Xpert MTB/RIF highly sensitive and speeds time-to-initiation of TB treatment among PLHIV in Botswana

“The sensitivity of the [Xpert MTB/Rif TB test] was very high when used for intensified case finding under programmatic conditions in Botswana,” according to Dr. Tefera Agizew, TB-HIV researcher for the Botswana-USA Partnership (BOTUSA), speaking at the 44th Union World Conference on Lung Health held from October 31 to November 4th in Paris.[1]

While South Africa’s well-funded crossover from smear microscopy-based TB diagnosis to using Xpert MTB/RIF as their first-line TB diagnostic has been highly publicised, there are limited reports from other countries in the region attempting to do the same — particularly from the perspective of addressing the case finding challenges among people living with HIV.

Dr Agizew’s presentation, during the first ever TB-HIV late breaker session at the Union meeting, described the results of an ongoing study that evaluated the impact of the switch to using the Xpert MTB/RIF for TB case finding among PLHIV and to see whether it decreased the time to initiation of TB treatment when used in ‘real world’ clinical settings in Botswana. Which it did.

“The [Xpert diagnostic] algorithm reduced time to antituberculosis treatment initiation and rapidly diagnosed rifampicin resistance,” he said. But, he added, difficulty collecting sputum in this population “limits Xpert’s potential to maximize TB case finding, especially for patients with advanced HIV disease.”

Botswana implements phased Xpert MTB/RIF implementation

Botswana has an HIV prevalence of around 17% in the general population and 24% among people of the childbearing age. One of the very first countries to begin offering antiretroviral therapy through its public health programme — first primarily in specialised ART clinics though the country is in the process of decentralising HIV care and treatment.

Botswana also has a high rate of TB — and a very high HIV-TB co-infection rate (around 64% in 2011).

In June 2012, the Botswana Ministry of Health adopted the WHO guidelines for the GeneXpert MTB/RIF test and incorporated it into their TB diagnostic algorithm. It was decided to implement the Xpert MTB/RIF in phases, but a study was designed to assess Xpert performance under programme conditions to inform the country’s further scale-up. There were two major primary objectives in this study:

  1. To evaluate the sensitivity of the new Xpert-based TB diagnostic algorithm for new adult ART enrolees compared to the pre-Xpert smear-microscopy-based algorithm in diagnosing culture-positive TB disease;
  2. To evaluate the impact of the whole “Xpert package” on all-cause mortality during the first 6-months of ART, among adult patients.

Agizew2“We introduced the Xpert diagnostic package which includes Xpert-based ICF algorithm, training of health facility personnel and Xpert device and testing services at point-of-care and other laboratory service sites. We used the stepped wedge design of the phased implementation of the Xpert diagnostic package,” said Dr Agizew.

[A stepped wedge randomised trial evaluates the sequential roll out of an intervention to individuals or clusters (sites) over more than one time period.[2] All participants or sites should have the intervention in place by the end of the study, but the order in which participants or sites receive the intervention is determined at random.]

There were thirteen GeneXpert devices installed to provide diagnostic services to 22 purposefully selected health facilities (four were point of care and the remainder were placed in centralized laboratories). Dr Agizew noted that these sites cover about half of the population living with HIV in Botswana. There were three cohorts in the trial: A prospective pre-Xpert implementation cohort; a prospective, post-Xpert implementation cohort; and a retrospective cohort.

The phased-in implementation has taken place over a 9-month period. The sample size is 9614: 3300 pre-Xpert implementation, and 6314 patients post-Xpert implementation — but the study is still ongoing.

Results thus far

Out of the 4331 patients evaluated so far; 1078 (25%) screened positive for one or more TB symptoms, 125 of whom were subsequently diagnosed with TB (3%). Out of these, 82 have culture confirmed TB of which 58 were diagnosed after having implemented Xpert. 54 of the 58 received positive Xpert results, which yielded a sensitivity of 91% for Xpert in this setting (rather high for an ART programme — however, culture may have missed a significant number of cases because of challenges collecting sputum, and other other types of specimens may not have been collected — see below). Rifampicin resistance was found in 4 patients (7.4 %).

Xpert has shortened the median time to antituberculosis treatment (drug sensitive). In the pre-Xpert algorithm for DS TB smear microscopy, the time to treatment was 9 days (IQR 2-28); and in the post-Xpert algorithm group, the time was 4 days (IQR 2-6). For MDR-TB post-Xpert, the median time to anti-TB treatment initiation was 20 days (IQR N/A).

However, a number of people have died in the study. In the prospective cohort, as of 20 September 2013 — there have been 91 deaths so far, 53 out of 1743 in the pre-Xpert cohort (3%); and 38/2588 (1.5%) of them in the post-Xpert cohort.

“We have not yet confirmed the cause of the mortality at this point,” said Dr Agizew, “but of these 91 deaths, 58 of them – or 64 percent – were screened positive [for TB symptoms] at enrolment. But unfortunately only 36 percent managed to give us a sputum sample. And this is one of the big challenges of GeneXpert. So even if we implement it in a marvellous way, if we don’t get sputum then we’ll fail there.

The median time of death, from enrolment, was 39.5 days IQR 18 – 89.5) and the CD4 was 81(IQR 44 – 167).

“This again is a group with very late presentation and advanced disease. Xpert was not helpful at all for this group,” said Dr Agizew.

Mechanical and operational challenges with implementation

Not everything went smoothly with the introduction of Xpert however. For instance, initially there were high error rates: an average of 8% of the tests — but as high as 15% at one site. Three per cent were invalid results (as high as 11% at some sites).

However, between the first and third quarter of the trial, there was a reduction of invalid results as well as errors. As has been noted in other sites, some of this may have been due to the cartridges in the early stage of the rollout, the country was using version 3 of the Xpert MTB/RIF cartridges but has now switched to version 4 — which seem to be more reliable.

Dr Agizew said that they also encountered many universal power supply (USB) breakdowns. USB is supposed to provide an uninterrupted power supply, but six of the thirteen UPS’s kept failing. Mostly Dr Agizew suspects that there were power outages or surges — the cause is still under investigation.

However, Dr Agizew attributed the errors “mainly due to human factor, and also to some extent the use of version 3 of the cartridges.” There was a high turnover of the lab technicians at the sites — and more training was required to bring new technicians up to speed.

Lessons learned have informed scale-up to additional sites in 2013 

Dr Agizew listed a number of the key operational lessons learnt in his setting:

1) Sputum Collection
“One of the big ones is the low sputum collection rate, and we were at 63%,” he said.

Recently, they’ve increased sputum collection to 74% and their goal is to increase it to more than 90% by taking the following steps: closer follow-up of the patient by study staff, patient coaching during sputum collection, having a poster/or job aide on sputum collection and sputum induction has been introduced now for routine program use.

“The sputum induction was part of our protocol but it was not part of the national TB guidelines, we were just putting it in just in case and mostly we opted to put it for the children’s sake [children with TB have difficulty producing good sputum specimens]. But lately we’ve learnt that if we don’t use sputum induction, we’ll be missing maybe patients who deserve Xpert. And more than two quarters of patients who died could not produce sputum, so what’s the use of Xpert then if we don’t have any sputum?”

[In another Late Breaker we’ll be reporting on, Dr Steve Lawn of the Desmond Tutu HIV Center, made the case for switching to another specimen entirely — urine — which most patients can provide more readily]

2) Failure to adhere to TB diagnostic algorithms:

Dr Agizew also noted that they found clinicians at the clinic were not adhering to the algorithm or felt uncomfortable ordering and reading chest x-rays for symptomatic patients who were smear negative or Xpert negative.

3) Inconsistent screening performed by clinic (non-study) staff:

The study did not train staff at the clinics to perform TB screening — it was assumed that they would just be following the national guidelines as the national programme had trained them to do.

“But they were not doing it,” said Dr Agizew.

4) Ongoing training and supervision are needed to counter the effects of substantial staff turnover/in rotation.

5) A good laboratory quality assurance plan is essential, with good supervision and also emphasising the use of the standard operating procedures.

6) There was also one laptop theft, and other misuse of the laptops in the setting, so the programme is switching over to desktops units (which people are less likely to try to take home.


Perhaps the most critical recommendation Dr Agizew offered was to follow Botswana’s example, and closely study the rollout of this (or any) new tool.

“We recommend countries to consider to evaluate and to monitor early implementation of Xpert to inform the program scale-up,” said Dr Agizew.

Dr Agizew also said programmes may also need to explore alternative reliable power supplies. The UPS they used, didn’t work in Botswana and are very expensive to replace, he said. One option they are considering is to put in solar panels.

Another recommendation he stressed is that programmes need to emphasise and monitor adherence to laboratory standard operating procedures.

“We also recommend to promoting high sputum collection rates, if possible by considering sputum induction for all symptomatic patients.. But most importantly, we need to identify rapid ways to diagnose TB in patients who fail to produce sputum,” he concluded.


[1] Agizew T. Impact of Xpert MTB/RIF on TB Case Finding among PLHIV, Time-to-Initiation of TB Treatment in Botswana. 44th Union World Conference on Lung Health, Paris, 2013, Late Breaker Abstract.

[2] BrownCA, Lilford RJ. The stepped wedge trial design: a systematic review. BMC Medical Research Methodology 2006, 6:54.

Leave a Comment

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.