Annex 1. Methods for evidence reviews and modelling

(1)

1

Annex 1

Annex 1.

Methods for evidence reviews and modelling

Questions for the 2011 update of the Guidelines for the programmatic management of drug-resistant tuberculosis (for Outcomes please see Table 2 in the main text of the Guidelines)

1.

1. At what prevalence of MDR-TB in any group of TB patients is rapid drug-At what prevalence of MDR-TB in any group of TB patients is rapid drug- susceptibility testing warranted to detect resistance to rifampicin and isoniazid susceptibility testing warranted to detect resistance to rifampicin and isoniazid or rifampicin alone on all patients in the group at the time of TB diagnosis, in or rifampicin alone on all patients in the group at the time of TB diagnosis, in order to prescribe appropriate treatment at the outset?*

order to prescribe appropriate treatment at the outset?*

2.

2. Among MDR-TB patients receiving appropriate treatment in settings withAmong MDR-TB patients receiving appropriate treatment in settings with reliable direct microscopy, is monitoring using sputum smear microscopy reliable direct microscopy, is monitoring using sputum smear microscopy alone rather than smear and culture, more or less likely to lead to the outcomes alone rather than smear and culture, more or less likely to lead to the outcomes listed in Table 2 of the guidelines?**

listed in Table 2 of the guidelines?**

3.

3. When designing regimens for patients with MDR-TB, is the inclusion of When designing regimens for patients with MDR-TB, is the inclusion of specific drugs (with or without documented susceptibility) more or less likely specific drugs (with or without documented susceptibility) more or less likely to lead to the outcomes listed below?***

to lead to the outcomes listed below?***

4.

4. When designing regimens for patients with MDR-TB, is the inclusion of fewerWhen designing regimens for patients with MDR-TB, is the inclusion of fewer drugs in the regimen (depending on the drug used, the patient’s history of its drugs in the regimen (depending on the drug used, the patient’s history of its use and isolate susceptibility) more or less likely to lead to the outcomes listed use and isolate susceptibility) more or less likely to lead to the outcomes listed in Table 2 of the guidelines?***

in Table 2 of the guidelines?***

5.

5. In patients with MDR-TB, is shorter treatment, compared with the duration In patients with MDR-TB, is shorter treatment, compared with the duration currently recommended by WHO, more or less likely to lead to the outcomes currently recommended by WHO, more or less likely to lead to the outcomes listed in Table 2 of the guidelines?***

listed in Table 2 of the guidelines?***

6.

6. In patients with HIV infection and drug-resistant TB receiving antiretroviralIn patients with HIV infection and drug-resistant TB receiving antiretroviral therapy, is the use of drugs with overlapping and potentially additive toxicities, therapy, is the use of drugs with overlapping and potentially additive toxicities, compared with their avoidance, more or less likely to lead to the outcomes compared with their avoidance, more or less likely to lead to the outcomes listed in Table 2 of the guidelines?****

listed in Table 2 of the guidelines?****

7.

7. Among patients with MDR-TB, is ambulatory therapy compared with inpatientAmong patients with MDR-TB, is ambulatory therapy compared with inpatient treatment, more or less likely to lead to the outcomes listed in Table 2 of the treatment, more or less likely to lead to the outcomes listed in Table 2 of the guidelines?*****

guidelines?*****

Reviewers of evidence

* Menzies D and Oxlade O, McGill University, Montreal, Quebec, Canada

** Mitnick C. D. et al, Harvard University, Boston, Massachusetts, United States

*** Menzies D and Bauer M, McGill University, Montreal, Quebec, Canada

**** Arentz M et al, University of Washington and University of California (San Francisco), United States

***** Fitzpatrick C et al, Stop TB Dept, Geneva, Switzerland

WHO/HTM/TB/2011.6a

(2)

2

Annex 1

1. At what prevalence of MDR-TB in any group of TB patients is rapid drug-susceptibility testing warranted to detect resistance to rifampicin and isoniazid or rifampicin alone on all patients in the group at the time of TB diagnosis, in order to prescribe appropriate treatment at the outset?

Descriptions of models and data

Overview

Hypothetical cohorts of newly diagnosed cases of sputum smear-positive active TB were followed through initial treatment and retreatment (if necessary). Failures and relapses of retreatment were given no further treatment. Two settings were examined:

no HIV and high HIV (HIV in active TB cases = 50%). In each setting, the cohort was stratified by HIV status (if appropriate) and underlying initial drug resistance. Three different initial drug resistance profiles were considered: (i) low MDR-TB and high mono-isoniazid; (ii) high MDR-TB and high mono-isoniazid; and (iii) high MDR-TB and low mono-isoniazid. Specific point estimates of initial prevalence of drug resistance were taken from specific countries that most closely match these scenarios, namely (i) Cambodia, (ii) the Russian Federation and (iii) South Africa. These scenarios are summarized in Table 1.

Table 1. Initial drug resistance scenarios considered in primary analysis

Drug resistance scenario (country used for point

estimates)

Initial drug resistance (%) (data from 4th annual WHO report (1))

Susceptible Mono-H Mono-R Mono-E Mono-S PDR MDR low MDR-TB and high

mono-H (Cambodia) 89.7% 4.7% 0.5% 0.0% 3.3% 1.8% 0%

high MDR-TB and high mono- H (Russian Federation)

64.7% 2.3% 0.2% 0.0% 7.2% 10.6% 15%

high MDR-TB and low

mono-H (South Africa) 92.1% 2.6% 0.3% 0.0% 1.7% 1.5% 1.8%

H = isoniazid; R = rifampicin; E = ethambutol; MDR = multidrug-resistance; PDR = polydrug-resistance, defined as a case with resistance to two or more drugs, but not to rifampicin (i.e. any resistance to rifampicin was grouped with MDR); S = streptomycin

Note: In all models, all single drug resistance and poly drug resistance was grouped together and considered as mono-H resistance, except mono-R which was grouped with MDR

(3)

3

Annex 1

We considered two broad groups of drug sensitivity tests (DST) that are WHO endorsed–based on how quickly results are obtained. These are traditional cultures and DST on solid media (such as LJ)–which provide results after 3 months, and rapid tests–

considered all tests which can provide results within 2 weeks or less. These include culture based tests such as MODS or TLA or automated liquid culture systems using direct cultures. Rapid tests also include amplification tests such as the Line Probe Assay or the GeneXpert, which provide results within 1-2 days. We did not consider these rapid tests individually, as we considered that the difference in providing results between one and 14 days to be clinically non-significant. Among the rapid tests we considered separately tests that provide results for R only and tests that can detect both H and R resistance. As well, we compared outcomes if the rapid DST was applied pre- treatment to all patients, or only for those who were still smear-positive after 2, 3 or 5 months of treatment. We also considered rapid DST for all patients who failed or relapsed–prior to retreatment.

For each type and timing of DST scenario, and each drug resistance setting, 1000 newly diagnosed active cases were followed through multiple courses of treatment. In each scenario we assumed that once DST was done, treatment was guided by, and appropriate for, each result. It was only when DST was not done (or the acquisition of drug resistance occurred after the DST has been done) that patients could be inappropriately treated.

Depending on the scenario, and type of drug resistance detected, two complete courses of treatment were administered. The treatment regimens used in each DST scenario are described below in more detail.

Outcomes predicted by the model include cure, failure, relapse, death, and acquired drug resistance. Disability adjusted life years (DALYs) were also predicted. Cost outcomes estimated include total costs for each DST strategy, cost per MDR-TB case prevented, cost per TB related death averted and cost per DALY averted. Transmission and subsequent secondary cases were not estimated.

Models were developed using decision analysis models and TreeAge Professional software. No discounting of outcomes and costs was applied, as the time frame of the analysis was limited to a single round of initial treatment and retreatment (if necessary).

Treatment regimens

Initial treatment regimens and outcomes without drug susceptibility testing

When DST was not performed, patients received the WHO recommended standard initial regimen (2HRZE/4HR). During initial treatment new cases can cure, die or fail.

Cases that cure can subsequently relapse. Default is not explicitly included as separate treatment category; however, treatment outcomes have been adjusted to take default rates into account. For HIV negative cases, except for those with underlying MDR,

(4)

4

Annex 1

the proportion of the hypothetical cohort that die during initial treatment is based on the most recent WHO global data (2). HIV uninfected MDR cases have the same spontaneous cure rate as untreated cases–25% (3), and annual mortality of 33% (4), since they are effectively untreated. Failure and relapse rates vary based on underlying drug resistance, according to a recent systematic review and meta-analysis of published randomized trials (5,6). Cases with underlying mono-H resistance or MDR have an increased likelihood of failure, relapse, death and acquired drug resistance as shown in table 2a.

Table 2a. Treatment outcomes after standardized initial treatment by underlying drug resistance

Cure Fail Relapse after cure Death during initial

treatment (6 months) DS Mono-H MDR DS Mono- H MDR DS Mon-o H MDR DS Mono- H MDR 96.2% 93.4% 25% 0.3% 3.1% 58.5% 3.1% 6.6% 2.5%/yr 3.5% 3.5% 16.5%

DS = drug sensitive; H = isoniazid; MDR = multidrug-resistance

Sources of data for drug-sensitive and mono-H treatment outcomes:

t Failure and Relapse (5) (6)

t Death: adapted from: WHO Global Report 2009 (1): Figure 1.26: Treatment outcomes for HIV-positive and HIV-negative patients

t Cure – complementary value (meaning 100% – [Failure + Death]) Sources of data for MDR treatment outcomes:

t Fail – complementary value t Relapse (3, 7)

t Death – as if untreated (data from pre-antibiotic era) (4) t Cure – equivalent to spontaneous resolution)(3)

Retreatment regimens and outcomes with no DST

If DST was not performed for cases who are smear-positive at 5 months or relapses, all failures or relapses after treatment completion (regardless of underlying drug resistance) received WHO standardized retreatment (“Category 2” which is 2SHRZE/1HRZE/5HRE). Retreatment outcomes varied according to underlying drug resistance as shown in table 2b. Cases with mono-H resistance or MDR had worse retreatment outcomes. Failures and relapses of standard retreatment were assumed to receive no further treatment.

(5)

5

Annex 1

Table 2b. Treatment outcomes after retreatment by underlying drug resistance^a

Cure Fail Relapse after cure Death during re-

treatment (8 months) DS Mono-H MDR DS Mono-H MDR DS Mono-H MDR DS Mono- H MDR

93.7% 90.9% 0 0.3% 3.1% 78% 3.1% 6.6% 2.5%/yr 6% 6% 22%

aSources of data: as per initial treatment

DS = drug sensitive; H = isoniazid; MDR = multidrug-resistance

Initial and retreatment regimens and outcomes following drug susceptibility testing

If DST was performed, patients received an appropriate regimen once the DST results are available. Patients with drug sensitive (DS) TB continued the WHO recommended standard initial regimen. Patients with mono-H resistant TB start (or restart-depending on the timing of DST) a 6-month regimen in which isoniazid (H) was replaced by levofloxacin throughout (2LfxRZE/4LfxR). Patients with MDR received a standardized 24 month regimen (6 Amikacin-Cs-Eth-Lfx-PZA-EMB/ 18 Cs-Eth-Lfx-PZA-EMB). If DST was performed prior to or during initial treatment and mono-H is detected, then for those who relapse or fail after the Levofloxacin regimen then they receive a standard retreatment regimen. Failures and relapses of standard retreatment were assumed to receive no further treatment. If MDR was previously detected and standardized MDR treatment given,patients will not have further treatment, even if they fail or relapse. The regimens given to different types of cases are summarized in Table 3.

Table 3. Summary of initial and retreatment regimens following drug susceptibility testing

Initial drug resistance Initial regimen Retreatment regimen

Drug sensitive Standard WHO 6-month

initial regimen: 2HRZE/4HR

WHO Cat II

retreatment regimen:

2SHRZE/1HRZE/5HRE Detected mono-isoniazid resistance As above, with Lfx replacing H.

2LfxRLZE/4LfxR

WHO Cat II

retreatment regimen:

2SHRZE/1HRZE/5HRE Detected multidrug-resistance Standardized WHO MDR 24

month regimen. 6 Amikacin- Cs-Eth-Lfx-PZA-EMB/ 18 Cs-Eth-Lfx-PZA-EMB

No retreatment

Treatment outcomes for mono-H cases treated with levofloxacin are assumed to be the same as those shown for drug sensitive cases given the WHO initial regimen (as per Table 2a). Retreatment outcomes are the same as those shown for mono-H cases in Table 2b. Treatment outcomes for MDR cases under appropriate therapy are summarized below in Table 4.

(6)

6

Annex 1

Table 4. Treatment outcomes for detected cases of multidrug-resistance (MDR) on standardized MDR treatment

Cure Fail Relapse after cure Death during treatment

67% 20% 3% 13%

HIV infection

The setting of South Africa was used to demonstrate the impact of HIV on predicted outcomes. Models were modified and results rerun with 50% of TB cases being infected with HIV. We assumed that the ability of each test to detect drug resistance will be the same in HIV positive or negative cases. We also assumed that during initial and retreatment treatment outcomes, other than death (see below), and including acquired drug resistance are the same for HIV positive and negative cases. These and other major assumptions in the model are summarized in Section titled Model Assumptions.

Death rates are assumed to differ between HIV positive and negative TB cases, particularly HIV cases with undetected MDR who will have higher mortality than HIV negative cases treated with inadequate regimens. Death rates for HIV positive cases with underlying DS and mono-H disease were obtained from WHO Global report 2009 (Figure 1.26. Treatment outcomes for HIV-positive and HIV-negative patients) (2). Undetected coinfected MDR cases are assumed to have 100% mortality and thus have no chance to spontaneously cure. Failure and relapse rates for HIV positive cases are assumed to the same as HIV negative cases, however they are adjusted for the higher mortality in HIV positive cases.

When MDR is detected in HIV positive cases, and co-infected cases are started on the standardized MDR regimen, the cure rate is assumed to be 29% (based on data from Galvani et al. ). The remainder of coinfected cases are assumed to die.

ART is not explicitly modelled, but taken into account by using treatment outcomes taken from settings where ART is provided. Costs associated with ART are not considered

Underlying initial drug resistance

In each setting, patients are stratified according to 3 different drug resistance profiles. The 3 initial drug resistance profiles were selected to represent 3 different country specific scenarios. Countries selected and their corresponding initial drug resistance profile (as reported by WHO (1)) are as follows: (i) Cambodia – high mono-H resistance (9.8%) and low MDR (0.5%); (ii) South Africa – low mono-H resistance (5.8%) and mid-range MDR (2.1%); and (iii) and the Russian Federation – high mono-H resistance (20.1%)

(7)

7

Annex 1

and high MDR (15.2%) (see Table 1). In all models, drug resistance was categorized as follows: susceptible to all; mono-isoniazid (H) resistance; or multidrug-resistance (MDR), defined as resistance to at least isoniazid (H) and rifampicin (R). All other mono-drug resistance and poly-drug resistance (PDR) was grouped with mono-H resistance, except mono-rifampicin (R) resistance, which was grouped with MDR. This simplification is for two reasons. First, it makes model construction and interpretation easier. More importantly, we are only modelling the detection of H and/or R resistance.

Hence treatment can not be adjusted in the presence of other resistance, so in all models these forms of resistance will receive the same management. Mono-R was considered the same as MDR because the key determinant of outcomes in MDR is the R resistance.

Modelling the acquisition of drug resistance

Cases with undetected underlying drug resistance that are not treated appropriately can acquire additional drug resistance during each course of treatment (8). Data on rates of acquisition of drug resistance for cases undergoing initial or retreatment with different categories of underlying initial drug resistance are taken from recent systematic reviews (5, 6, 9). Tables 5 and 6 provide a summary of acquisition of drug resistance rates.

Cases with undetected mono-H resistance can acquire MDR during standardized initial and retreatment. If there is a delay in performing DST, or obtaining DST results, then cases can also acquire resistance during that interval while being treated with the standardized initial regimen, particularly if they have H-mono-resistance. Other forms of drug resistance are not considered, nor whether the resistance pattern of MDR-TB could be further worsened (amplified) by standardized initial and retreatment, since we are considering tests that detect H and R only. These assumptions hold for all cases, regardless of HIV status.

Table 5a. Probability of failures acquiring drug resistance after a standardized 6-month regimen, by initial drug sensitivity^a

Initial drug resistance category Probability of drug resistance among failures after treatment Sensitive (%) Mono-H (%) MDR (%)

Sensitive to all 38.5 44.2 17.3

Mono-H – 0 100.0

a Derived from published reviews (5, 9); these estimates are from additional secondary analysis that were not published.

H = isoniazid; MDR = multidrug-resistance

(8)

8

Annex 1

Table 5b. Probability of relapses having different types of drug resistance after a standardized 6-month regimen, by initial drug sensitivity^a

Initial drug resistance category Probability of drug resistance among relapses after treatment Sensitive (%) Mono-H (%) MDR (%)

Sensitive to all 94.1 3.7 2.3

Mono-H – 100 0

a Derived from published reviews (5, 9); these estimates are from additional secondary analysis that were not published.

H = isoniazid; MDR = multidrug-resistance

Table 6a. Probability of failures with mono-isoniazid (H) resistance, treated with levofloxacin, or standard retreatment acquiring multidrug-resistance^a

Initial drug resistance category Probability of drug resistance among failures after treatment

Mono-H (%) MDR (%)

Mono-H (receiving levofloxacin regimen and standardized

retreatment)

82.7 17.3

a Rates as per drug sensitive in Table 5a.

Table 6b. Probability of relapses with mono-isoniazid (H) resistance, treated with levofloxacin, or standard retreatment acquiring multidrug-resistance (MDR)^a

Initial drug resistance category Probability of drug resistance among relapses after treatment

Mono-H (%) MDR (%)

Mono-H (receiving levofloxacin regimen and standardized

retreatment)

97.7 2.3

a Rates as per drug sensitive in Table 5a.

Interventions for drug susceptibility testing

Use and type of drug susceptibility testing

Three types of DST, all performed pre-treatment, were compared to status quo scenario:

1. Status quo scenario: no DST and all patients receive the same standardized treatment 2. Solid DST: DST detecting R and H and performed in all patients at the time of

diagnosis. Results obtained in 3 months. Appropriate treatment delayed for 3 months.

(9)

9

Annex 1

3. Rapid DST: DST detecting R only and performed in all patients at the time of diagnosis. Results and appropriate treatment immediate (as above all tests providing results within 2 weeks were considered as rapid tests).

4. Rapid DST: DST detecting R and H and performed in all patients at the time of diagnosis. Results and appropriate treatment immediate.

For all 3 types of DST the tests were assumed to have 100% sensitivity and specificity for H and/or R resistance (100% accuracy), regardless of HIV status. No other drug resistance is detected.

Timing of drug susceptibility testing

In order to weigh the benefits and costs of performing DST in more patients, against the risks from delaying appropriate therapy, four time points of test administration were evaluated only for the use of a rapid DST test detecting R and H.

1. Baseline – Rapid DST (detecting H and R) is performed prior to starting initial treatment: In this scenario, all 1000 cases will incur the cost of the DST test.

Treatment to all patients will be given appropriately based on DST.

2. Rapid DST (detecting H and R) performed only if smear-positive after 2 months of treatment. In this scenario, only 18% of patients will be smear-positive after 2 months, and thus fewer will require DST. However appropriate treatment will be delayed and treatment outcomes will be less favorable for those with underlying H-R or MDR. Those with underlying mono-H resistance may have acquired further resistance, and those with MDR may have died.

3. Rapid DST (detecting H and R) performed only if smear-positive after 3 months of treatment. Only 8% of patients will be smear-positive after 3 months and need DST, however outcomes will be even worse for those with underlying drug resistance.

4. Rapid DST (detecting H and R) performed only if smear-positive after 5 months of treatment. Only failures will be smear-positive, so DST will be needed for only 4–8%; however, outcomes will be much worse.

5. Rapid DST (detecting H and R) performed only if smear-positive after 5 months treatment (failures), or with relapse after treatment (i.e. all retreatment cases have DST). Defaulters are taken into account through treatment outcomes, but are not explicitly modelled. DST will be needed for 10–20% (10) of all patients who are treated. Treatment outcomes during initial treatment will be worse for those with H-R or MDR. Those with underlying mono-H resistance may have acquired further resistance, and those with MDR may have died.

(10)

10

Annex 1

Costs to the health system

The direct costs considered in this analysis include the test cost for DST, drug costs, as well as DOT and medical follow up health-system costs associated with initial, retreatment, H-R and MDR regimens. DST costs used in the analysis were derived from county specific data provided by Hojoon Soon from FIND/McGill University.

DST costs are summarized in Table 7.

Table 7. Cost of drug susceptibility testing (DST) test

DST strategy Cost of test (US$)

No DST –

Solid DST 20

Rapid HR 40

Rapid R 40

HR = isoniazid plus rifampicin; R = rifampicin

In sensitivity analysis the cost of the DST tests was varied in order to estimate the impact of different costs for the rapid tests as these costs are still not well established.

The details of this and other sensitivity analyses are described in the Section titled Sensitivity analyses.

Drug costs for each treatment regimen were calculated using individual drug costs provided by the drug procurement agency the Global Drug Faculty (11). Unit cost for DOT and medical follow up were obtained from a recent survey of health-system costs in Brazil (12). These unit costs were extrapolated for the 6-month initial treatment regimen, 8 month retreatment regimen and standard MDR treatment regimens. All of these direct health-system costs are summarized in Table 8 below.

Costs were originally obtained from direct surveys conducted in Brazil, but are assumed to be appropriate for South Africa in this analysis, as the GDP per capita (PPP) for the two countries were almost identical in 2007 (13).

(11)

11

Annex 1

Table 8. Details of regimens and associated costs (USD$ 2007) of treatment and costs to the health system Regimen Duration of regimenRegimen costaSingle visit to doctor (2007 USD) # of visits to doctor per regimen (1 per month)Single DOT costb (2007 USD)# of DOT visits per regimenTOTAL health- system cost Initial standardized 6mth2RHZE/4RH blisters = $16.62$22.706$7.52=5 visit/wk for 8.7 wks =3 visit /wk for 17.3 wks Total=95 visits

$867 Retreatment standardized 8mth2RHZES/1RHZE/5RHE blisters, water for inj, 2 syringes = $29.43

$22.708$7.52=5 vst/wk for 8.7 wks (injectable) =44 injectable visits =5 visit /wk for 4.3 wks =3 visit /wk for 21.7 wks =87 non injectable visits Total=131 visits

$1527 H-R regimen- substitute Quinolone for H

6mthInitial regimen, with Levoquin (Lfx) 2RLfxZE/4RLfx = $43.33

$22.706$7.52=5 visit /wk for 8.7 wks =3 visit /wk for 17.3 wks Total=95 visits

$894 Standardized MDR24mths6Cm/18CsLfxEthPAS no syringe, no water for inj = $2,118.79

$22.7024$7.52=5 visit /wk for 26 wks (injectable) =130 injectable visits =5 visit /wk for 70 wks =350 non injectable visits Total=480 visits

$7250 a GDF costs–estimated by Thomas Moore, staff member of GDF b Data obtained from survey of health-system costs conducted in Rio de Janeiro, Brazil, 2007 (12). Costs were originally obtained from direct surveys conducted in Brazil, but are assumed to be appropriate for South Africa, as the GDP per capita (PPP) for the two countries was almost identical in 2007 (13).

(12)

12

Annex 1

Incorporating disability-adjusted life-years

The health impact of disability associated with any disease or injury is summarized by the disability-adjusted life-year (DALY), a metric that combines the burden of mortality and morbidity (non-fatal health problems) into a single number (14). Regardless of the characteristics (e.g. sex, race) of the population under consideration, a “disability weight” is generated for everyone living in a year in a specified health state.

A disability weight is a weight factor that reflects the severity of the disease on a scale from 0 (perfect health) to 1 (equivalent to death). In most situations, disabilities with a larger impact have a larger disability weight. The disability weights used for the Global Burden of Disease estimates for 2004 for active Tuberculosis was 0.271 (with a range of 0.264 to 0.294 with different ages) (14).

Evidently, DALY is usually used as a negative concept, where a larger DALY equates to a larger burden of disease. However, in cost-effectiveness studies, the effectiveness measure is a positive concept where a larger value is better. As such, WHO recommends that the complement of the disability weights are used in cost-effectiveness studies, such that “perfect health” is valued at 1 and “death” is valued at 0. In this way, the positive health impact of an intervention compared with another is expressed as DALYs averted (15).

An example is provided below to illustrate how DALYs are calculated:

For an individual who lived for 40 years with perfect health (complement of disability weight = 1) and 10 years with TB (complement of disability weight = 1- 0.271 = 0.729), the estimated number of DALYs would be (40 x 1) + (10 x 0.729) = 47.29. Now, suppose TB treatment was given to the individual, they cured after 2 years, and the individual died at the age of 80 from a non-TB related cause. In this scenario, the person would live for 78 years with perfect health and only 2 years with active TB, which is equivalent to 79.458 DALYs. Consequently, 79.458 – 47.29 = 32.168 DALYs were averted by providing the TB treatment.

DALYs are provided as an additional outcome for all base case scenarios and some sensitivity analyses.

Description of scenarios

Primary analysis

A brief description of each scenario considered is provided below. A summary of the treatment regimens given pre- and or post- DST, but underlying drug resistance are provided in Table 9 at the end of this section.

(13)

13

Annex 1

a. Scenario 1 – Status quo

No DST conducted. All cases given standard initial and standard retreatment regardless of underlying drug resistance.

b. Scenario 2 – Testing for isoniazid and rifampicin susceptibility on solid media (solid DST)

All cases are given DST, which detects R and H resistance, at the time of diagnosis.

However there is a 3-month delay before DST results are obtained, so treatment with standard initial treatment is given, while awaiting results. Once DST results are obtained cases that were initially mono-H are restarted on a 6-month regimen containing levofloxacin (2LfxRZE4LfxR) and those that initially had MDR are switched to the standard 24-month MDR treatment. Any acquired drug resistance during the 3-month standardized initial regimen will not be detected; hence these patients are treated based on their drug resistance at the time of the DST (i.e. initial drug resistance). Those with DS or Mono-H who fail, or relapse will receive standardized retreatment. Those who had initial MDR were given the MDR regimen. If MDR cases fail or relapse after MDR treatment, no further treatment is given.

c. Scenario 3 – Rapid test that detects R resistance only

All cases given DST at the time of diagnosis, which detects R resistance only. All cases, except those with R resistance, would be given a standardized initial and retreatment regimen. For those cases with R resistance, the standard MDR regimen would be given.

d. Scenario 4 – Rapid test with DST that detects H and R resistance

All cases given DST at time of diagnosis, which detects R and H resistance. Drug sensitive cases are given standard initial therapy, mono-H cases are given the levofloxacin regimen and MDR cases are given the standard MDR regimen. Drug sensitive and mono H cases that fail or relapse are given standardized retreatment. If MDR cases fail or relapse after MDR treatment, no further treatment is given.

Timing scenarios – secondary analysis

e. Scenario 5a – Perform rapid HR DST- if smear-positive after 2 months of treatment

All cases are first given a standardized regimen for 2 months. A rapid DST is performed only for those who are still smear-positive after 2 months treatment. Mono-H cases are then switched to the levofloxacin regimen, and MDR cases to the standard MDR regimen. Those who have acquired drug resistance during the first two months will be detected and put on appropriate therapy. Standardized retreatment is given to any who fail or relapse, except MDR patients that receive standard MDR therapy–and no further therapy if they fail or relapse. ADR that develops after the two months DST, or during subsequent retreatment is not detected.

(14)

14

Annex 1

f. Scenario 5b – Perform rapid HR DST- if smear-positive after 3 months of treatment

All cases are first given a standardized regimen for 3 months. A rapid DST is then performed only for those who are still smear-positive after 3 months of treatment.

Mono-H cases are switched to a levofloxacin regimen, and MDR cases to a standard MDR regimen. Those who have acquired drug resistance during the first 3-month treatment will be detected and put on appropriate therapy. Those that fail or relapse subsequently receive standardized retreatment, except MDR patients who fail or relapse after standard MDR therapy do not get any further treatment. ADR that develops after the first 3 months, or during subsequent retreatment is not detected.

g. Scenario 5c – Delay rapid HR DST until 5 months post diagnosis – failures only All cases are started on standardized initial therapy. A rapid test is then performed on those who are still smear-positive at 5 months time (i.e. failures). Once DST is done, mono H cases are switched to a levofloxacin regimen, and MDR cases changed to a standard MDR regimen. Those who have acquired drug resistance during the partial regimen will be detected and put on appropriate therapy. DS cases (there will be very very few) receive standardized retreatment. ADR that develops during Levo or retreatment is not detected.

h. Scenario 5d – Delay test until 5 months post diagnosis – failures and relapses All cases started on standardized initial therapy. A rapid test performed on those who are smear-positive at 5 months time (i.e. failures) and all relapses. Once DST is done those with DR switched to appropriate regimen. Mono-H cases are switched to a levofloxacin regimen, and MDR cases to a standard MDR regimen. Those who have acquired drug resistance during the initial treatment will be detected and put on appropriate therapy.

DS cases (there will be very few) receive standardized retreatment. ADR that develops during levofloxacin or retreatment is not detected.

(15)

15

Annex 1

Table 9. Summary of regimens by drug resistance and scenario for drug susceptibility testing ScenarioDSMono-HMDR Pre-DSTPost-DSTPre-DSTPost-DSTPre-DSTPost-DST 1. Status Quo- No DST1. Standard initial tx 2. Standard retxN/A1. Standard initial tx 2. Standard retxN/A1. Standard initial tx 2. Standard retxN/A 2. Solid DST1. Standard initial tx2.Remainder of Standard initial tx 3.Standard retx

1. Standard initial tx1.Levofloxacin 2.Standard retx1. Standard initial tx2. Standardized MDR 3. Rapid RN/A1.Standard initial tx 2.Standard retxN/A1.Levofloxacin 2.Standard retxN/A1.Standardized MDR 4. Rapid HRN/A1.Standard initial tx 2.Standard retxN/A1.Levofloxacin 2.Standard retxN/A1.Standardized MDR 5a. Rapid HR- 2 mths1. Standard initial tx2.Remainder of standard initial tx 3.Standard retx

1. Standard initial tx2.Levofloxacina 3.Standard retx1. Standard initial tx2.Standardized MDR 5b. Rapid HR- 3 mths1. Standard initial tx2.Remainder of standard initial tx 3.Standard retx

1. Standard initial tx2.Levofloxacina 3.Standard retx1. Standard initial tx2.Standardized MDR 5c. Rapid HR- 5 mths- Failures only1. Standard initial tx2.Standard retx1. Standard initial txFor Failures: 2.Levofloxacinb For Relapses: 2. Standard retx

1. Standard initial txFor Failures: 2.Standardized MDR For Relapses: 2.Std Retx 5d. Rapid HR- 5 mths- Failures and relapses1. Standard initial tx2.Standard retx1. Standard initial tx2.Levofloxacinb1. Standard initial tx2. Standardized MDR DS = drug sensitive; DST = drug susceptibility testing; HR = isoniazid plus rifampicin; MDR = multidrug-resistance a Mortality with levofloxacin regimen is 0% because we assumed that all deaths would occur during the first round of initial treatment bMortality with levofloxacin as second regimen is 6%, so it is comparable with retreatment

(16)

16

Annex 1

Model costs (US$) by strategy

a. Strategy 1 – Status quo

Costs for DS, mono H, MDR:

t for those who die during initial treatment- get 2 months of costs of standard initial regimen= $289.07

t all cases that survive get full cost of standard initial regimen- $867.22

t all cases that fail or relapse after initial treatment get full cost of retreatment regimen- $1527.03

b. Scenario 2 – Testing for isoniazid and rifampicin susceptibility on solid media (solid DST)

DS:

t all cases get DST cost at time of diagnosis =20$

t those who die during first 3 months of treatment get 2months of costs for drug sensitive regimen (assume death happens at end of first 2 months of treatment)= $289.07

t those who survive get full cost of drug sensitive regimen =$867.22

t all drug sensitive failures and relapses of initial therapy get cost of full second regimen= $1527.03

mono-H:

t all cases get DST cost at time of diagnosis = 20$

t those who die during partial therapy get 2 months of costs for drug sensitive regimen (assume death happens at end of first 2 months of treatment)= $289.07 t those who survive and H resistance is detected get partial cost of drug sensitive

regimen (ie 3 months)= $433.61 and then switch to Levofloxacin regimen =

$893.93

t if H resistance is not detected get remainder of drug sensitive regimen cost=$433.61

t all mono H failures and relapses of initial therapy get std retreatment cost

=$1527.03 MDR:

t all cases get DST cost at time of diagnosis = 20$

t those who die during partial therapy get 2 months of costs for drug sensitive regimen= $289.07

t those who survive get partial cost of drug sensitive regimen (ie 3 months)=

$433.61, and then switch to standard MDR regimen =$ 7250.79 c. Scenario 3 – Rapid test that detects R resistance only

DS:

t all cases get cost of DST test=40$

(17)

17

Annex 1

t all drug sensitive cases that survive get standard initial regimen= $867.22 t all drug sensitive cases that fail or relapse after initial treatment get standard

Retreatment regimen= $1527.03 mono-H:

t all cases get DST cost = 40$

t all mono-H cases that survive get std initial regimen: $867.22

t all mono-H cases that fail or relapse after initial treatment get full cost of retreatment regimen: $1527.03

Initial MDR:

t all cases get DST cost at time 0 = 40$

t for those who die during treatment- get 2 months of costs of standard MDR regimen= 625$

t all MDR cases that survive get full cost of standard MDR regimen =$ 7250.79 d. Scenario 4 – Rapid test with DST results for isoniazid and rifampicin

DS:

t all cases get cost of DST test=40$

t all drug sensitive cases that survive get standard initial regimen= $867.22 t all drug sensitive cases that fail or relapse after initial treatment get standard

Retreatment regimen= $1527.03 mono-H:

t for those who die during initial treatment- get 2 months of costs of Levofloxacin regimen=298$

t all mono H resistant cases that survive full get cost of Levofloxacin regimen=

$893.93

t all mono H failures and relapses of initial therapy get std retreatment cost

=$1527.03 MDR cases:

t for those who die during treatment- get 2 months of costs of standard MDR regimen= 625$

t all MDR cases that survive get full cost of standard MDR regimen =$ 7250.79 e. Scenario 5 –: Delay test until 2 months post diagnosis

DS:

t cases that survive until 2 months of therapy and still smear-positive get cost of DST test =40$

(18)

18

Annex 1

t for those who die during partial- get 2 months of costs for DS regimen=

$289.07

t all living DS failures and relapses of initial therapy get cost of full second regimen= $1527.0

mono-H:

t cases that survive until 2 months of therapy and still smear-positive (more Mono-H will be smear pos than DS cases) get cost of DST test=40$

t those who die during partial therapy get 2 months of costs for DS regimen=

$289.07

t those who survive get partial cost of DS regimen (ie 2 months)= $289.07 and then switch to Levofloxacin regimen = $893.93

t mono H failures and relapses of initial therapy get std retreatment cost

=$1527.03 MDR:

t cases that survive until 2 months of therapy get cost of DST test (all MDR will be smear-positive)=40$

$289.07

t those who survive get partial cost of DS regimen (ie 3 months)= $289.07, and then switch to standard MDR regimen =$ 7250.79

f. Scenario 5b – Delay test until 3 months post diagnosis

DS:

$289.07

mono-H:

t mono H cases that survive until 3 months of therapy and still smear-positive get cost of DST test=40$

$289.07

=$1527.03 MDR:

t MDR cases that survive until 3 months of therapy get cost of DST test (all MDR will be smear-positive)=40$

(19)

19

Annex 1

$289.07

t those who survive get partial cost of DS regimen (ie 3 months)= $433.61 and then switch to standard MDR regimen =$ 7250.79

g. Scenario 5c – Delay test until 5 months post diagnosis – failures only DS:

$289.07

H:

t mono H cases that survive until 5 months of therapy and still smear-positive get cost of DST test=40$

$289.07

t those who survive get partial cost of DS regimen (ie 5 months)= $722.68 and then switch to levofloxacin regimen = $893.93

=$1527.03 MDR:

t MDR cases that survive until 5 months of therapy get cost of DST test =40$

$289.07

h. Scenario 5d – Delay test until 5 months post diagnosis – failures and relapses DS:

t drug sensitive cases that survive until 5 months of therapy complete and still smear-positive get cost of DST test =40$

$289.07

mono-H:

t mono-H cases that survive until 5 months of therapy and still smear-positive get cost of DST test=40$

(20)

20

Annex 1

$289.07

=$1527.03 MDR:

t MDR cases that survive until 5 months of therapy get cost of DST test =40$

$289.07

Model assumptions

General

t cohort considered will have newly diagnosed active TB cases t all cases are smear-positive

t all cases are M. Tuberculosis t all cases have pulmonary disease

t No DST will be the base case scenario in order to maximize expected differences between base case and alternative DST scenarios

t models do not take into account secondary cases arising from transmission from active cases

t costs and outcomes are not discounted as time frame is assumed to be short

HIV-related

t the ability of each test to detect drug resistance will be the same in HIV positive and negative cases.

t during initial and retreatment the acquisition of drug resistance will be the same for HIV positive and negative cases.

t The proportion of cases that remain smear-positive following 2, 3 or 5 months of initial therapy is assumed to be the same in HIV positive and negative cases t Undiagnosed and hence inappropriately treated MDR- TB cases with HIV will

have 100% case fatality.

t Co-infected cases with drug sensitive or mono-H resistance treated with standardized initial therapy have higher mortality than HIV negative cases (approximately 5X as per WHO data)

t ART is not explicitly modelled, but taken into account by using treatment outcomes from settings where ART is available

t Costs associated with ART are not considered

(21)

21

Annex 1

Classification of drug resistance

t in all models drug resistance was categorized as follows: susceptible to all, mono isoniazid (H) resistance or multidrug-resistance (MDR).

t MDR is defined as resistance to at least isoniazid (H) and rifampicin (R), or mono-rifampicin resistance.

t all other types of mono drug resistance or poly-drug resistance (except mono-R) will be considered as mono H.

Drug sensitivity tests

t tests have 100% sensitivity and specificity for the type of resistance they are designed to detect.

t in primary analysis (not secondary analysis that consider different timing of DST), DST is performed only once at the time of initial diagnosis.

t DST is for H and R first-line drugs only. DST for EMB or PZA and second- line drugs are not considered. Hence MDR treatment must be standardized.

(Acquisition of other forms of resistant e.g. ETH, PZA or SLD not considered.) t rapid DST gives results within 2 weeks–this has no impact on clinical outcomes

Treatment

t following DST 100% of cases will start appropriate treatment immediately.

t in each scenario treatment will be guided by, and appropriate for, each result.

It is only when DST is not done that patients may be inappropriately treated.

t when standardized initial treatment is given it is assumed to be the Category I WHO regimen (2HRZE/4HR)

t when standardized retreatment is given it is assumed to be the Category II WHO regimen (2HRZES/1HRZE/5HRE)

t Mono-H resistant cases (if detected) will receive 2LfxRZE/4LfxR–which will have equivalent efficacy as Cat1 in DS cases.

t when an MDR standardized treatment is given it is assumed to be the following- 6CmCsLfxEthPAS / 18CsLfxEthPAS

t default is not included as a treatment outcome, however cure and fail rates are adjusted to account for defaulters

t optimal programme assumed–all cases are assumed to remain on treatment (i.e. no default)

t optimal programme assumed–all relapses and failures are detected and will enter retreatment

t Adverse events taken into account through costs only

Acquisition of drug resistance:

t during inappropriate treatment only two forms of drug resistance can be acquired; pan-sensitive strains can become mono-H resistant or MDR, and

(22)

22

Annex 1

mono-H resistance can become MDR. (as this is all that can be detected with tests being evaluated).

Costs:

t ambulatory health-system costs only t hospitalization costs excluded

t out-of-pocket patients costs excluded

t indirect costs associated with mortality excluded

t indirect costs associated with disability/sickness excluded

DALYs

t DALYs are only included for TB related morbidity and mortality t HIV-related DALYs are not considered in any analyses

t calculations assume that the mean age of developing active TB disease is 35 years (as per IPD data)

t life expectancy estimates are based on WHO estimates for those who live to age 35

Sensitivity analyses

All sensitivity analyses were conducted using the drug resistance profile of South Africa because it has a moderate initial prevalence of MDR (1.8%) and a mono H resistance of 6%. Sensitivity analyses were restricted to the two highest ranking scenarios: (Rapid DST for HR pre-treatment in all patients; and Rapid DST for HR for those who are smear-positive after 2 months of initial treatment).

The following key parameters were varied in sensitivity analysis:

1. Varying prevalence of TB cases co-infected with HIV from 0% to 85%. This analysis was conducted in order to understand how results may vary in different settings of HIV.

2. Varying the prevalence of initial MDR from 0% to 30%. This analysis was conducted in order to understand how results may vary in different settings of initial MDR prevalence.

3. Varying the prevalence of initial mono-H from 0% to 40%. This analysis was conducted in order to understand how results may vary in different settings of initial monoH prevalence.

4. Varying the effectiveness of Levofloxacin from baseline to a worst case scenario for mono-H cases (i.e. outcomes for mono H when on standard initial treatment). This parameter was varied because it is a key assumption in the analysis, yet there is only limited data available to support the assumption.

5. Varying the cost of rapid HR test from 1$ to 1000$. This parameter was varied to assess the impact on each scenario of using different DST methods with different

(23)

23

Annex 1

costs. There is great uncertainty about the real costs to implement any of the new rapid DST methods. The true costs, in LMIC, of the equipment, construction, training, supplies, services, maintenance and repairs for liquid culture methods (Bactec, or MGIT), Nitrate Reductase Assay (NRA), Microscopic Observation Drug susceptibility (MODS), Thin layer agar (TLA), Line Probe Assay (LPA), or GeneXpert are largely unknown. Hence the cost for the rapid test cost was varied widely to understand how results could vary if additional implementation costs were included.

6. Excluding DOT and medical follow up cost. This analysis was conducted in order to get a clearer understanding to what extent patient medical visits and DOT costs were driving results.

7. DST test sensitivity and specificity. Lower sensitivity would result in MDR and H cases being treated with standardized treatment. Missed MDR cases would have increased mortality and missed mono H cases will have increased acquired drug resistance. Reduced specificity (false positive tests) is assumed to result in some unnecessary treatment of DS and mono H resistant cases for MDR-TB; this will result in increased cost and increased default rates given the lengthy treatment plus known toxicities. Default rates will be increased to those reported in MDR trials (23%, as per preliminary results of the Menzies IPD meta analysis). However the failure and death rates will remain the same as drug sensitive and mono H cases treated with standard initial treatment.

Published estimates of DST test sensitivity and specificity used to derive ranges using in sensitivity analysis are summarized in Table 10.

Table 10. WHO-endorsed MDR diagnostic tools

Test family WHO endorsed tests in the family

Sens for H Spec for H Sens for R Spec for R Rapid

cultures Liquid culture and DST 100 (if done on smear+)

100 (if done on smear+)

100 (if done on smear+) Non-commercial rapid

cultures

MODS NRA

94–99 95–98

88–99 99–100

94–99 95–98

95–100 99–100 Molecular

assays Line probe assays (Hain GenoType MTBDRplus)

76–90 97–100 95–99 97–99

Automated NAAT (Xpert) NA NA 94–99 96–99

[after : Ling D et al. ERJ 2008; Minion J et al. Lancet Infect Dis 2010; Boehme CC et al.

NEJM 2010 ;WHO. Policy statement on non-commercial and DST methods, 2010]

Ranges of DST test sensitivity and specificity used in sensitivity analysis are shown in Table 11. Ranges will be applied to the best test scenario of Rapid HR at diagnosis.