Iem et al. BMC Infectious Diseases (2023) 23:341 BMC Infectious Diseases https://doi.org/10.1186/s12879-023-08330-9 RESEARCH Open Access Pooling sputum testing to diagnose tuberculosis using xpert MTB/RIF and xpert ultra: a cost-effectiveness analysis Vibol Iem1,2*, John S. Bimba3, Victor S. Santos4,5, Jose Dominguez6, Jacob Creswell7, Silaphet Somphavong8, Tom Wingfield2,9, Jahangir A. M. Khan10† and Luis E. Cuevas2† Abstract Background The World Health Organization (WHO) recommends the diagnosis of tuberculosis (TB) using molecular tests, such as Xpert MTB/RIF (MTB/RIF) or Xpert Ultra (Ultra). These tests are expensive and resource-consuming, and cost-effective approaches are needed for greater coverage. Methods We evaluated the cost-effectiveness of pooling sputum samples for TB testing by using a fixed amount of 1,000 MTB/RIF or Ultra cartridges. We used the number of people with TB detected as the indicator for cost- effectiveness. Cost-minimization analysis was conducted from the healthcare system perspective and included the costs to the healthcare system using pooled and individual testing. Results There was no significant difference in the overall performance of the pooled testing using MTB/RIF or Ultra (sensitivity, 93.9% vs. 97.6%, specificity 98% vs. 97%, p-value > 0.1 for both). The mean unit cost across all studies to test one person was 34.10 international dollars for the individual testing and 21.95 international dollars for the pooled testing, resulting in a savings of 12.15 international dollars per test performed (35.6% decrease). The mean unit cost per bacteriologically confirmed TB case was 249.64 international dollars for the individual testing and 162.44 international dollars for the pooled testing (34.9% decrease). Cost-minimization analysis indicates savings are directly associated with the proportion of samples that are positive. If the TB prevalence is ≥ 30%, pooled testing is not cost-effective. Conclusion Pooled sputum testing can be a cost-effective strategy for diagnosis of TB, resulting in significant resource savings. This approach could increase testing capacity and affordability in resource-limited settings and support increased testing towards achievement of WHO End TB strategy. Keywords Tuberculosis, Xpert MTB/RIF, Xpert Ultra, Pooling, Cost-effectiveness, Cost-minimization analysis † Jahangir A. M. Khan and Luis E. Cuevas Joint senior authors. *Correspondence: Vibol Iem vibol.iem@lstmed.ac.uk Full list of author information is available at the end of the article © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Iem et al. BMC Infectious Diseases (2023) 23:341 Page 2 of 11 Introduction resources savings, the diagnostic accuracy, the cost to Tuberculosis (TB) was the second leading cause of detect one person with bacteriologically confirmed TB, death by an infectious disease after Coronavirus disease and the potential increase in testing capacity and TB case 2019 (COVID-19) [1]. The World Health Organization detection. (WHO) recommends to provide upfront molecular tests (mWRDs) for the diagnosis of TB and at least rifampi- Materials and methods cin resistance to all individuals with presumptive TB In this cost-effectiveness analysis, a total of 3,076 individ- [2]. mWRDs include the Xpert MTB/RIF [3] (MTB/ uals with presumptive TB were enrolled from two stud- RIF) and Xpert MTB/RIF Ultra (Ultra) [4], which are ies conducted in Lao PDR (840 individuals per study) [8], semi-automated and simultaneously detect Mycobac- two studies in Nigeria (500 individuals per study) (Bimba terium tuberculosis complex and markers of rifampi- et al. in press), and one study in Brazil (396 individuals) cin resistance using the GeneXpert platform. The Ultra [9], which are described in more detail below. assay is currently the recommended Xpert assay, based WHO defines an individual with presumptive TB as on its increased sensitivity, which improves the detec- anyone who shows symptoms or signs suggestive of TB. tion of paucibacillary TB [5]. Several high TB burden The most common symptom of pulmonary TB is persis- countries such as South Africa and Uganda have tran- tent, productive cough, often accompanied by other non- sitioned towards use of Xpert as the upfront test for TB specific respiratory symptoms (shortness of breath, chest diagnosis. However, despite efforts made by National TB and back pains, hemoptysis) and/or constitutional symp- Programmes, mWRDs are still not used globally as the toms (loss of appetite, weight loss, fever, night sweats, upfront test for TB diagnosis for many people. This is and fatigue) however screening tests such as chest x-ray because of the high cost ($US 9.98 per test at FIND nego- can also be used to identify people with presumptive TB tiated price) and mWRDs being predominantly available despite lack of symptoms [10]. only at higher levels of the TB laboratory network with Pools were created by mixing four consecutive sam- better infrastructure and more qualified human resources ples. Pooled samples were then tested with Xpert MTB/ [6]. Consequently, due to the high costs of the test, car- RIF or Ultra assays. Pools and their corresponding indi- tridges are often rationed, and many tests are only used vidual results were compared to determine the level of as reflex tests once people have been diagnosed, and agreement. more centralized testing can lead to longer turnaround Studies were cross-sectional surveys, conducted during time. PCF programmatic activities. In this approach, which is To maintain sufficient TB testing capacity and cope a patient-initiated pathway to TB diagnosis, individuals with these challenges, one practice that has re-emerged with symptoms suggestive of TB present spontaneously during the Covid-19 pandemic is pooled testing. In this to the health facility for the health worker to initiate the approach, several specimens collected from different investigation for TB using a diagnostic algorithm with presumptive TB cases are pooled (mixed) together and sufficient sensitivity and specificity to diagnose TB [11]. tested as a group in a single assay. If the pooled test is In Lao PDR [8] and Nigeria (Bimba et al. in press), two negative, it is then assumed all samples included in the independent studies were conducted during PCF of peo- pool are negative. If the pooled test is positive, it means ple with presumptive TB, one using Xpert MTB/RIF, and at least one sample included in the pool is positive, and the other Ultra. The study from Brazil assessed the per- individual re-testing of samples is needed to identify the formance of pooled testing with Xpert Ultra only [9]. positive sample(s) (Fig. 1). A systematic review published in 2021 concluded this method was highly sensitive and Statistical analysis specific and can substantially increase testing capacity Categorical data were summarized using descriptive sta- with savings up to 27–31% in cartridges alone, depend- tistics and chi-squared tests were used to test for statis- ing on the prevalence of TB in the population tested [7]. tically significant differences, where appropriate. Pooled However, data on cost-effectiveness are currently limited test results (MTB-positive or MTB-negative) were com- to assay savings on the basis of the number of cartridges pared with the four corresponding Xpert MTB/RIF or that would have been required to test all specimens when Xpert Ultra individual test results and their agreement using individual vs. pooled testing as part of individual was assessed by calculating the Kappa coefficient. The evaluations. kappa values and their interpretations were as follows: In this study, we conducted a cost-effectiveness analy- <0, no agreement; 0–0.19, very weak agreement; 0.20– sis of the pooled testing strategy of Xpert MTB/RIF in 0.39, weak agreement; 0.40–0.59, moderate agreement; comparison with Xpert Ultra, during passive case find- 0.60–0.79, substantial agreement; and 0.8–1.0, excellent ing (PCF) routine activities. Between each method, we agreement [12]. compared the costs to test 1,000 patients, the potential Iem et al. BMC Infectious Diseases (2023) 23:341 Page 3 of 11 Fig. 1 Flow diagram of the sputum processing* *Same test was used for individual and pooled testing (either Xpert MTB/RIF or Xpert Ultra). Cost-effectiveness analysis Cost analysis is a technique that involves the system- We measured the cost-effectiveness of pooled testing vs. atic collection, categorization, and analysis of costs of individual testing by comparing the number of individu- any intervention [13]. Potential savings were calculated als that would be bacteriologically confirmed using each by comparing all resources required to test all speci- method. mens using pooled and individual testing by analyz- ing the costs of each TB detection method. We used Iem et al. BMC Infectious Diseases (2023) 23:341 Page 4 of 11 Table 1 Resources costs assumptions for TB diagnosis by Xpert an ingredient-based, top-down approach, in which all MTB/RIF and Ultra (unit costs from the Global Drug Facility categories of inputs were listed alongside all quantities products catalogs “Ordering List of TB Medicines or Diagnostics, needed to perform all tests annually, for both the individ- Medical Devices and other health products”) ual and pooled testing approach (Table 1). Price Cost Cost per test (International (US$) per test $) The GeneXpert instruments set, biosafety cabinet and (US$) Lao Nigeria Brazil autoclaves, and other small equipment (uninterruptible PDR power supply, timer, vortex) were considered as “capital Supplies required items”. The cost of equipment was determined by using for Xpert test: Pro- the estimated lifetime of capital items in years to which vided by Cepheid we then applied an annuity factor to estimate the cost GeneXpert 15,105 0.5815 1.977 1.522 1.242 per year. The useful time of the capital items reported instrument here was based on annual warranty cost with a 5-year Laptop HP or DELL 2395 0.1126 0.383 0.295 0.240 expected lifetime [14]. brand Biosafety We also listed and quantified all recurrent items equipment needed to perform all the tests over one year, with the Laboratory coats 20 0.0018 0.006 0.005 0.004 cost of all items needed annually. The base level cost of Autoclave 22,000 0.5042 1.714 1.320 1.077 MTB/RIF and Ultra testing were the same. All the Xpert Equipment cartridges, laboratory supplies, disposable personal UPS 1500 VA 1,500 0.0577 0.196 0.151 0.123 protective equipment, biosafety supplies, and human Timer 30 0.0012 0.004 0.003 0.002 resources were considered as “recurrent items”. We then Vortex 220 0.0085 0.029 0.022 0.018 divided the total annual cost for capital and recurrent Stationery items by the number of tests performed annually to esti- Indelible labelling 2 0.0033 0.011 0.009 0.007 mate the unit cost to perform one test. Values for each marker country were adjusted for international dollars by using Pens (red and blue 1 0.0017 0.006 0.004 0.004 DEC (World Bank’s Development Economics depart- or black) ment) alternative conversion factor (local currency units Supplies required per US$) and purchasing power parity conversion factor, for Xpert test gross domestic product (local currency units per interna- MTB/RIF cartridges 499 9.9800 33.930 26.130 21.312 Laboratory tional $) from the World Bank (2021 data). supplies We then compared both approaches to calculate the Sterile screw- 83.5 0.0835 0.284 0.219 0.178 difference in the money invested for testing 1,000 con- capped speci- secutive individuals, the number of people who could be men collection tested for TB when using a fixed amount of 1,000 car- containers tridges, and the costs per bacteriologically confirmed TB Paper towels 2 0.0067 0.023 0.017 0.014 case detected. The cost of pooled testing also included Personal protec- the cost of retesting all specimens from positive pools tive equipment individually. Thus, our cost-effectiveness analysis was Disposable gloves 20 0.0800 0.272 0.209 0.171 able to demonstrate a cost-saving outcome if pooled test- Surgical masks 21.5 0.0344 0.117 0.090 0.073 Biosafety supplies ing cost less than individual testing while detecting at Disposable 20 0.0320 0.109 0.084 0.068 least the same or higher numbers of TB cases. autoclave bags (LxW = 35"x25”) Results Disposable 20 0.0160 0.054 0.042 0.034 Pooled testing diagnostic accuracy (Table 2) autoclave bags In Lao PDR, in the Xpert MTB/RIF survey, 77/81 (sensi- (LxW = 19"x14”) tivity 95.1%, 95%CI 87.8-98.6%) pools containing ≥ 1 posi- Tuberculocidal 45 0.0270 0.092 0.071 0.058 tive sample tested MTB-positive and 4/81 (4.9%, 95%CI disinfectant solution 0.003 L per test 1.4-12.2%) tested MTB-negative. All 129/129 pools con- Human resources taining MTB-negative samples tested MTB-negative Laboratory techni- 250 0.5000 1.700 1.309 1.068 (specificity 100%, 95%CI 97.2-100%), with 98.1% agree- cian (40 h/week, 4 ment (Kappa: 0.959). In the Xpert-Ultra survey, 70/70 weeks/month)* (sensitivity 100%, 95%CI 94.9-100%) pools containing ≥ 1 Total number of 6000* 11.7141 40.91 31.50 25.69 MTB-positive sample tested MTB-positive and 140/140 test/years (specificity 100%, 95%CI 97.4-100%) pools containing * Lao PDR data Iem et al. BMC Infectious Diseases (2023) 23:341 Page 5 of 11 only MTB-negative samples tested MTB-negative, with (93.7-99.2%) (93.9-98.6%) 100% agreement (Kappa: 1). 422/434 Overall (97.2%) 97.6% In Nigeria, 46/50 (92%, 95%CI 80.8-97.8%) positive 97% pools tested Xpert MTB/RIF MTB-positive and 71/75 - - - (94.7%, 95%CI 86.9-98.5%) negative pools tested MTB- Pos negative (agreement 93.6%, Kappa = 0.867). In compari- son, 36/42 (86%, 95%CI 71.5-94.6%) positive pools tested 61 1 Pooled n = 99 Xpert-Ultra MTB-positive and 82/83 (98.8%, 95%CI (86.9-99.0%) (85.1-99.9%) (96.0%) 93.5-99.8%) negative pools tested negative (agreement 94.4%, Kappa = 0.871). There was no statistically signifi- Brazil 95.3% 97.1% 95/99 0.913 Neg cant difference in sensitivity (p-value = 0.33) or specific- 34 3 ity (p-value = 0.14) for pooling with Xpert MTB/RIF or Xpert Ultra. In Brazil, 99 pools were tested, of which 62 (62.6%) Pos 35 Pooled n = 125 7 had MTB-detected and 37 (37.4%) MTB-not detected, (71.5–94.6%) (93.5-99.8%) including six (6.1%) with MTB-trace. The agreement of 117/125 Nigeria (93.6%) individual and pooled testing was 96.0% (Kappa of 0.913). 85.7% 98.8% 0.851 Neg Pooling had sensitivity of 95.3% (95%CI 86.9–99%) and 82 1 specificity of 97.1% (95%CI 85.1–99.9%). There was no significant difference in the overall agreement across all studies with individual testing Pos 70 when pooling either Xpert MTB/RIF (96.4% agreement Pooled n = 210 0 (n = 323/335, CI 95% 93.7-98.1%) or Ultra (97.2% agree- (97.4–100%) (94.9-100%) Lao PDR ment (n = 422/434, CI 95% 95.1-98.5%), p-value = 0.529. 210/210 ULTRA (100%) 100% 100% 1.000 There was also no significant difference in the overall Neg 140 0 performance across all studies when pooling with either Xpert MTB/RIF or Ultra (sensitivity 93.9% (n = 123/131, CI 95% 87.9-97.1%) vs. 97.6% (n = 166/170, CI 95% 93.7- (87.9–97.1%) (94.7–99.4%) 99.2%), p-value = 0.105, and specificity 98% (n = 200/204, 323/335 Overall CI 95% 94.7-99.4%) vs. 97% (n = 256/264, CI 95% 93.9- (96.4%) 93.9% 98% 98.6%, p-value = 0.467, respectively). - - - Testing capacity and number of bacteriologically confirmed TB cases (Table 3) Pos 46 In Lao PDR, pooled testing using a fixed number of 1,000 Pooled n = 125 4 (80.8–97.8%) (86.9–98.5%) Xpert MTB/RIF cartridges would miss 5.1% (n = 10/197) of the TB cases. However, pooled testing would gener- 117/125 Nigeria (93.6%) 92.0% 94.7% 0.867 ate an increase of 62% in the number of people screened Neg 71 4 (1,000 vs. 1,620) leading to an increase of 54% in the Table 2 Agreement of individual and pooled tests absolute number of the TB cases identified despite the 10 missing TB cases (121 vs. 187 (197 − 10)). Pooled test- Pos ing using a fixed number of 1,000 Ultra cartridges would 77 Pooled n = 210 0 generate an increase of 71.5% in the number of people (87.8–98.6%) (97.2–100%) tested (1,000 vs. 1,715) and 71.5% in the absolute number MTB/RIF Lao PDR 206/210 (98.1%) 95.1% of TB cases identified (111 vs. 191), with no missing TB 100% 0.959 Neg 129 cases. 4 In Nigeria, pooled testing using a fixed number of 1,000 Xpert MTB/RIF cartridges would miss 5.8% (n = 13/223) of the TB cases. However, pooled testing would gener- At least one positive ate an increase of 44% in the number of people screened All four negative (1,000 vs. 1,440) leading to an increase of 45.3% in the Agreement Sensitivity Specificity absolute number of TB cases identified despite the 13 Individual (95% CI) (95% CI) missing TB cases (144 vs. 210 (223 − 13)). Pooled testing Kappa using a fixed number of 1,000 Ultra cartridges would miss Iem et al. BMC Infectious Diseases (2023) 23:341 Page 6 of 11 Table 3 Cost analysis of each strategy (individual vs. pooled) by country, by assay (MTB/RIF vs. Ultra) using a fixed amount of 1,000 cartridges Assay Country Testing Number of individuals tested Potential missed Bacteriologically Strategy among TB cases confirmed cases Xpert MTB/RIF Lao PDR Individual N = 1,000 Reference 121 Pooling N = 1,620 (62% increase) 5.1%, n = 10/197 187 (54% increase) Nigeria Individual N = 1,000 Reference 144 Pooling N = 1,440 (44% increase) 5.8%, n = 13/223 210 (45.3% increase) Xpert ULTRA Lao PDR Individual N = 1000 Reference 111 Pooling N = 1,715 (71.5% increase) - 191 (71.5% increase) Nigeria Individual N = 1,000 Reference 120 Pooling N = 1,858 (85.8% increase) 9.6%, n = 27/280 253 (110.7% increase) Brazil Individual N = 1,000 Reference 240 Pooling N = 1,142 (14.2% increase) 3.3%, n = 9/275 265 (10.4% increase) Table 4 Cost analysis of each strategy (individual vs. pooled) by country, by assay (MTB/RIF vs. Ultra) to test 1,000 presumptive TB patients Assay Country Testing Proportion Savings Cost per test Nb of Bacte- Cost per bacte- Strategy positive (%) (International cartridges riologically riologically con- $) confirmed firmed TB case cases (International $) Xpert MTB/RIF Lao PDR Individual 12.1% Reference 40.91 1,000 121 338.07 Pooling 36.7% 37.9% 25.42 617 115 221.04 Nigeria Individual 14.4% Reference 31.50 1,000 144 218.77 Pooling 40% 34.6% 20.61 650 136 151.53 Xpert ULTRA Lao PDR Individual 11.1% Reference 40.91 1,000 111 368.53 Pooling 33.3% 41.2% 24.04 583 111 216.56 Nigeria Individual 12.0% Reference 31.50 1,000 146 215.77 Pooling 28.8% 45.7% 17.10 538 136 125.77 Brazil Individual 24% Reference 25.69 1,000 240 107.06 Pooling 62.6% 12.1% 22.57 876 232 97.30 Overall Individual 34.10 249.64 Pooling 21.95 162.44 9.6% (n = 27/280) of the TB cases. However, pooled test- that would vary among different settings shows costs of ing would generate an increase of 85.8% in the number of the cartridge assay was the major determinant in the unit people screened (1,000 vs. 1,858) leading to an increase cost per test variation, accounting for 85.2% of the cost to of 110.7% in the absolute number of TB cases identified test one person with presumptive TB. despite the 27 missing TB cases (120 vs. 253 (280 − 27)). The overall unit cost across all studies (1,000 individ- In Brazil, pooled testing using a fixed number of ual sample size population) to test one person was 34.10 1,000 Ultra cartridges would miss 3.3% (n = 9/275) of international dollars for the individual testing and 21.95 the TB cases. However, pooled testing would generate international dollars for the pooled testing, resulting in a an increase of 14.2% in the number of people screened savings of 12.15 international dollars per test performed (1,000 vs. 1,142) and 10.4% in the number of TB cases (35.6% decrease). The overall unit cost per bacteriologi- identified despite the 9 missing TB cases (240 vs. 265). cally confirmed TB case was 249.64 international dollars for the individual testing and 162.44 international dollars Costs of detection methods (Table 4) for the pooled testing (34.9% decrease). Cost-minimization analysis Figure 3a shows there is a linear correlation between Since the detection of TB cases by individual and pooled the prevalence of the disease in the population tested and testing, with both Xpert MTB/RIF and Ultra was not sig- the proportion of positive pools. This has a direct impact nificantly different, we compare only the costs of tests and on the savings: the lower the proportion of positive pools, accept the least costly one as the cost-effective method by the higher the savings in assay costs (Fig. 3b), since fewer utilizing the cost-minimization analysis technique [13]. pools require individual testing. Consequently, the lower The univariate sensitivity analysis (Fig. 2a and b) on other the proportion of positive pools, the higher the increase parameters that could affect the cost-effectiveness and of testing capacity (Fig. 3c). Iem et al. BMC Infectious Diseases (2023) 23:341 Page 7 of 11 Fig. 2 Parameters affecting the pooled testing cost-effectiveness Based on these findings, by applying a forecast forward Varying the number of samples per pool may also help from the trendline of the graph in Fig. 3b, we can observe improve cartridge savings [15]. Pooled testing demon- savings disappear when the proportion of positive pools strated high sensitivity and specificity with both Xpert is ≥ 75%. Inductively, when applying a forecast forward MTB/RIF and Xpert Ultra. At a time when international on graph 2a, a 75% proportion of positive pools corre- donors are increasingly requesting countries to commit sponds to a 30% prevalence of TB. Therefore, when the to co-financing mechanisms for the procurement of tests prevalence of TB is ≥ 30%, pooled testing is unlikely to from government-funded schemes, the pooling method still be cost-effective. is relevant to help National TB Programs cope with these funding gaps. Discussion Discrepancies between individual and pooled tests only Our findings suggest that pooled sputum testing may occurred among pauci-bacillary samples with high Xpert improve the efficiency of GeneXpert-based testing in a CT values. This suggests that some samples with low variety of settings. Depending on the local TB prevalence, DNA concentrations fall below the assay’s limit of detec- pooled testing could potentially enable the screening and tion once mixed in the pool. Consequently, some patients testing of larger numbers of people more cost-effectively. with paucibacillary disease could be missed by pooling, Iem et al. BMC Infectious Diseases (2023) 23:341 Page 8 of 11 Fig. 3 Effect of the prevalence of the disease on the amount of savings by pooling method especially if testing is based on Xpert MTB/RIF. How- higher number of individuals could be tested leading to ever, if we look at the resources needed to screen this a higher absolute number of bacteriologically confirmed fixed number of patients, the savings will allow a higher cases within a fixed time period with a fixed amount of number of patients to be tested using the same amount resources, despite the number of TB cases. Pooled test- of resources. Therefore, under the pooling approach, a ing will allow a faster catch-up and more cost-effective Iem et al. BMC Infectious Diseases (2023) 23:341 Page 9 of 11 strategy to find the people with TB compared to indi- and the savings were a function of the estimated underly- vidual testing. Moreover, Cepheid will discontinue the ing prevalence of the disease (proportion of people with production of the Xpert/MTB RIF assay in 2023, and the MTB-positive results) in the setting where the pooled Global Laboratory Initiative from the Stop TB Partner- testing was implemented and their distribution within ship, has issued practical guidance to plan and imple- the pools. When the proportion of individuals with posi- ment a smooth transition from use of Xpert MTB/RIF to tive tests is lower, there are more MTB-negative pooled Xpert MTB/RIF Ultra cartridges, ensuring uninterrupted tests which do not require further testing, leading to service and avoiding cartridge wastage [16]. If countries higher savings. In the study from Brazil [9], the propor- choose to implement pooled testing going forward, only tion of individual samples MTB-detected was much Xpert Ultra will be available, which has better sensitivity higher (24%), and many more MTB-detected pools and agreement compared to Xpert MTB/RIF. required further individual testing (62.6%), resulting in A small number of individual samples included in reduced cost savings (12%). Pooling therefore works well MTB-positive pooled test results would return an when there is a low TB prevalence, with more negative MTB-negative result when re-tested individually. These than positive results [19]. This is an important practi- are unexpected results since the Xpert MTB/RIF and cal factor to consider before implementation of pooled the Ultra are highly specific and are not expected to testing, as the proportion of positive pools varies signifi- yield false-positive results [5]. However, in other stud- cantly according to the population to be tested. Extrapo- ies assessing the performance of the pooled testing for lations from results reported here confirm the findings SARS-Cov-2 [17], these false-positive pooled test results from previous study showing that in population where happened on rare occasions with pools displaying bor- the disease prevalence is above 30%, the proportion of derline high CT values suggesting very small quantity pools returning an MTB-positive results would be high of genetic material, and the authors have attributed it to (75%), leading to no savings due to the high number of cross-contamination during samples handling and pro- deconvolution [20]. Adjusting the number of samples per cessing. In general, for all diagnostic tests, false-positive pool may increase the efficiency of pooling based on the results occur more frequently in low prevalence settings expected prevalence [15]. Pooling is not a universal solu- [18], and this is why for instance the WHO recommends tion and National TB Programmes need to be cautious as repeat Xpert test with a fresh sample whenever rifampi- to where and when to apply it. Laboratories should deter- cin resistance is detected for an individual from groups mine the TB prevalence based on a rolling average of the with low risk of RR/MDR-TB, despite the high specific- positivity rate of their own testing and for different popu- ity of the assay [2]. It is therefore important to properly lations/groups. Indeed, clinical history of the patients organize the workflow of samples with adequate labora- to be tested by the pooling method must be considered, tory commodities, clear standard operating procedures especially in settings where HIV is prevalent. PLHIV to avoid any clerical errors or risks of contamination. The have low sputum bacillary loads, and mixing those sam- interpretation of trace results should also be interpreted ples into a pool with MTB-negative samples will increase cautiously. If a pool returns a trace result, all samples the risk of getting a false negative pooled test result due included in that pool should be retested individually in to the dilution factor. However, if the proportion of TB- order to determine if the pooled trace result is due to HIV coinfected is high, the risk of false negative results a very low load of bacilli that became trace due to the may be minimized by the increased likelihood of samples dilution effect, or if a trace sample was indeed included containing more than one positive specimen in the pool. in the pool. If individual testing of samples from the Other studies have shown the dilution effect was not trace pooled test shows there was a sample with a very homogeneous, as pools with multiple positive samples low result, or one or more trace results samples, those often had the same or lower CT value than individual patients need to be managed according to their national samples [17], thus indicating that the combination of diagnostic algorithm considering pulmonary or extra- multiple positive samples in a pool increases the total pulmonary TB, HIV status, age, and prior TB treatment. amount of genetic material and compensates for the dilu- However, since all samples included in MTB-positive tion effect. Laboratories can then determine when the pools are re-tested individually, the false-positive pooled positivity rate is low enough to justify the implementa- test results would have no impact since the individual tion of a pooling strategy [19]. Moreover, the use of the test result is used to guide the clinical management of the pooling method should be a dynamic strategy following patients. the evolution of the TB prevalence in the selected area Our results demonstrate that pooling samples can sig- and the positivity rate of laboratory results. nificantly increase testing capacity, while simultaneously reducing the resources needed for TB mass testing. The unit cost for testing each person with presumptive TB Iem et al. BMC Infectious Diseases (2023) 23:341 Page 10 of 11 Study limitations Conclusions Results reported here are focused on the costs minimiza- Our results demonstrate the repeatability, reliability, tion of the pooled testing for TB diagnosis, but param- consistency, and accuracy of the pooling method in a eters included for the analysis did not encompass all variety of settings with both Xpert MTB/RIF and Xpert actual costs. For example, the costs for maintenance of Ultra, in PCF approach. The low frequency of false-neg- the instruments were not included. These costs comprise ative results and the high degree of specificity makes the price of the spare parts such as the module (900$ this approach a cost-effective strategy for large scale TB per refurbished module, 3000$ for new module) or the testing at reduced costs. This can allow resource limited annual calibration (Xpert Check calibration kit at 450$ countries to catch up with the WHO End TB strategy tar- per kit per machine), shipment, purchasing and supply gets despite the reversal of progress due to the Covid-19 management (PSM) costs and the manpower to carry out pandemic. the calibration or replace faulty elements. Maintenance Acknowledgements and servicing were recognized as major bottlenecks for The authors would like to thank the National Tuberculosis Control Center the scale up of the GeneXpert instrument to a lower level (NTC) of Lao PDR for authorizing the study. We are also grateful to the local in the laboratory network [6]. The absence of local autho- health authorities in Nigeria and Brazil that allowed access to the study sites and for their coordination locally. We also thank Jim Read and the Global rized service providers from Cepheid and limited capac- Health Trial Unit, LSTM for generating databases and data curation during ity of end-users for maintenance have led to high rates of primary data collection of pooling studies. module failures in different settings [21]. Including main- This paper is dedicated to the memory of our dear friend, mentor, and co-worker Professor Luis Eduardo Cuevas, who passed away while this paper tenance costs in the analysis would therefore significantly was being peer-reviewed. increase the actual unit cost of the test but would likely make pooling more cost-savings. Author Contribution The study was designed by VI, JAMK and LEC. Data collection and analysis Secondly, we have not included the costs for repeat- were conducted by VI, JSB and JAMK. VI, JAMK, TW and LEC wrote the first ing all Xpert tests with non-valid results (invalid, error, draft of the manuscript. All authors made substantial contributions to the no results) in the calculation. This is an important fac- writing and editing of the manuscript. The final version has been read and approved by all authors. tor because the rate of non-valid Xpert results can sig- nificantly vary from one setting to another, impacting the Funding costs and cost-effectiveness of the pooled testing. Some This research was funded in part by a TB REACH, Stop TB Partnership grant supported by Global Affairs Canada (STBP/TBREACH/GSA/2020-04, awarded to studies have reported abnormally high rates of non-valid LEC), the Global Fund to Fight AIDS, Tuberculosis and Malaria (LAO-T-GFMOH results, with 10.6% (range 5.9–16.3%) in nine countries grant) and the National Tuberculosis Control Program budget from the implementing Xpert MTB/RIF [22], 7.2% (range 4–17%) Ministry of Health of Lao PDR. TW is supported by grants from the Wellcome Trust, UK (209075/Z/17/Z); the Medical Research Council, Department for in India [6], and 11% for Nigeria [23]. These high rates of International Development, and Wellcome Trust, UK (Joint Global Health non-valid results were attributed to either the environ- Trials, MR/V004832/1), the Medical Research Council, UK (MR/V028618/1); the ment with high temperature and/or dust, or due to poor Academy of Medical Sciences, UK; and the Swedish Health Research Council, Sweden. adherence to standard operating procedures. Thirdly, this study focuses on the cost minimisation Data Availability for the diagnosis of TB using pooled testing compared The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request. to individual testing. We therefore did not assess the impacts of earlier TB diagnosis and TB treatment ini- Declarations tiation, nor did we incorporated into the analysis the cost-effectiveness of preventing additional disease trans- Ethics approval and consent to participate mission. Cost-effectiveness analyses are more robust All methods were carried out in accordance with the Declaration of Helsinki as well as in accordance with relevant national guidelines and regulations. when the number of people correctly diagnosed and Need for ethical approval and informed consent were waived by the National started on treatment is included along with costs and TB Control Center of Lao PDR and the Liverpool School of Tropical Medicine outcomes related to treatment, survival and disability, Research Ethics Committee, UK (Ethical waiver 20–037) for studies conducted in Lao PDR. using cost per disability-adjusted life year (ref ). Many Studies in Nigeria were approved by the Health Research Ethics Committees model-based economic evaluations (ref ) predicted that of the Liverpool School of Tropical Medicine, UK, and the Federal Capital Xpert would be cost-effective through a reduction in Territory, Nigeria (numbers 20–037 and FHREC/2020/01/29/10-04-20, respectively). All patients attending the centers were asked to read and tuberculosis-related mortality and/or reduction in the confirm they had understood the study information leaflets and underwent overtreatment of tuberculosis (ref ). Given that more consent procedures. Individuals were included after they had provided written cases are detected with pooling, more patients will be ini- informed consent to participate. The study in Brazil was approved by the Committee on Ethics in Research tiated on treatment leading to less transmission, so likely with Human Beings at the Universidade Federal de Alagoas, Brazil (CAAE that the pooling strategy would be more cost-effective if number 45432821.2.0000.5013) and the Liverpool School of Tropical Medicine these parameters are incorporated into the model. Research Ethics Committee, UK (Ethical waiver 20–037), who both waived the need for informed consent. Iem et al. BMC Infectious Diseases (2023) 23:341 Page 11 of 11 Consent for publication 7. Cuevas LE, Santos V, Lima SVMA, et al. Systematic review of pooling Sputum Not applicable. as an efficient method for Xpert MTB/RIF tuberculosis testing during the COVID-19 pandemic. Emerg Infect Disease J. 2021;27(3):719. Competing interests 8. Iem V, Chittamany P, Suthepmany S, et al. Pooling sputum for Xpert MTB/ The authors declare no competing interests. RIF and Xpert Ultra testing during the Covid-19 pandemic in Lao People’s Democratic Republic. PLOS Global Public Health. 2022;2(4):e0000116. Author details 9. Santos VS, Allgayer MF, Kontogianni K et al. Pooling of sputum samples to 1 National Tuberculosis Control Center, Lao People’s Democratic Republic, increase tuberculosis diagnostic capacity in Brazil during the COVID-19 Vientiane, Laos pandemic. Int J Infect Dis 2023. 2 Liverpool School of Tropical Medicine, Pembroke Place, 10. World Health Organization. Implementing the WHO Stop TB Strategy: A Liverpool L3 5QA, UK Handbook for National Tuberculosis Control Programmes. 2008: 1, Case 3 Zankli Research Centre, Bingham University, Karu, Nigeria detection. 4 Departamento de Medicina, Universidade Federal de Sergipe, Lagarto, 11. World Health Organization. WHO consolidated guidelines on tuberculosis. Brazil Module 2: screening – systematic screening for tuberculosis disease. 2021. 5 Programa de Pós-Graduação em Ciências da Saúde, Universidade 12. Landis JR, Koch GG. The measurement of observer agreement for categorical Federal de Sergipe, Aracaju, Brazil data. Biometrics. 1977;33(1):159–74. 6 Institut d’Investigació Germans Trias i Pujol, CIBER Enfermedades 13. Drummond MF, Sculpher MJ, Claxton K, et al. Methods for the economic Respiratorias, Universitat Autònoma de Barcelona, Barcelona, Spain evaluation of Health Care Programmes. Oxford University Press; 2015. 7 Stop TB Partnership, Innovations and Grants, Geneva, Switzerland 14. Foundation for Innovative New Diagnostics. GeneXpert negotiated prices. 8 Lao Association for Medical Laboratory Sciences, Lao People’s 2021. Democratic Republic, Vientiane, Laos 15. Abdalhamid B, Bilder CR, McCutchen EL, et al. Assessment of Specimen 9 WHO Collaborating Centre for Tuberculosis and Social Medicine, Pooling to conserve SARS CoV-2 Testing Resources. Am J Clin Pathol. Department of Global Public Health, Karolinska Institutet, Solna 2020;153(6):715–8. 17177, Sweden 16. Global Laboratory Initiative. Planning for country transition to Xpert MTB/RIF 10 Health Economics and Policy Unit, School of Public Health and Ultra cartridges. 2017. Community Medicine and Centre for Health Governance, University of 17. Iem V, Xangsayarath P, Chittamany P, et al. Pooling samples to increase testing Gothenburg, Gothenburg, Sweden capacity with Xpert Xpress SARS-CoV-2 during the Covid-19 pandemic in Lao People’s Democratic Republic. PLoS ONE. 2022;17(9):e0275294. Received: 2 February 2023 / Accepted: 15 May 2023 18. Centers for Disease Control and Prevention. Guidance for Antigen Testing for SARS-CoV-2 for Healthcare Providers Testing Individuals in the Community. 2022. 19. Centers for Disease Control and Prevention. Interim Guidance for Use of Pool- ing Procedures in SARS-CoV-2 Diagnostic and Screening Testing. 2021. 20. Williams BG. Optimal pooling strategies for laboratory testing. 2010; p. arXiv:1007.4903. References 21. Albert H, Nathavitharana RR, Isaacs C, et al. Development, roll-out and impact 1. World Health Organization. Global tuberculosis report 2021. 2021. of Xpert MTB/RIF for tuberculosis: what lessons have we learnt and how can 2. World Health Organization. WHO consolidated guidelines on tuberculosis. we do better? Eur Respir J. 2016;48(2):516. Module 3: diagnosis - rapid diagnostics for tuberculosis detection, 2021 22. Creswell J, Codlin AJ, Andre E, et al. Results from early programmatic update. 2021. implementation of Xpert MTB/RIF testing in nine countries. BMC Infect Dis. 3. Lawn SD, Nicol MP. Xpert MTB/RIF assay: development, evaluation and 2014;14(1):2. implementation of a new rapid molecular diagnostic for tuberculosis and 23. Gidado M, Nwokoye N, Nwadike P, et al. Unsuccessful xpert MTB/RIF results: rifampicin resistance. 2011(1746 – 0921 (Electronic)). the nigerian experience. Public Health Action. 2018;8(1):2–6. 4. Chakravorty S, Simmons AM, Rowneki M et al. The New Xpert MTB/RIF Ultra: improving detection of Mycobacterium tuberculosis and Resistance to Rifampin in an assay suitable for point-of-care testing. mBio 2017: 8(4). Publisher’s Note 5. World Health Organization. WHO meeting report of a technical expert con- Springer Nature remains neutral with regard to jurisdictional claims in sultation: non-inferiority analysis of Xpert MTF/RIF Ultra compared to Xpert published maps and institutional affiliations. MTB/RIF. 2017. 6. Raizada N, Sachdeva KS, Sreenivas A et al. Feasibility of decentralised deploy- ment of Xpert MTB/RIF test at lower level of health system in India. PLoS ONE 2014: 9(2): e89301.