This was a cross sectional study performed between late May to early July 2018, which utilized the remaining sputum samples submitted and tested routinely for TB diagnosis at Bugando Medical Centre TB laboratory. Bugando Medical Centre is a 1000-bed tertiary hospital located in Mwanza, Tanzania serving about 15.0% of the Country’s population of about 60 million.

Sputum samples from Presumptive TB patients attending TB clinic at Bugando Medical Centre were submitted to TB laboratory for routine diagnosis of TB using Xpert MTB/RIF (Cepheid Inc., Sunnyvale, CA, US). The TB results from the individual sputum sample were used as a reference standard to examine the pools results and were concealed from the investigating laboratory technicians. A total of 250 samples from the remainder of samples from presumptive TB patients submitted for routine Xpert MTB/RIF testing at Bugando Medical Centre Laboratory were used serially for pooling analysis. All methods were carried out in accordance with standard operating procedures adopted from manufacturer’s instructions [14] .

Residualsputum samples with sufficient volume to make up a pool were serially collected until the sample size was reached.

The residual sputum samples were de-identified and used to make pools of five samples per pool with each sample contributing 0.5 mL. Buffer solution was added to each sputum sample before splitting. These pooled samples were transferred into 5 mL falcon tubes, each tube making one pool. The laboratory registration (identification) number for each sample in a pool was noted in linkage with the pool number to ensure no mix up of samples and pools during analysis. The laboratory personnel working with the MTB/RIF testing for pooled samples was blinded for the results from the individual MTB/RIF sputum testing. The MTB/RIF testing for both individual and pooled samples was carried out following the tertiary hospital laboratory standard operating procedures adopted from the manufacturer’s instructions (Cepheid Inc., Sunnyvale, CA, US, 2010) [14] .

Data from questionnaires and laboratory forms were entered into the computer using Microsoft Excel 2016 (Microsoft Corporation, WA, US) and then imported to STATA version 13 (StataCorp, College Station, TX, US) for analysis. Continuous data were summarized using median with interquartile range [IQR] whereas categorical data were summarized using frequency and proportions (percent). To determine the sensitivity and specificity of pooled 5 samples testing strategy we employed 2 × 2 contingency using individual sputum samples Xpert MTB/RIF results as the reference standard.

The ethical approval to conduct this study was obtained from the joint Bugando Medical Centre/Catholic University of Health and Allied Sciences (BMC/CUHAS) ethics review board (CREC 672/2018). A waiver of informed consent was obtained from the ethics committee/institutional review board of Central Pathology Laboratory to use the samples for this study.

Of the 250 samples, 143 (57.2%) were from male patients. The median age for all patients was 35 (IQR 27 - 40) years. Table 1 shows the demographic information.

Of the 50 pooled sputum samples, MTB was detected in 17 (34.0%) pools whereas 33 pooled sputum samples (66.0%) had no MTB detected. Of the 17 MTB positive pools, 10 (58.8%) were detected as medium while the remaining 7 (41.2%) pools were detected as low MTB. Of note, there was no pool that had a high MTB detected. Furthermore, none of the pools was detected to have Rifampicin resistance.

Finally results from the individual sputum samples MTB/RIF testing were retrieved for all 250 samples and there were 28 (11.2%) samples in which MTB was detected whereas 222 (88.8%) samples had no MTB detected. Of the 28 samples in which MTB was detected, 2 (7.1%) had high MTB detected results, 13 (46.4%) medium MTB detected and the rest 13 (46.4%) had low MTB detected. None of the sputum samples was Rifampicin resistant. Following re-analysis of 17 positive pools for MTB using individual sputum samples MTB/RIF results as a reference standard, all 28 (100%) individual sputum samples tested positive for

Xpert MTB/RIF was detected by pooling. The number of individual MTB positive samples ranged from 1 to 3 per pool, with 3 (17.6%) pools contained two positive individual samples whereas 4 (23.5%) pools contained three positive individual samples. Furthermore, of the 33 negative pools for MTB, and none of the negative pools contained the positive individual sample for MTB. These results showed that all the positive and negative individual sputum samples were correctly identified, with an agreement of 100% between individual and pooled sputum on Xpert MTB/RIF testing. Table 2 summarizes the comparison of Cycle Thresholds Results of Pooled versus Individual sputum samples.

The pooling of samples has the effect of increasing the dilution (decreasing MTB concentration) as compared to the individual sample analysis. This is shown by the increase in cycle thresholds whereby pooled samples had slightly higher cycle thresholds than the original individual sample cycle threshold. Table 3 summarizes the comparison of cycle thresholds and dilutions between the individual sputum samples and pooled sputum samples.

Pooling sputum Xpert MTB/RIF strategy detected all the positive MTB cases that were also detected from the individual Xpert MTB/RIF testing. There was a 100% agreement between results from pooling sputum Xpert MTB/RIF and individual Xpert MTB/RIF results. The sensitivity and specificity were 100% respectively.

The strategy of pooling sputum, as was used in this scenario, has saved a total of 115 cartridges by using 135 cartridges instead of 250 that would have been used for individual MTB/RIF testing. This salvage is equivalent to 46.0%. Therefore considering that in year 2019 the Hospital performed 2803 Xpert MTB/RIF tests and using this strategy we could have saved 1290 cartridges.

This study has shown a level of agreement of 100% between individual and pooled sputum on Xpert MTB/RIF testing with a sensitivity and specificity of 100% respectively. These results are in-line with two other studies conducted in Vietnam by Ho et al., and Phuong et al., by pooling 2 sputum samples per pool, where a sensitivity of 88% and 96% were detected respectively [9] [10] . Whereas studies that used 4 samples per pool in Nigeria (Abdurrahman et al.), Vietnam (Ho et al.) Cambodia (Chry et al.) and Brazil (Santoz et al.) found the sensitivity of 88%, 94%, 95% and 100% respectively [8] [9] [11] [12] . Our specificity of 100% is similar to the previous studies done by pooling 4 sputum samples per pool in Nigeria by Abdurrahman et al., and in Cambodia by Chry et al., and found the specificity of 99% and 100% respectively [8] [11] .

Xpert MTB/RIF Pooled Sputum Strategy is expected to detect higher levels of TB cases in active case finding than in passive case finding, as in the active case finding there may be more risk missing. In our study we used passive case finding and got comparable results with those from studies that also utilized passive case finding from Nigeria and Vietnam [8] [10] . Surprisingly, an active case finding study in Cambodia that employed the active case finding had comparable and very good results with that the studies utilized passive case finding. This could be attributed to the fact that in Cambodia study by Chry et al., they used Xpert MTB/RIF Ultra [11] . These findings underpin the utility of Xpert MTB/RIF Pooled Sputum Strategy using ultra for screening of pulmonary tuberculosis in active case finding.

The concentration of MTB for positive samples was reduced in pooled sputum samples as compared to the individual samples; this was shown by the increase in cycle thresholds whereby pooled samples had slightly higher cycle thresholds than the original individual sample cycle threshold. Among all the positive individual sputum samples there was no one with a very low MTB detected as samples were pooled without the knowledge of the individual sample results. This scenario partly explains why we have a sensitivity of 100% apart from the dilution effect observed after pooling samples together. The dilution effect has also expressed itself in this study whereby there were two samples with high MTB detected but high TB was not detected in any of the pools. This effect was also observed in a study done by Abdurrahman et al., [8] whereby samples with very low MTB detected turned negative when included in pools.

The strategy has proved to be cost-effective by saving 115 Xpert MTB/RIF test cartridges. This implies that if the technique is employed; 50 cartridges would be used to test the 50 pooled sputum samples and re-testing individual sputum samples making the 17 pools with detected MTB results (85 cartridges) making a total of 135 test cartridges. This salvage is equivalent to 46.0%. Therefore considering that in year 2019 the Hospital performed 2803 Xpert MTB/RIF tests and using this strategy we could save 1290 cartridges. Taking into consideration of the concession price of US$ 9.98 per cartridge [15] this strategy would have saved a total of US$ 12,874 for year 2019 expenditures on cartridges. Furthermore, this study has shown that performing fewer numbers of tests using this technique directly saves time as it has also been shown in previous studies in Nigeria, South Africa, Vietnam, Cambodia and Brazil [5] [7] - [12] . Studies were done in Nigeria by Abdurrahman et al., and the one in Cambodia by Chry et al., have also demonstrated the cartridge savings of 31% and 27% respectively [8] [11] . Of note, our cartridge savings of 46.0% is higher than those from these two previously reported studies due to the fact that we pooled 5 samples compared to previous studies that pooled 4 samples [8] [11] . Our study is limited by the fact that we did not establish the optimum pooling ratio.