Feasibility and Sensitivity of Saliva GeneXpert MTB/RIF Ultra for Tuberculosis Diagnosis in Adults in Uganda

ABSTRACT The objective of this prospective observational study carried out at China-Uganda Friendship Hospital-Naguru in Kampala, Uganda, was to determine the performance of GeneXpert MTB/RIF Ultra (Xpert Ultra) molecular testing on saliva for active tuberculosis (TB) disease among consecutive adults undergoing TB diagnostic evaluation who were Xpert Ultra positive on sputum. We calculated sensitivity to determine TB diagnostic performance in comparison to a composite reference standard of Mycobacterium tuberculosis liquid and solid cultures on two spot sputum specimens. Xpert Ultra on a single saliva sample had a sensitivity of 90% (95% confidence interval [CI], 81 to 95%) relative to the composite sputum culture-based reference standard, similar to the composite sensitivity of 87% (95% CI, 77 to 94%) for fluorescence microscopy (FM) for acid-fast bacilli on two sputum smears. The sensitivity of salivary Xpert Ultra was 24% lower (95% CI for difference, 2 to 48%; P = 0.003) among persons living with HIV (71%; 95% CI, 44 to 90%) than among persons living without HIV (95%; 95% CI, 86 to 99%) and 46% higher (95% CI, 14 to 77%; P < 0.0001) among FM-positive (96%; 95% CI, 87 to 99%) than among FM-negative (50%; 95% CI, 19 to 81%) patients. The semiquantitative Xpert Ultra grade was systematically higher in sputum than in a paired saliva sample from the same patient. In conclusion, molecular testing of saliva for active TB diagnosis was feasible and almost as sensitive as molecular testing of sputum in a high TB burden setting. IMPORTANCE Tuberculosis is among the leading causes of morbidity and mortality worldwide, in large part because >3 million people go undiagnosed and untreated each year. Sputum has been the mainstay for TB diagnosis for over a century but can be difficult for patients to produce. In addition, the vigorous coughing required during sputum collection can lead to infection of nearby individuals and health workers. In this case-only study, applying the ultra-sensitive GeneXpert MTB/RIF Ultra molecular diagnostic assay to saliva detected 90% of culture-confirmed TB cases among 81 adults who were undergoing TB evaluation at the outpatient department of a general hospital in Uganda and tested sputum GeneXpert MTB/RIF Ultra positive. These results suggest that saliva may be a feasible and sensitive alternative to sputum for TB diagnosis, thereby meeting two key metrics proposed by the World Health Organization in its target performance profile for a nonsputum test for TB.


Introduction
Thank you for the opportunity to review this manuscript. The authors have provided evidence that saliva may a suitable specimen for the diagnosis of TB, while acknowledging that future studies are needed in order to address lingering questions. The authors collected saliva in a population of sputum Xpert positive patients and showed that there were similar rates of detection of tuberculosis in saliva Xpert Ultra as compared to sputum smear microscopy using sputum culture as the reference standard. While a population of sputum Xpert positive patients is convenient, it does raise the concern that such a population is more likely to test positive with saliva Xpert Ultra as well as sputum culture, thus inflating the sensitivity higher than would be observed in a sputum Xpert negative population. Saliva samples were treated at a 1:1 ratio with Cepheid's Sample Reagent. There are potential biosafety considerations using this lower concentration of Sample Reagent that the authors did not address in this study. The authors are encouraged to comment on biosafety concerns as well as temper their conclusions given the selective population in this study.
Major Comments 1. The authors chose to treat the saliva samples with Cepheid Sample Reagent at a 1:1 ratio, which is used for some other non-pulmonary samples such as cerebrospinal fluid. In contrast, sputum is treated at a 2:1 Sample Reagent to sputum ratio. As reported, other groups, such as the South Africa/South Korea study (1) had treated saliva at a 2:1 ratio. The purpose of the sample reagent is two-fold, to liquefy viscous samples, as well as to render bacilli unviable and thus lowering the biohazard risk. As reported in Helb et al. (2) killing assays in spiked sputum were utilizing two volumes of Sample Reagent per volume of sputum. That grouped showed that at that concentration, after 15 minutes, viability was reduced by at least 8-logs in sputum. This is an important consideration in the safety of laboratory staff to ensure that they are not unnecessarily exposed to viable bacilli. To the reviewer's knowledge, no such killing assays have been conducted for saliva samples at a lower Sample Reagent concentration. If samples are handled within a biosafety cabinet at all times, concerns regarding a lower Sample Reagent concentration are alleviated; however, if treated samples are opened on an open bench, lab workers could potentially be exposed to viable bacilli. The authors are encouraged to specify that samples should be handled within a biosafety cabinet at all times or conduct killing assays to show that lower concentrations of Sample Reagent still result in an adequate reduction in bacilli viability. 2. In the Abstract, the authors state, "The objective... was to determine the performance of GeneXpert MTB/RIF Ultra (Xpert) testing on saliva for active tuberculosis (TB) disease among consecutive adults undergoing diagnostic evaluation." This statement is misleading. While the parent study did enroll consecutive adults undergoing TB evaluation, in the sub-study, saliva samples were only collected among adults that had already tested positive for tuberculosis using sputum Xpert. This has important implications in the interpretation of the reported sensitivity. As is noted in the Discussion regarding Wood et al. (3), case-control study designs are prone to inflate diagnostic accuracy. The authors of this study do not acknowledge this same limitation in their study design, which is effectively a "caseonly" study population. The authors do note the limitation of not being able to determine specificity, but do not note the limitation of inflated sensitivity. The authors are encouraged to acknowledge this major limitation throughout the manuscript. 3. Saliva was collected "at least two hours after sputum collection." Is this a sufficient period of waiting? Why was saliva not collected before sputum? It could be reasonably assumed that the process of expectorating sputum would leave some bacilli in the mouth that could then later be released in saliva. Had the order of collection been reversed (saliva first and then sputum), one might expect detection in saliva Xpert to be lower. The authors are strongly encouraged to discuss the rationale for this methodology, and discuss whether future studies should explore sample collection order.

Staff Comments:
Preparing Revision Guidelines To submit your modified manuscript, log onto the eJP submission site at https://spectrum.msubmit.net/cgibin/main.plex. Go to Author Tasks and click the appropriate manuscript title to begin the revision process. The information that you entered when you first submitted the paper will be displayed. Please update the information as necessary. Here are a few examples of required updates that authors must address: • Point-by-point responses to the issues raised by the reviewers in a file named "Response to Reviewers," NOT IN YOUR COVER LETTER. • Upload a compare copy of the manuscript (without figures) as a "Marked-Up Manuscript" file. • Each figure must be uploaded as a separate file, and any multipanel figures must be assembled into one file. For complete guidelines on revision requirements, please see the journal Submission and Review Process requirements at https://journals.asm.org/journal/Spectrum/submission-review-process. Submissions of a paper that does not conform to Microbiology Spectrum guidelines will delay acceptance of your manuscript. " Please return the manuscript within 60 days; if you cannot complete the modification within this time period, please contact me. If you do not wish to modify the manuscript and prefer to submit it to another journal, please notify me of your decision immediately so that the manuscript may be formally withdrawn from consideration by Microbiology Spectrum.
If your manuscript is accepted for publication, you will be contacted separately about payment when the proofs are issued; please follow the instructions in that e-mail. Arrangements for payment must be made before your article is published. For a complete list of Publication Fees, including supplemental material costs, please visit our website. 90%) than among persons living without HIV (95%, 95%CI 86-99%) and 46% lower (95% CI 11 for difference 14-77%, p<0.0001) among sputum microscopy positive (96%, 95% CI 87-99%) 12 than among sputum microscopy negative patients (50%, 95% CI 19-81%). Semi-quantitative 13 Xpert grade was higher in sputum than in paired saliva samples from the same patient. In 14 conclusion, saliva specimens appear to be feasible and similarly sensitive to sputum for active 15 TB diagnosis using molecular testing, suggesting promise as a non-sputum diagnostic test for 16 active TB in high-burden settings. ver the last quarter century, improvements in diagnosis and treatment of people with 2 tuberculosis (TB) have gradually reduced mortality, but large gaps in detection and 3 treatment persist that contribute to substantial ongoing morbidity and mortality [1]. Among 4 several available strategies to facilitate rapid, same-day diagnosis of TB, testing sputum with the 5 GeneXpert MTB/RIF Ultra molecular assay [2,3] is the most sensitive and most readily 6 available approach. Unfortunately, there are several operational challenges associated with 7 collecting sputum for diagnosis of pulmonary TB. First, coughing during sputum expectoration 8 or sputum induction generates aerosols that may facilitate TB transmission [4]. Second, some 9 individuals are unable to produce sputum, including young children, those with dry cough, and 10 the severely ill/severely debilitated. Given these limitations of sputum for TB diagnosis, in 2014 11 the World Health Organization (WHO) issued guidelines for developers of a future non-sputum 12 test for active TB diagnosis, including a target product profile suggesting that it should have a 13 minimum diagnostic accuracy similar to sputum GeneXpert MTB/RIF on sputum smear-negative 14 individuals (i.e., sensitivity ≥68%, specificity ≥98%) [5]. 15 16 One alternative sample type with great promise for diagnosis of pulmonary TB is saliva, which is 17 easy to collect, with minimal risk of generating aerosols. Although Stop TB Partnership 18 guidelines discourage collection of salivary sputum samples because they have lower diagnostic 19 yield for acid-fast bacilli (AFB) by microscopy or culture, the diagnostic yield of TB molecular 20 testing appears to be more promising. In a previous study of 1782 smear-negative adults 21 undergoing evaluation for active TB, for example, we found that salivary sputum provided a 22 substantially higher diagnostic yield and sensitivity for culture-positive TB than other sputum 23 O 4 types, implying incremental value to using oral samples at least as a supplement to sputum [6]. 24 Using a different sampling technique, oral swabs, Wood and colleagues showed that oral nylon 25 swabs repeatedly tested positive for TB via IS6110 polymerase chain reaction testing in 90% of 26 South African patients with TB confirmed by sputum GeneXpert MTB/RIF testing, suggesting 27 that TB is present in the oral cavity [7]. A subsequent study of 50 adults with possible TB in 28 Uganda found similar sensitivity of 88%, albeit with lower specificity. Saliva is also now widely 29 used for molecular diagnosis of COVID-19, where it has high sensitivity, even among patients 30 without symptoms [8]. Using saliva as a diagnostic specimen in the COVID-19 context has been 31 shown to reduce aerosol exposure for health workers and eliminate the need for personal 32 protective equipment because it is self-collected [9]. This raises the possibility that saliva alone 33 could be used as a TB diagnostic when paired with next generation and ultra-sensitive molecular 34 participants, a study nurse collected demographic and clinical information using a structured 52 questionnaire, and then provided standardized instructions to expectorate sputum into three 53 separate cups "on the spot" [10]. Trained study staff examined the first sample using direct 54 auramine-O fluorescence microscopy (FM) [11,12]  greater than two weeks, while 71 (88%) reported subjective fever within the past seven days. 75 106 (93%) reported weight loss, including 46 (57%) with weight loss ≥5 kg. 14 (17%) reported no 107 ambulatory limitation; 47 (58%) were mildly limited with ambulation, and 20 (25%) were 108 severely affected but not bedbound. 67 patients (83%) were AFB smear-positive, including 30 109 (37%) with an AFB microscopy smear grade of 3+, 18 (22%) with a grade of 2+, nine (11%) 1+, 110 and 10 (12%) had 1-9 AFB seen per 100 high-powered fields. 13 (16%) were AFB smear-111 negative and one (1%) was missing an AFB smear microscopy result ( We also compared the semi-quantitative results of bacilli by GeneXpert for both saliva and 126 sputum, as shown in Table 2 In a prospective, observational study of consecutive sputum GeneXpert-positive TB patients in a 136 high-burden setting, we showed that diagnosis of TB using GeneXpert Ultra on saliva is feasible 137 and had a high sensitivity relative to a rigorously defined reference standard. This finding has 138 9 significant implications for the diagnosis of TB and potentially also for TB infection control. 139 Using sputum specimens for TB diagnosis poses a number of challenges, since some individuals 140 such as those with non-productive cough and young children find expectoration challenging, and 141 the associated generation of sputum aerosols poses an infection control risk for health care 142 workers and nearby patients [14]. The development of novel testing strategies that employ non-143 sputum samples for TB has been identified as a priority by the WHO, and the sensitivity 144 measured in our study is consistent with WHO's minimum target-product profile for a non-145 sputum-based test, with similar sensitivity to sputum GeneXpert among a population of 146 predominantly sputum microscopy-positive and HIV-negative individuals. Although our 147 alternative strategy of salivary GeneXpert exceeds WHO's optimal targets for cost ($4) and turn-148 around time (20 minutes) for a non-sputum-based test, if GeneXpert on saliva were shown to 149 perform well in populations for whom sputum collection is less feasible for the reasons described 150 above, the willingness to pay for and wait for results might be higher. 151

152
The use of saliva for molecular diagnosis of TB was first described in a convenience sample of 153 52 adult TB patients in Japan who were evaluated using a lab-developed, nested PCR assay that 154 was shown to have a sensitivity of 98% [15]. A more recent study of 44 sputum smear-and 155 culture-positive TB patients, including 35 in South Africa and 9 in South Korea, reported on 156 saliva as having a very low sensitivity of 39% for TB testing [16]. Sputum mycobacterial load 157 was similarly high (100% smear-positive in the South Africa/South Korea study vs. 87% in our 158 study), so these differences in diagnostic performance might be attributable to differences in 159 either sample collection or specimen processing. For example, participants were instructed to 160 rinse their mouths prior to specimen collection in the South Africa/South Korea study but not in 161 our study. Second, the South Africa/South Korea study diluted one part of the sample in two 162 parts of sample reagent as recommended by the manufacturer for sputum, while we used a 1:1 163 dilution ratio as recommended for cerebrospinal fluid, another extra-pulmonary specimen 164 without a mucoid matrix [17]. Finally, we used the GeneXpert MTB/RIF Ultra cartridge, which 165 has ten-fold better analytic sensitivity than the earlier generation GeneXpert MTB/RIF cartridge. 166 To our knowledge, we are among the first to report the performance of GeneXpert MTB/RIF 167 Ultra on saliva. 168 169 Previous studies have examined the sensitivity of a variety or oral specimens for diagnosis of 170 TB. We previously showed that oropharyngeal wash specimens paired with a lab-developed PCR 171 assay had a high sensitivity for TB diagnosis in reference to sputum mycobacterial culture on 172 previously frozen and thawed sputum, but a subsequent study failed to confirm these results [18, 173 19]. A study of Mtb PCR on buccal swabs of South African TB patients and US controls showed 174 high sensitivity (90%) and specificity (100%), although the case-control design may have 175 inflated diagnostic accuracy [7]. A recent study from the US was among the first to show that 176 saliva is a viable and accurate specimen for diagnosis of SARS-CoV2, and more sensitive and 177 less variable than nasopharyngeal swab specimens [20]. Another study carried out in Thailand 178 using saliva for diagnosis of SARS-CoV2 showed similar results, with saliva providing a 179 sensitivity of 84% and a specificity of 99% [21]. Collectively, these studies suggest that saliva is 180 a very promising novel specimen for diagnosis of respiratory tract infections. 181

182
There were a few limitations to our study. First, because the primary study objective was to 183 evaluate feasibility and preliminary sensitivity, we did not include patients with non-productive 184 cough or children, two ideal target populations for salivary testing. If, as seems plausible, these 185 populations have more paucibacillary disease, diagnostic sensitivity could be lower in these 186 populations, as suggested by the lower sensitivity observed among sputum smear-negative 187 individuals and persons living with HIV. However, in the current study, we found that even 188 though saliva is more paucibacillary than sputum as assessed by GeneXpert's semi-quantitative 189 measurement of mycobacterial load, diagnostic sensitivity was similar between the two specimen 190 types, likely because of the extremely low threshold of analytic sensitivity of the GeneXpert 191 Ultra assay [22]. Secondly, to conserve costs in this preliminary study, we did not enroll non-TB 192 patients to serve as controls, a choice that prevented us from estimating diagnostic specificity.    Legend: Shading intensity is proportional to the frequency of paired results by semi-quantitative grade across the two sample types.

Editors' Comments
Comments: C1. Abstract appears misleading in its characterization of the study population, which seems to imply that all persons being evaluated for TB were included in the study population. when in fact, it was only sputum Xpert positive that were included in this sub-study. By not including those who were sputum Xpert negative (but could have been sputum culture positive), the sensitivity is likely higher than would have been observed had everyone in the parent study been included. The reviewers suggest a revision of the manuscript to describe a feasibility study and minimize any discussion of sensitivity. I agree with them. We do acknowledge that the generalizability of our sensitivity estimates remains in question given the high smear-positivity rate, as discussed below in our response to Reviewer 3, comment C5. C1. This manuscript reports the important finding that saliva samples analyzed by GeneXpert Ultra may serve as a non-sputum alternative for TB diagnosis. Identifying viable non-sputum samples is critical to the fight against TB, and it can greatly benefit the patients and healthcare workers.

C2. Additionally
I would consider the study presented in this manuscript to be a successful proof-of-concept study, where the population and methods gave the saliva samples the best chance to perform well relative to sputum testing. It is certainly key step in establishing saliva as a diagnostic sample, though additional research will be necessary before saliva is more widely collected and used.
Specifically, the following points should be addressed: We thank Reviewer 1 for the positive reception and for recognizing the importance of the findings in the context of the proof-of-concept study design. Below, we have addressed each of Reviewer 1's comments: C2. Line 20: Replace "of" with "for". Done.
C4. Line 24: Add specific sensitivity data for the study referenced (Reference 6). Added.
C5. Lines 28-29: A citation is needed here. We have added a citation but also modified this text because the original work referenced has been superseded by a larger and more recent study from the same group.
C7. Lines 62-64: By collecting saliva after sputum, it is possible that the bacilli from the sputum remained in the mouth. I'd like this addressed as a caveat/limitation of this study and for follow-up studies to collect saliva before sputum (or at least compare before vs after).
We understand Reviewer 1's concern and have added this insight to the Discussion as an additional study limitation, incorporating additional hypotheses for evaluation as suggested by other Reviewers. It appears on page 15, at the end of the second-to-last paragraph: "Last, because saliva was collected after sputum, residual bacilli left in the mouth after coughing may have exaggerated the diagnostic yield of saliva collected without cough. Alternatively, expectoration may have depleted the mouth of Mtb bacilli, reducing sensitivity. Future studies should compare the yield with and without coughing before saliva collection."

RESPONSE TO REVIEWERS
C8. Line 118: Saliva Xpert sensitivity relative to sputum Xpert, and sputum Xpert sensitivity relative to the combined sputum culture reference standard should also be calculated and included here.
Thank you for this suggestion, which we have implemented in the Results at the top of page 9. We now report both saliva Xpert Ultra sensitivity and sputum Xpert Ultra sensitivity relative to a primary reference standard of composite sputum culture, and also relative to a secondary reference standard of sputum Xpert Ultra. We have revised the sentence to clarify that we are referring to diagnostic sensitivity, in this case the proportion of confirmed TB cases who were TB positive on saliva PCR testing (note that the cited article does not specify the reference standard used to confirm the TB): "The use of saliva for molecular diagnosis of TB was first described in a convenience sample of 52 adult TB patients in Japan in whom testing of saliva using a lab-developed, nested PCR assay targeting the 65 kD mycobacterial antigen detected TB in 98% of those with confirmed TB [19]." C10. Line 155: When you state "very low sensitivity of 39% for TB testing [16]" what sample/test is this relative to? And were the samples tested in reference 16 analyzed by GeneXpert or manual PCR?
These saliva samples were tested by sputum Xpert Ultra and the sensitivity was determined relative to mycobacterial culture. We have now added this information: We now reference the more recent Wood/Andama publication.
C12. Line 192: Reference 23 doesn't test saliva, so it is not relevant. It should be removed.
We understand the reviewer's concern and have revised the text to clearly specify that the citation does not include saliva. We have chosen to retain this information (a Cochrane systematic review of extrapulmonary specimens for TB) to show that the Xpert assay has high analytic specificity across a range of body fluid types (not only sputum), while clarifying some of the reasons why specificity might be lower in saliva, which we feel may help guide future research on the specificity of saliva Xpert. C3. Lines 32 to 34 (also lines 162-164). The sensitivity was 46% higher (not lower) in smear-positive vs smearnegative.

RESPONSE TO REVIEWERS
Thank you, we have corrected this in the Results and in the Abstract.

RESPONSE TO REVIEWERS
C4. A potential confounding aspect of the study design is that the saliva samples were collected after the sputum samples were collected. It is not obvious that waiting 2 hours is sufficient time to avoid potential 'contamination' of the saliva samples with bacteria introduced into the mouth during sputum collection.
Yes, we do understand the Reviewer's concern. As detailed above in response to Reviewer 1, comments C7 and C13, this has been added to the limitations of the study.
C5. The authors should discuss how the data from this somewhat unusual population of test subjects can be extrapolated to the typical population of persons being evaluated for TB. That is, in the overall pool of recruited test subjects, more than half were bacteriologically confirmed TB patients and almost half were smear positive. This pool of test subjects was further restricted to Xpert-positive test subjects, 87% of whom were smear positive.
Thank you for this comment. We have discussed this issue as the first limitation of the Discussion on page 14, in the first full paragraph: "First, our study population included an exceptionally high proportion of sputum smear-positive individuals. In addition, because our primary study objective was to evaluate feasibility and preliminary sensitivity, we did not include patients with non-productive cough or children, two ideal target populations for salivary testing who are likely to have more paucibacillary disease. Diagnostic sensitivity is likely to be lower in paucibacillary populations, as suggested by the lower sensitivity that we observed among sputum smear-negative individuals and among persons living with HIV. However, in the current study, we found that even though saliva is more paucibacillary than sputum as assessed by Xpert Ultra's semi-quantitative measurement of mycobacterial load, diagnostic sensitivity was similar between the two specimen types, likely because of the extremely low threshold of detection and high analytic sensitivity of the Xpert Ultra assay [11]." We have addressed the concern about restriction to Xpert-positive test subjects in our above response to the Editor's comment C1.
This is a good suggestion, which we have implemented in the Discussion at the top of page 14: "…we showed that diagnosis of TB using Xpert Ultra on saliva is feasible and had a high sensitivity relative to a rigorously defined reference standard of composite sputum mycobacterial culture." C7. Line 195. 44 smear-negative samples? Line 198 suggests it should be 44 smear-positive samples.
We apologize for the ambiguity -we meant smear-and-culture-positive samples but have now corrected this as follows: Yes, thank you for pointing this out. We have defined and adopted the shorthand phrases "Xpert Ultra" and "pre-Ultra Xpert MTB/RIF" to ensure clarity throughout the manuscript.

Introduction
C1. Thank you for the opportunity to review this manuscript. The authors have provided evidence that saliva may a suitable specimen for the diagnosis of TB, while acknowledging that future studies are needed in order to address lingering questions. The authors collected saliva in a population of sputum Xpert positive patients and showed that there were similar rates of detection of tuberculosis in saliva Xpert Ultra as compared to sputum smear microscopy using sputum culture as the reference standard. While a population of sputum Xpert positive patients is convenient, it does raise the concern that such a population is more likely to test positive with saliva Xpert Ultra as well as sputum culture, thus inflating the sensitivity higher than would be observed in a sputum Xpert negative population.
Thank you for this excellent point, which we have addressed above in our response to the Editor's comment C1.
C2. Saliva samples were treated at a 1:1 ratio with Cepheid's Sample Reagent. There are potential biosafety considerations using this lower concentration of Sample Reagent that the authors did not address in this study. The authors are encouraged to comment on biosafety concerns as well as temper their conclusions given the selective population in this study.
Thank you for this insightful comment, which we have addressed above in our response to the Editor's comment C2.

Major Comments
C3. The authors chose to treat the saliva samples with Cepheid Sample Reagent at a 1:1 ratio, which is used for some other non-pulmonary samples such as cerebrospinal fluid. In contrast, sputum is treated at a 2:1 Sample Reagent to sputum ratio. As reported, other groups, such as the South Africa/South Korea study (1) had treated saliva at a 2:1 ratio. The purpose of the sample reagent is two-fold, to liquefy viscous samples, as well as to render bacilli unviable and thus lowering the biohazard risk. As reported in Helb et al. (2) killing assays in spiked sputum were utilizing two volumes of Sample Reagent per volume of sputum. That grouped showed that at that concentration, after 15 minutes, viability was reduced by at least 8-logs in sputum. This is an important consideration in the safety of laboratory staff to ensure that they are not unnecessarily exposed to viable bacilli.
To the reviewer's knowledge, no such killing assays have been conducted for saliva samples at a lower Sample Reagent concentration. If samples are handled within a biosafety cabinet at all times, concerns regarding a lower Sample Reagent concentration are alleviated; however, if treated samples are opened on an open bench, lab workers could potentially be exposed to viable bacilli. The authors are encouraged to specify that samples should be handled within a biosafety cabinet at all times or conduct killing assays to show that lower concentrations of Sample Reagent still result in an adequate reduction in bacilli viability.

RESPONSE TO REVIEWERS
As noted in our response to the Editor's comment C2, we have added a limitation about the biosafety concerns of treating saliva at a ratio of 1:1, including the importance of using a biosafety cabinet. We have also highlighted the need for additional studies of the risks and benefits of different SR:saliva dilutions.
C4. In the Abstract, the authors state, "The objective... was to determine the performance of GeneXpert MTB/RIF Ultra (Xpert) testing on saliva for active tuberculosis (TB) disease among consecutive adults undergoing diagnostic evaluation." This statement is misleading. While the parent study did enroll consecutive adults undergoing TB evaluation, in the sub-study, saliva samples were only collected among adults that had already tested positive for tuberculosis using sputum Xpert. This has important implications in the interpretation of the reported sensitivity. As is noted in the Discussion regarding Wood et al. (3), case-control study designs are prone to inflate diagnostic accuracy. The authors of this study do not acknowledge this same limitation in their study design, which is effectively a "case-only" study population. The authors do note the limitation of not being able to determine specificity, but do not note the limitation of inflated sensitivity. The authors are encouraged to acknowledge this major limitation throughout the manuscript.
Thank you for this insightful point, which highlights the issue of selection bias in case-only diagnostic studies. We have added this as an additional limitation on page 14 of the Discussion, along with additional analyses summarized in our above response to the Editor's comment C2: "Although case-only studies are more cost-efficient, they have the limitation of potentially inflating sensitivity by excluding patients who would have been diagnosed with TB by a more sensitive inclusion criterion, such as sputum mycobacterial culture or clinical evaluation [25]. However, we found that only two of the 153 patients screened for this study were likely to be sputum Xpert Ultra-negative and culture-positive, making the selection bias too small to meaningfully influence our sensitivity estimates." C5. Saliva was collected "at least two hours after sputum collection." Is this a sufficient period of waiting? Why was saliva not collected before sputum? It could be reasonably assumed that the process of expectorating sputum would leave some bacilli in the mouth that could then later be released in saliva. Had the order of collection been reversed (saliva first and then sputum), one might expect detection in saliva Xpert to be lower. The authors are strongly encouraged to discuss the rationale for this methodology, and discuss whether future studies should explore sample collection order.