong commitment and leadership from the NTP, Ministry of Health and a core group of
professionals;2) identification of a suitable institute, organization or agency to lead and manage the survey; 3) adequate
laboratory capacity, especially for culture; 4) compliance with the regulations of the national regulatory authority; 5)
reliable and timely procurement and logistics; 6) funding; 7) assurance of security in the field for survey teams and
participants; 8) data management; 9) community participation; 10) expert review and clearance of protocols, including
ethical clearance; 11) external support and technical assistance.
![Page 18: BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy (background document 1), it is important](https://reader036.fdocuments.us/reader036/viewer/2022081620/6105b8559ff6c663dd579060/html5/thumbnails/18.jpg)
18
2.2.3 Ensure Good Clinical Data Management Practices
The strengthened monitoring and associated data quality assurance mechanisms described in
section 2.2.2 should help to resolve many of the data-related challenges faced in recent surveys.
In addition, use of relational databases, field data entry and barcodes should also help improve
the overall quality of data, including minimizing errors in personal identifiers.
2.2.4 Invest more resources in the work required once results are finalized, especially to
ensure the timely production of survey reports and effective communication of findings and
their implications
It has often taken considerable time to produce a survey report (more than one year in eight
countries). The presence of a permanent full-time survey monitor (suggested in section 2.2.2)
could help to address this challenge, since one of their responsibilities would be to provide
regular reports with material that could subsequently be used in the final survey report. More
generally, more resources (people with the right skills and time for report writing, and funding
for production costs including editing and printing) need to be committed to this task when the
survey budget is first developed and approved.
Recent experience in a few countries also highlights the importance of discussing possible survey
results and their implications in advance, and maintaining communication as results emerge with
key decision makers (e.g. planners, policy makers, those with responsibility for communicable
diseases in the Ministry of Health). During discussions, particular emphasis should be given to:
survey validity; quality assurance procedures; monitoring including external monitoring; how
survey findings provide valuable information for decision making on policies, prioritization and
future budgeting for TB control. When to engage with national and local media also needs to be
considered.
The last chapter of the Lime Book, on “Analysis and reporting”, focused on best-practice methods
for the analysis of survey data and how to present results.12
The book does not include a
subsequent chapter on the production of a survey report and communication of results. Such
additional guidance could be valuable to countries implementing surveys in future.
12 This guidance was subsequently updated and published in a journal article: Floyd et al. Analysis of tuberculosis
prevalence surveys: new guidance on best-practice methods. Emerging Themes in Epidemiology 2013, 10:10 013,
http://www.ete-online.com/content/10/1/10
![Page 19: BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy (background document 1), it is important](https://reader036.fdocuments.us/reader036/viewer/2022081620/6105b8559ff6c663dd579060/html5/thumbnails/19.jpg)
19
3. Where should prevalence surveys be prioritized post-2015?
3.1 Criteria used in 2007
The criteria for prioritizing a prevalence survey agreed by the Task Force in December 2007 and
subsequently published in a WHO Policy Paper (2009) and the Lime Book [1] are shown in Table
3. In 2007, there were 53 countries that met these criteria. Among them, a subset of 22 global
focus countries was identified. These global focus countries were selected on the basis of their
estimated share of global and regional TB disease burden and to ensure inclusion of countries
from different parts of Africa. An important reason for defining a subset of focus countries was
the relatively limited availability of technical expertise and experience to support countries to
implement surveys for the first time. This necessitated prioritization of countries given the
impossibility of providing adequate support to up to 53 countries. Following the implementation
of surveys in 19 countries 2009–2015 alongside concerted efforts to build capacity at global,
regional and national levels including via collaboration between implementing countries, there
are now many more people available to provide technical assistance to countries that have not yet
implemented surveys as well as to support repeat surveys. If adopted, the simplifications to
survey design discussed in section 2 should also make it easier to implement surveys.
Table 3. The criteria used to identify countries eligible to conduct a national survey of the
prevalence of TB disease during the period 2008–2015
Criteria Explanations
Group 1
1. Estimated prevalence of smear-positive TB ≥100
per 100 000 population; and 2. Accounts for ≥1% of the estimated total number of
smear-positive TB cases globally; and 3. Case detection rate (CDR) for smear-positive TB
≤50% or >100%
• Major contribution to global burden of TB • Sample size small enough to make surveys feasible in
terms of cost and logistics • Excludes countries whose contribution to the global
burden of TB is insignificant for the purposes of global
and regional assessments of burden and impact • CDR≤50% or >100% indicates weak reporting
systems and problematic TB estimates, respectively
Group 2
1. Estimated prevalence of smear-positive TB≥70 per
100 000 population; and 2. Accounts for ≥1% of the estimated total number of
smear-positive TB cases globally; and 3. Estimated HIV prevalence rate in the adult
population (15 to 49 years)≥1%
Less stringent criteria for the TB prevalence rate, but
incorporates countries with high HIV prevalence and
therefore where there is potential for a rapid increase
in TB incidence and prevalence rates
Group 3
1. Estimated prevalence of smear-positive TB≥200
per 100 000 population; and 2. Accounts for ≥0.5% of the estimated total number
of smear-positive TB cases globally
Less stringent criteria for a country’s contribution to
the global burden of disease, but incorporates countries
with particularly high TB prevalence rates
Group 4
1. Nationwide survey implemented 2000–2007 or 2. Nationwide survey planned before 2010
Prior survey data allow monitoring of trends. High
motivation of NTP to conduct a survey
![Page 20: BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy (background document 1), it is important](https://reader036.fdocuments.us/reader036/viewer/2022081620/6105b8559ff6c663dd579060/html5/thumbnails/20.jpg)
20
3.2 Proposed epidemiological criteria for prioritizing a national prevalence survey
post-2015
Post-2015, there are two major dimensions that need to be considered when determining whether
a TB prevalence survey should be prioritized.
1. The added value of survey results. This will be bigger in settings where the available
routine surveillance data on TB (from notification and vital registration systems) are
poorly informative. In countries that have implemented a survey in the past, a repeat
survey will provide not only a point estimate and insights into the current burden of TB
disease and how to address it, but will allow measurement of trends, which can be used
to make inferences about the impact of interventions.
2. Survey feasibility. This includes the required sample size and logistics, expected
participation rate and representativeness, and completeness and quality of data.
These factors and associated criteria can be considered separately for a) countries that conducted
a survey between 2007 and 2015 and b) countries that did not implement a survey between 2007
and 2015.
Table 4 shows the proposed epidemiological criteria for prioritizing a national prevalence survey
post-2015, for these two groups of countries. Figure 13 shows the countries that implemented a
survey between 2007 and 2015. Figure 14 shows the Group 2 countries that did not implement a
national survey between 2007 and 2015 and that meet the criteria for implementing a survey
post-2015. These countries are almost exclusively in Africa, plus India, Afghanistan and Papua
New Guinea.
Table 4. Suggested epidemiological criteria for assessing whether a country could consider
implementing a prevalence survey post-2015, for two major groups of countries
Criteria Explanations
Group 1 - Countries that conducted a national prevalence survey, 2007-2015* (Figure 13)
1. Estimated prevalence of bacteriologically
confirmed TB ≥2.5 per 1000 population aged ≥15
years during the previous survey: and
2. >5 years since the last survey.*
• Sample size small enough (less than 70,000
individuals) to make surveys feasible in terms of cost
and logistics;
• Time between surveys sufficient to allow a
statistically meaningful comparison of prevalence.
Group 2 - Countries that did not implement a national prevalence survey 2007–2015 (Figure 14)
1. Estimated TB incidence** ≥ 1.5 per 1000
population/year (all forms, all ages); and
2. No nationwide vital registration system with
standard coding of causes of deaths; and
3. Infant mortality rate > 10/1000 live births.
• Sample size** small enough (less than 70,000
individuals) to make surveys feasible in terms of cost
and logistics, accounting for added uncertainty due to
the use of rapid molecular tests with performance that
may be inferior to culture;
• No reliable direct measurement of TB disease burden;
• Indirect indicator of low access to quality health
services as defined in the Standards and Benchmarks
for TB surveillance and vital registration [11].
* Surveys conducted prior to 2000 may lack comparability with surveys implemented according to the screening and
diagnostic algorithm recommended in the Lime Book. A WHO workshop held in Cambodia in 2012 recommended a
period of 7–10 years between two surveys. Designs 1-3 (section 2.2) of the planned survey may be adapted to include
microscopic examination of smears performed in laboratory confirmed cases to allow comparability of results with the
previous survey. ** Country-specific prevalence estimates may not be published by WHO post-2015, except for countries with
prevalence survey results. For sample size determinations, prevalence in the ≥15 years age group may be predicted
from incidence.
![Page 21: BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy (background document 1), it is important](https://reader036.fdocuments.us/reader036/viewer/2022081620/6105b8559ff6c663dd579060/html5/thumbnails/21.jpg)
21
Figure 13. Countries that conducted a national prevalence survey, 2007-2015 (see Appendix
3).13
Figure 14. Countries that did not conduct a national prevalence survey 2007–2015 that meet
the proposed Group 2 criteria for prioritizing a survey (see Appendix 3).
For any country that meets the epidemiological criteria shown in Table 4, it is then crucial to
assess the feasibility of a survey, using the prerequisites for implementing a survey defined in the
Lime Book as a framework. For a survey to be feasible, the following are necessary:
1. There is strong commitment and leadership from the NTP, Ministry of Health and a core
group of professionals;
2. A suitable institute, organization or agency to lead and manage the survey can be
identified;
3. There is adequate laboratory capacity;
4. X-ray equipment can comply with the regulations of the national regulatory authority;
5. Reliable and timely procurement and logistics is possible;
6. Funding is available;
7. Security in the field for survey teams and participants can be assured;
8. Data management can be done according to recommended standards;
9. Community participation is likely to be sufficiently high, including in urban areas;
13
It is already recognised that China and Thailand are unlikely to conduct another national survey given
the relatively low burden in terms of rates, the possibility of direct measurement of trends from
surveillance data and the expected low participation among urban/mobile populations. A subnational
survey could be considered e.g. Western China, North Eastern Thailand. Similarly, a repeat survey in
Gambia and Rwanda may also be affected by a low burden of TB.
![Page 22: BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy (background document 1), it is important](https://reader036.fdocuments.us/reader036/viewer/2022081620/6105b8559ff6c663dd579060/html5/thumbnails/22.jpg)
22
10. External support and technical assistance are available if needed. This is likely to be
especially important for countries implementing a survey for the first time.
With respect to the expected participation rate, it is worth noting that while several countries have
successfully implemented repeat prevalence surveys [12–14], surveys were discontinued in the
Republic of Korea as a result of: (i) lower survey acceptability in an increasingly urbanized and
modern environment and associated reductions in participation rates; and (ii) the large sample
sizes required to obtain a prevalence estimate with satisfactory precision when prevalence had
fallen to much lower levels [14]. If it is anticipated that participation will not be sufficiently high
in urban areas, and especially if the share of the population living in urban areas has increased
since the last survey, then a survey may be ruled out based on this criterion alone. A good recent
example is the survey in Thailand, where participation was very low (below 50%) in Bangkok. If
available, recent observations from active case finding for TB, particularly in urban
environments, may provide an indication of likely participation rates. Other large scale health-
related surveys may also provide an indication of the likely level of survey participation.
![Page 23: BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy (background document 1), it is important](https://reader036.fdocuments.us/reader036/viewer/2022081620/6105b8559ff6c663dd579060/html5/thumbnails/23.jpg)
23
Appendix 1. JAGS/WinBUGS model specification
Design 1, screening and no confirmation of positive tests
model {
m ~ dbin(theta, n)
theta <- se*phi + (1-sp)*(1-phi)
se ~ dbeta(291.9, 143.8)
sp ~ dbeta(767.34, 15.66)
phi ~ dbeta(1, 1)
}
Designs 2-3, with or without screening, all positive tests are confirmed
model {
m ~ dbin(theta, n)
theta <- se*phi
se ~ dbeta(291.9, 143.8)
phi ~ dbeta(1, 1)
}
![Page 24: BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy (background document 1), it is important](https://reader036.fdocuments.us/reader036/viewer/2022081620/6105b8559ff6c663dd579060/html5/thumbnails/24.jpg)
24
Appendix 2: Adjustments required to a survey without chest X-ray or symptom
screening that relies on Xpert testing of all survey participants, with confirmatory
test of Xpert positive results 14
Design 3: No chest X-ray or symptom screening, with all participants tested using Xpert
followed by use of a different confirmatory test (or test with Xpert on a second sample) for
participants with Xpert-positive results
A similar analytical approach to that used for Design 2 can be used when X-ray screening is not
feasible or where TB investigations are added to a health survey. The implications of a low
positive predictive value of the test used in a population that is not screened (therefore less likely
to have TB) would likely require the use of confirmatory tests for case management purposes. If
positive Xpert test results are confirmed with a second test, the modelling approach of Design 2
is used:
1. define m as the number of Xpert-positive cases who also had a positive confirmatory test
(using a different testing technique)
2. set specificity to 100% (Xpert false-positive results are no longer included in m) and
simplify the likelihood function:
3. define n’ as the effective sample size of the survey (the effective sample size is the
number of eligible people who were tested with Xpert divided by the sampling design effect) and
calculate m’ so that m’/n’ = m/n
With an effective sample size n’=20000, and m’=120 confirmed cases (adjusted to effective
sample size),
14 Confirmatory tests may not be necessary if the main test is Xpert Ultra or a new rapid test with high specificity.
![Page 25: BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy (background document 1), it is important](https://reader036.fdocuments.us/reader036/viewer/2022081620/6105b8559ff6c663dd579060/html5/thumbnails/25.jpg)
25
Appendix 3:
3a: Countries that have conducted a national TB prevalence survey from 2007-2015,
and their contribution to the global burden (incidence) of TB.
Country
Proportion (%) of estimated global
incidence (2014)
Bangladesh 3.7
Cambodia 0.6
China 9.6
DPRK 1.1
Ethiopia 2.1
Gambia <0.1
Ghana 0.5
Indonesia 10.5
Kenya 1.1
Lao PDR 0.1
Malawi 0.4
Mongolia 0.1
Myanmar 2.0
Nigeria 5.9
Pakistan 5.2
Philippines 3.0
Rwanda 0.1
Sudan 0.4
Thailand 1.2
Uganda 0.6
UR Tanzania 1.8
Viet Nam 1.3
Zambia 0.7
Zimbabwe 0.4
![Page 26: BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy (background document 1), it is important](https://reader036.fdocuments.us/reader036/viewer/2022081620/6105b8559ff6c663dd579060/html5/thumbnails/26.jpg)
26
3b: Countries that did not conduct a national prevalence survey during 2007–2015 that
meet the proposed Group 2 criteria for prioritizing a survey, and their contribution to
the global burden (incidence) of TB.
Country
Proportion (%) of estimated global
incidence (2014)
Afghanistan 0.6
Angola 0.9
Bhutan 0.0
Botswana 0.1
Central African Republic 0.2
Cote d'Ivoire 0.4
Cameroon 0.5
DRC 2.5
Congo 0.2
Djibouti 0.1
Micronesia <0.1
Gabon 0.1
Guinea 0.2
Guinea-Bissau 0.1
Equatorial Guinea <0.1
Haiti 0.2
India 22.4
Kiribati <0.1
Liberia 0.1
Lesotho 0.2
Madagascar 0.6
Marshall Islands <0.1
Mozambique 1.6
Namibia 0.1
Nepal 0.5
Papua New Guinea 0.3
Sierra Leone 0.2
Somalia 0.3
Swaziland 0.1
Chad 0.2
Timor-Leste 0.1
Tuvalu <0.1
South Africa 4.7
![Page 27: BACKGROUND DOCUMENT 4e Prevalence surveys of TB ......context of a new era of the Sustainable Development Goals (SDGs) and the End TB Strategy (background document 1), it is important](https://reader036.fdocuments.us/reader036/viewer/2022081620/6105b8559ff6c663dd579060/html5/thumbnails/27.jpg)
27
References
1 WHO. Tuberculosis prevalence surveys: a handbook. 2011.
http://www.who.int/tb/advisory_bodies/impact_measurement_taskforce/resources_documen
ts/thelimebook/en/
2. Glaziou P, Sismanidis C, Pretorius C, Floyd K. Methods used by WHO to estimate the
Global burden of TB disease. arXiv [q-bio.QM]. 2016; published online March 1.
http://arxiv.org/abs/1603.00278.
3 Glaziou P, van der Werf MJ, Onozaki I, et al. Tuberculosis prevalence surveys: rationale
and cost. Int J Tuberc Lung Dis 2008; 12: 1003–8.
4 Dye C, Bassili A, Bierrenbach AL, et al. Measuring tuberculosis burden, trends, and the
impact of control programmes. Lancet Infect Dis 2008; 8: 233–43.
5 World Health Organization. Definitions and Reporting Framework for Tuberculosis, 2013
revision, updated December 2014. http://www.who.int/tb/publications/definitions/en/
6 Hoa NB, Cobelens FGJ, Sy DN, Nhung NV, Borgdorff MW, Tiemersma EW. Diagnosis
and treatment of tuberculosis in the private sector, Vietnam. Emerg Infect Dis 2011; 17:
562–4.
7 Walusimbi S, Bwanga F, De Costa A, Haile M, Joloba M, Hoffner S. Meta-analysis to
compare the accuracy of GeneXpert, MODS and the WHO 2007 algorithm for diagnosis of
smear-negative pulmonary tuberculosis. BMC Infect Dis 2013; 13: 507.
8 Plummer M, Others. JAGS: A program for analysis of Bayesian graphical models using
Gibbs sampling. In: Proceedings of the 3rd international workshop on distributed statistical
computing. Technische Universit, Wien, Austria, 2003: 125.
9 Renyi A. Probability Theory. New York: Dover Publications Inc, 2007.
10 International Conference on Harmonisation Working Group. ICH harmonised tripartite
guideline: guideline for good clinical practice E6 (R1). In: International Conference on
Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human
Use. Washington, DC, 1996.
11 WHO. Standards and Benchmarks for tuberculosis surveillance and vital registration
systems. WHO, 2014
http://apps.who.int/iris/bitstream/10665/112673/1/9789241506724_eng.pdf?ua=1.
12 Onozaki I, Law I, Sismanidis C, Zignol M, Glaziou P, Floyd K. National tuberculosis
prevalence surveys in Asia, 1990-2012: an overview of results and lessons learned. Trop
Med Int Health 2015; 20: 1128–45.
13 Wang L, Zhang H, Ruan Y, et al. Tuberculosis prevalence in China, 1990-2010; a
longitudinal analysis of national survey data. Lancet 2014; 383: 2057–64.
14 Hong YP, Kim SJ, Lew WJ, Lee EK, Han YC. The seventh nationwide tuberculosis
prevalence survey in Korea, 1995. Int J Tuberc Lung Dis 1998; 2: 27–36.