Plasma calprotectin as a biomarker of mortality at antiretroviral treatment initiation in advanced HIV – pilot study

Background: In advanced HIV, significant mortality occurs soon after starting antiretroviral treatment (ART) in low- and middle-incomes countries. Calprotectin is a biomarker of innate response to infection and inflammatory conditions. We examined the association between plasma calprotectin collected before ART treatment and mortality among individuals with advanced HIV. Methods: We conducted a pilot case-cohort study among HIV infected adults and adolescents over 13 years old with CD4+ <100/mm3 at ART initiation at two Kenyan sites. Participants received three factorial randomised interventions in addition to ART within the REALITY trial (ISRCTN43622374). Calprotectin collected at baseline (before ART) and after 4 weeks of treatment was measured in archived plasma of those who died within 24 weeks (cases) and randomly selected participants who survived (non-cases). Association with mortality was assessed using Cox proportional hazards models with inverse sampling probability weights and adjusted for age, sex, site, BMI, viral load, randomised treatments, and clustered by CD4+ count (0-24, 25-49, and 50-99 cells/mm3). Results: Baseline median (IQR) plasma calprotectin was 6.82 (2.65–12.5) µg/ml in cases (n=39) and 5.01 (1.92–11.5) µg/ml in non-cases (n=58). Baseline calprotectin was associated with age, neutrophil count and the presence of cough, but not other measured indicators of infection. In adjusted multivariable models, baseline calprotectin was associated with subsequent mortality: HR 1.64 (95% CI 1.11 - 2.42) and HR 2.77 (95% CI 1.58 - 4.88) for deaths during the first twenty-four and four weeks respectively. Calprotectin levels fell between baseline and 4 weeks among both cases and non-cases irrespective of randomised interventions. Conclusions: Among individuals with advanced HIV starting ART in Kenya, plasma calprotectin may have potential as a biomarker of early mortality. Validation in larger studies, comparison with other biomarkers and investigation of the sources of infection and inflammation are warranted.


Introduction
Approximately one-quarter of individuals newly diagnosed with HIV in sub-Saharan Africa have advanced disease at presentation (IeDEA and ART Cohort Collaborations et al., 2014). Advanced HIV is characterized by immunosuppression, infection and immune activation which may independently drive mortality despite antiretroviral therapy (ART). Measurements of soluble biomarkers such as soluble CD14, C-reactive protein and IL-18 have highlighted that inflammation and innate immune responses predict all-cause mortality, cardiovascular events, and other morbidities in HIV infected individuals, even after starting ART treatment (Duprez et al., 2012;Kuller et al., 2008;Sandler & Douek, 2012). Overall, innate immune activation seems more important than T-cell activation for disease progression in sub-Saharan Africa (Hunt et al., 2016;Serrano-Villar et al., 2014). Inflammation and co-infection can also arise from disruption of intestinal tight junctions leading to increased mucosal permeability. Translocation from the intestine of bacteria and their products including lipopolysaccharides has been demonstrated in some (Brenchley et al., 2006;Marchetti et al., 2013;Nazli et al., 2010;Sandler & Douek, 2012), but not all studies (Fitzgerald et al., 2019).
Calprotectin is a soluble 24 kDa dimer of calcium-binding proteins S100A8 and S100A9 (Brophy & Nolan, 2015) produced by neutrophils and other cells following activation in response to infection and inflammation. Calprotectin, measured in either stool or plasma, is a recognised biomarker of inflammation and bacterial infections including sepsis (Banerjee et al., 2015;Bjarnason, 2017;Huang et al., 2016;Jonsson et al., 2017;Simm et al., 2016;Walsham & Sherwood, 2016). The S100A9 subunit of plasma calprotectin has been associated with reduced immune reconstitution after ART (Drozd et al., 2016), enhanced antimicrobial defence transiently induced by antiviral treatment (Muller et al., 1994) and HIV-associated neurocognitive disorders (Colon et al., 2016). Njunge et al. recently demonstrated an association between increased plasma calprotectin and early post-discharge mortality among HIV-uninfected children hospitalized for severe acute malnutrition (Njunge et al., 2019).
We considered that calprotectin may be of value as a prognostic biomarker in advanced HIV and conducted a pilot study to evaluate associations between plasma calprotectin and mortality in individuals with advanced HIV infection prior to ART initiation who participated in the Reduction of Early Mortality in HIV Infected Adults and Children Starting Antiretroviral Therapy (REALITY) trial (Hakim et al., 2017) (Kityo et al., 2018) (Mallewa et al., 2018.

Study population
The REALITY trial (ISRCTN43622374) enrolled HIV-infected adults and children aged five years or more with a CD4 + T cell count <100 cells/mm 3 and without previous ART treatment at 8 sites in Kenya, Uganda, Malawi, and Zimbabwe. Participants in the REALITY trial were enrolled between August, 2013 and April, 2015 and randomised to three factorial treatments compared to standard of care: enhanced antimicrobial prophylaxis (single-dose albendazole, 5 days of azithromycin, 12 weeks of fluconazole (100 mg), and 12 weeks of fixed-dose combination of cotrimoxazole (800/160 mg)/isoniazid (300 mg)/pyridoxine (25 mg) once daily) (Hakim et al., 2017); additional raltegravir (Kityo et al., 2018); and ready to use supplementary food (RUSF) (Mallewa et al., 2018). This pilot study capitalised on a broader ongoing immunology case-cohort sub-study that included study participants aged 13 years or more with a sample set of plasma, baseline stool, PBMCs and data at two Kenyan sites: Kilifi County Hospital and the Academic Model for the Prevention and Treatment of HIV/AIDS Centre at Moi Teaching Referral Hospital, Eldoret.

Study design
This pilot was a case-cohort study. The REALITY trial had enrolled a total of 139 participants in Kilifi, of whom 29 (20%)

Amendments from Version 1
Abstract, in methods, clarified sub-study minimum age 13 years (R2); that calprotectin was measured before ART initiation; that mortality to 24 weeks was used (as per primary endpoint of trial) rather than 48 weeks extended follow up (R1&2); the analysis is more succinctly described; in results, effect sizes 'for death' changed to 'for subsequent mortality' and that 4-week changes involved cases who survived to 4 weeks and controls.
Methods: study population: paragraph 1, description of the parent trial moved up and in paragraph 2, moved up the names of the sites for this sub-study; study design: paragraph 1, reworded participant selection percentages (R1&2); study design, paragraph 2 and ELISA tests, paragraph 1: stated baseline samples taken prior to ART initiation.
Methods, statistical analysis: clarified regression analysis indicating it was stepwise, including variables with biological plausibility, retaining variables with P<0.1 (R1); better explaining stratification and methods for examining randomised interventions in the last paragraph.
Results, under baseline plasma calprotectin, added regression coefficients to the text, reworded infection markers (R1), and in Table 2 added interpretations (R1); under association with mortality, added hazards ratios to the text (R2); Table 3, added CD4 strata to the footnote (R1); under plasma calprotectin after 4 weeks, added regression coefficients to the text (R2); Table 4, added row for all participants and clarified cases must have survived 4 weeks to be included here (R1), indicating N=78 below Table 4 and Figure 2.
Discussion, paragraph 1, clarification of 24-week mortality and moved up comments on antimicrobial prophylaxis; paragraphs 3 & 4, expanded discussion of sample size and other limitations (R1&2); under conclusions, clarified sample timing and added potential practical usage (R1).
Any further responses from the reviewers can be found at the end of the article REVISED died, and 195 in Eldoret, of whom 14 (7.2%) died ( Figure 1). However, in Kilifi baseline CD8 measurements were missing on approximately one-third of participants due to reagent unavailability at specific time periods (i.e., missing at random) and one-quarter were also missing stored specimens for similar reasons. Therefore, the required number of participants in Kilifi were selected from those with complete samples and baseline CD8 available to ensure that data could be obtained, and weighted (see below) to reflect the original trial population. The immunology case-cohort sub-study randomly selected 45% of all participants at Kilifi and 10% of study participants at Eldoret as a sub-cohort in order to reflect enrolment and mortality in the full REALITY trial, stratified by CD4 + count (0-24, 25-49, and 50-99 cells/mm 3 ) to avoid imbalance in this exposure, plus any remaining unselected deaths by week 24 (the trial primary endpoint). Non-cases were randomly selected from those who survived through 24 weeks from the trial database using the uniform random number function in STATA.
Deaths were weighted as 1 and non-deaths were weighted as 106/42 in Kilifi and 174/16 in Eldoret. These values were chosen in order that the sample represented the full trial population in terms of deaths and survivors at these sites using the inverse probability of selection from all REALITY patients ≥13 years of age alive at week 24 (regardless of immunology sub-study membership and available samples). Demographic, clinical and laboratory data were collected during the REALITY trial using standardised case report forms (Hakim et al., 2017). Complete blood counts, including neutrophils and CD4 + counts were done at local laboratories. Samples analysed were those prior to receiving ART at baseline, then 4 weeks after starting ART.
Enzyme-linked immunosorbent assay (ELISA) Plasma calprotectin was measured in duplicate at recommended dilutions using a solid-phase enzyme-linked immunosorbent assay (ELISA) human calprotectin kit (Hycult Biotech HK379-02) according to the manufacturer's instructions and the absorbance read at 450 nm using a Synergy 4 (BioTek) plate reader. Calprotectin was measured using samples collected at baseline, before ART treatment was initiated, and at four weeks after ART initiation. The four-week time period allowed an appropriate time frame for the detection of immunological changes following treatment initiation.

Statistical analysis
Data were analysed using STATA version 13.1 (STATA corp. TX, USA). The baseline characteristics of the selected study participants were presented as mean (SD) for normally distributed variables, as median with interquartile range (IQR) for nonnormally distributed variables, and as numbers (percentage) for categorical data. To compare differences in characteristics between cases and non-cases, Student's t-test and Mann-Whitney rank-sum test (of non-normally distributed variables) were used. Chi-square or Fisher's exact tests were used to compare proportions and Spearman's correlation to assess correlation.
Factors associated with plasma calprotectin levels at baseline were assessed using a stepwise linear regression model including age, sex, site, CD4 + , viral load and neutrophil count, but excluding the randomised interventions. Individual clinical and laboratory features of infection with biological plausibility: the presence of fever, cough, diarrhoea, known tuberculosis, measured body temperature, CD8 + T cell count, and cryptococcal antigen test were then added one by one, retaining only those which were statistically significant at P<0.1. The beta coefficient indicated the strength of the effect on plasma calprotectin.
The association between baseline plasma calprotectin and mortality was assessed using a stratified Cox proportional hazards regression model with inverse probability weights (Buchanan et al., 2014) with three strata by CD4 count (0-24, 25-49, and 50-99 cells/mm 3 ) to reflect the proportions selected in the casecohort design to help address bias, as described above. Time at risk was defined from enrolment to 24 weeks which was the primary endpoint of the parent trial. Mortality within the first four weeks, when most deaths occurred, was also assessed. The hazard ratio per unit increase in calprotectin was estimated, after adjustment for age, sex, BMI, log viral load, site and each of the randomised treatment arms. Akaike (AIC) and Bayesian (BIC) information criteria were calculated to assess model performance.
The Wilcoxon paired sign-rank test was used to assess the change in plasma calprotectin between baseline and after four weeks following treatment. Generalised linear modelling, adjusted for age, sex and site and the three randomised interventions and stratified by CD4 + group as above was used to examine relative risk for subsequent mortality.

Ethical considerations
The study protocol was approved by the Kenya Medical Research Institute (KEMRI) Scientific and Ethics Review Committee approval number SSC 2231. Written informed consent was obtained from all participants using local languages, which included permission for storage and testing for this work.

Characteristics of study participants
Of the selected 97 participants, 39 were cases and were 58 non-cases ( Figure 1). The baseline characteristics of the participants are shown in Table 1. A total of 16/39 (41%) cases died between enrolment and four weeks, and a further 23/39 (59%) cases died between four and 24 weeks. Death occurred at a median of 31 (Interquartile range, IQR 18-72) days after enrolment. Cases had a lower BMI, haemoglobin and CD8 + T cell counts than non-cases, and fewer cases were randomized to receive RUSF in univariate analyses.

Discussion
This pilot study focused on assessing the association with pre-ART plasma calprotectin and mortality within 24 weeks among patients with advanced HIV disease defined by a very low CD4 + cell count (<100cells/mm 3 ). Plasma calprotectin at the time of ART initiation was associated with subsequent mortality within 24 weeks as well as in the first 4 weeks. There was no evidence that the enhanced opportunistic infection intervention in the REALITY trial, which was associated with a 27% reduction in mortality to 24 weeks, (Hakim et al., 2017) affected changes in calprotectin during the first 4 weeks. Interestingly, although none of the three interventions significantly affected the change in plasma calprotectin in this pilot study, the point estimate for enhanced antimicrobial prophylaxis, which was associated with reduced mortality in the parent trial (Hakim et al., 2017) was in a negative direction. However, the sample size for this exploratory analysis was limited. This pilot study requires validation in a larger sample and in other clinical settings before being used to guide further investigations or specific interventions.
To the best of our knowledge, this is the first study looking at plasma calprotectin in the context of mortality in advanced HIV. Previous studies have focused on faecal calprotectin as a biomarker of enteropathy, which is typically elevated in HIV-positive compared to HIV-negative individuals and progressively increases with a reduction in CD4 + T cell count   opportunistic infections among advanced HIV-positive patients, rather than only inflammation due to infection with HIV. However, markers of infection markers apart from cough were not statistically significantly associated with baseline plasma calprotectin. This could result due to the small number of samples tested, hence insufficient power.
The main strength of our study is that it was carried out in typical African hospital-based HIV clinics in which all eligible patients from both urban and peri-urban areas were recruited and is thus a reasonable reflection of advanced HIV patients in sub-Saharan Africa. Moreover, the use of ELISA is feasible in these setups, and already in application as a confirmatory test for HIV positive diagnosis. Besides the sample size of this pilot study, limitations included that despite weighting there was potential for bias during sample selection as included participants were required to have a full set of samples, which varied by site, and children under thirteen years old were excluded. The parent trial only enrolled patients with advanced disease, therefore we recommend that future studies be carried on patients with less advanced HIV to elucidate whether plasma calprotectin has similar predictive value, along with other proteins, metabolites and cytokines.

Conclusions
Findings from this pilot study suggest that plasma calprotectin at the time of ART initiation has value in predicting early mortality among HIV patients with advanced disease. There is at least one quantitative serum calprotectin lateral-flow test available, and ELISA-or lateral flow-based tests may be useful in patient care. However, further validation of plasma calprotectin as a clinical tool is needed before incorporating the biomarker to guide enhanced investigation for infections, more frequent follow up or specific interventions.

Data availability
Underlying data De-identified REALITY trial data are available from MRC CTU at UCL, which encourages optimal use of data by employing a controlled access approach to data sharing, incorporating a transparent and robust system to review requests and provide secure data access consistent with the relevant ethics committee approvals. All requests for data are considered and can be initiated by contacting mrcctu.ctuenquiries@ucl.ac.uk Quoting "REALITY Trial Immunology".

Rupak Shivakoti
Columbia University Mailman School of Public Health, New York, NY, USA This is a pilot study assessing the relationship of plasma calprotectin, a marker of inflammation, with mortality among individuals with advanced HIV initiating treatment. The authors use a nested case-cohort design to measure inflammation at time of ART initiation and use data from the parent study to assess mortality during treatment. Their results show that plasma calprotectin could be useful marker for subsequent mortality, especially within the next 4 weeks, in individuals with advanced HIV. This is a clearly written manuscript and the design seems appropriate. There are, however, a few major questions that should be addressed.
Comments: Given differences in immunity for children compared to adults, the percentage of children in this study should be specified and it would be helpful to know whether the baseline levels of calprotectin was different between children and adults. It is likely that the study is underpowered to look at association with mortality separately for children and adults but the authors should discuss whether combining them (including under 5 as suggested in discussion) in one analysis might or might not make sense for this specific marker.
○ It is not clear why 45% of Kilifi participants vs. only 10% of Eldoret were selected for the random subcohort. It is also not clear how this case-cohort was selected from the larger immunology case-cohort.

○
The case-cohort design has a random subcohort. How many cases out of 39 were part of the random subcohort? It is not clear from reading this what the sample size of the random ○ subcohort is.
In the methods, the authors say: "The linear regression to examine whether subsequent death and the 3 randomized interventions were associated with the change in plasma calprotectin between baseline and 4 weeks" The mortality analysis is confusing as written. I don't think the results were presented in the manuscript? For example, it is not clear what the exposure and outcome variable for this analysis is. If the exposure is change in plasma calprotectin, then the outcome is presumably death after 4 weeks (which would exclude many deaths) and a linear regression does not seem appropriate. But if the exposure is death (i.e. before 4 weeks), then a 4 week sample would not be available.
○ ○ Sample size should be stated for the tables and figures. Was the random subcohort used in any of the presented analysis? E.g. due to random selection, it may be most useful for Table  2 analysis if there are enough cases in the subcohort.
○ Given the small sample size, selecting covariates for multivariable model based on significance on univariate models can present some issues. Lack of significance could be due to the limited power for this specific study although a real association may exist.

If applicable, is the statistical analysis and its interpretation appropriate? Partly
Are all the source data underlying the results available to ensure full reproducibility? Yes

Are the conclusions drawn adequately supported by the results? Yes
Competing Interests: No competing interests were disclosed.

I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.
Author Response 29 Sep 2020

James Berkley,
We thank the reviewer for their comments providing an opportunity to improve our manuscript. Please find out point-by-point responses below. Reviewer 2 Given differences in immunity for children compared to adults, the percentage of children in this study should be specified and it would be helpful to know whether the baseline levels of calprotectin was different between children and adults. It is likely that the study is underpowered to look at association with mortality separately for children and adults but the authors should discuss whether combining them (including under 5 as suggested in discussion) in one analysis might or might not make sense for this specific marker.
In the parent study, children recruited were 5-12 years, adolescents were 13-17 years and adults were 18 years and above. In this pilot sub-study we only included participants ≥ 13 years and above. As such, data from children is out of scope and not available for this plot study. It is not clear why 45% of Kilifi participants vs. only 10% of Eldoret were selected for the random subcohort. It is also not clear how this case-cohort was selected from the larger immunology case-cohort.

The broader case cohort study involved two of the sites in the broader ongoing immunology casecohort sub-study which randomly selected 45% of all Kilifi participants and 10% of Eldoret study participants as a sub-cohort reflecting survival and mortality in the full REALITY trial. This has now been clarified on page 6.
In the methods, the authors say: "The linear regression to examine whether subsequent death and the 3 randomized interventions were associated with the change in plasma calprotectin between baseline and 4 weeks" The mortality analysis is confusing as written. I don't think the results were presented in the manuscript? For example, it is not clear what the exposure and outcome variable for this analysis is. If the exposure is change in plasma calprotectin, then the outcome is presumably death after 4 weeks (which would exclude many deaths) and a linear regression does not seem appropriate. But if the exposure is death (i.e. before 4 weeks), then a 4 week sample would not be available.

We have revised clarified this in the manuscript in methods on page 9 and results on page 11. In the main analysis, factors associated with plasma calprotectin levels at baseline were assessed using a linear regression model. We have now analysed the associations between change in plasma calprotectin between baseline and 4 weeks and deaths after 4 weeks by calculating relative risk.
Sample size should be stated for the tables and figures. Was the random subcohort used in any of the presented analysis? E.g. due to random selection, it may be most useful for Table  2 analysis if there are enough cases in the subcohort.

The sample size has been added to tables and figures. All analyses include the random subcohort and all additional deaths as is usual in a nested case-cohort study.
Given the small sample size, selecting covariates for multivariable model based on significance on univariate models can present some issues. Lack of significance could be due to the limited power for this specific study although a real association may exist.
We agree, and we commented in the discussion that this pilot study should be validated in a larger study.
It is useful to also see the hazards and 95% CI (not just p-values) in text of results instead of just the

Results:
Major comments: The authors described variables associated with baseline calprotectin, but this analysis is not described in the methods. Please add methods for these results in the methods section and explain why this was done.
○ Results mention deaths within 4 weeks as an outcome, but this is not include in the methods. Please ensure all results have methods described.

Methods: Study design:
Major comments: Please clarify why there were 45% randomly selected from Kilfi but only 10% randomly selected from Eldoret? How were these percentages determined for selection? ○ Minor comment: Please define "baseline" in the context of calprotectin samples collected at baseline and 4 weeks after treatment initiation. So baseline is prior to ART initiation, yes? Please clarify.

Statistical Analysis:
Major comments: Please clarify the linear regression analysis. Was the linear regression model was univariable or was multivariable? This seems to be describing stepwise model building, yes? It states that the authors only retained those variables which were statistically significant. How was the order of variable addition determined? Generally, model building should not rely solely on statistical significance. Variables should be determined on the basis of plausibility a priori as well. A variable may be associated with an outcome, but especially in a small pilot, this might not be statistically significant due to insufficient power, so significance should not be the determining factor for its inclusion in the final model. Important to look at point estimates in addition to p-values and known relationships between the variables in identifying confounders to include in subsequent adjusted analyses.

○
Please clarify in the Cox proportional hazard regression, how was this stratified by CD4 count. What strata were used and how were these determined? ○ Please clarify the following in relation to survival time: the abstract states cases died within ○ 24 weeks and non-cases were those who survived for 48 weeks. Manuscript states time at risk was 24 weeks and there is no mention of 48 weeks. What about those who died between 24 and 48 weeks? And the results present 4 weeks as an outcome end point. Please ensure all results have associated methods and that the Abstract has consistent language with the manuscript main body.
Why was the Wilcoxon paired sign-rank test was performed (final paragraph of Statistical Analysis). Please explain the objective/purpose of this analysis and how the covariates included in the model were determined.

Characteristics of study participants:
Minor comments: Is enrollment misspelled (enrolment)? ○ Final sentence is grammatically incorrect. I think deleting the "lower" before "hemoglobin"would correct this.

Baseline plasma calprotectin:
Minor comments: Please define what "crude" means.

Major comments:
The following language should be adjusted to reduce reliance on p-values for scientific interpretation: "other putative markers of infection (CD8+ T cells, fever, tuberculosis, body temperature, diarrhoea, and cryptococcal antigen test) did not emerge as factors affecting plasma calprotectin levels." Please change this language to specify these were not "statistically significant." This could be a result of low power. The point estimates should also be discussed, especially since this is a small pilot.

Plasma calprotectin after 4 weeks:
Please show or at least describe the following data: "There was no evidence that any of the three randomized interventions affected the change in calprotectin between baseline and 4 weeks (data not shown)." Rephrase to describe lack of statistical significance and also discuss the point estimates.

Discussion:
Minor comment: The following could be clarified: "This pilot study focused on assessing plasma calprotectin mortality among patients with advanced HIV disease determined by a very low CD4+ cell count." Reword to include more detail: "This pilot study focused on assessing the association ○ with pre-ART plasma calprotectin and mortality within 24 weeks among patients with advanced HIV disease defined by a very low CD4+ cell count (<100cells/µl)." Major comments: The following is confusing: "It appeared that calprotectin levels at ART initiation are likely to be predictive of deaths occurring in the first 4 weeks." Please use more concrete language to describe your main finding. Also, it is unclear if 24 weeks or 4 weeks is your survival end point. Calprotectin levels were associated with both, right? ○ Please describe your limitations specifically. "However, the sample size for this exploratory analysis was limited. This pilot study requires validation in a larger sample and in other clinical settings before being used to guide further investigations or specific interventions." Should be explicit about your limitations, how they may have impacted your results, and thus why next steps are needed. You do this at the end of the discussion section, so confusing why you mention this in two places. This is also the first time the term "exploratory" is used in the manuscript. Please include this in the methods so it's clear what you mean by this term.
○ Please adjust language in the following: "However, markers of infection markers apart from cough were not associated with baseline plasma calprotectin." The phrase "not associated" is not ideal. Better to state: "However, markers of infection markers apart from cough were not statistically significantly associated with baseline plasma calprotectin." And discuss the point estimate. Why do you think cough associated but others not? Is this just an issue of multiple comparisons and a spurious association or do you have insufficient power to detect other infection markers? I think this is worth further discussion.

○
The authors mention "…potential for bias during sample selection as included participants were required to have a full set of sample" but you state earlier that you are confident this was missing at random, so this is assumed not to have biased your results, right? Please include this information here if you are confident in the MAR assumption and it's impact on your analysis.

○
Can you please discuss if it will be feasible for the ELISA test to be used in low/middle income locations where it could be of most benefit due to high rates of ART initiation at an advanced HIV stage? Do you think ART clinics have the infrastructure and resources to perform this test to inform patient care or are costs and lab resources potentially another barrier?

Are sufficient details of methods and analysis provided to allow replication by others? Partly
If applicable, is the statistical analysis and its interpretation appropriate?
Thank you, this has been added to the bottom of page 8. Please clarify why there were 45% randomly selected from Kilfi but only 10% randomly selected from Eldoret? How were these percentages determined for selection?
The broader case cohort study involved two of the sites in the broader ongoing immunology casecohort sub-study which randomly selected 45% of all Kilifi participants and 10% of Eldoret study participants as a sub-cohort reflecting survival and mortality in the full REALITY trial. This has now been clarified on page 6. Please define "baseline" in the context of calprotectin samples collected at baseline and 4 weeks after treatment initiation. So baseline is prior to ART initiation, yes? Please clarify. Calprotectin was measured using samples collected at baseline, before ART treatment was initiated, and four weeks after ART treatment initiation. We have clarified this on page 7. Please clarify the linear regression analysis. Was the linear regression model was univariable or was multivariable? This seems to be describing stepwise model building, yes? It states that the authors only retained those variables which were statistically significant.
How was the order of variable addition determined? Generally, model building should not rely solely on statistical significance. Variables should be determined on the basis of plausibility a priori as well. A variable may be associated with an outcome, but especially in a small pilot, this might not be statistically significant due to insufficient power, so significance should not be the determining factor for its inclusion in the final model. Important to look at point estimates in addition to p-values and known relationships between the variables in identifying confounders to include in subsequent adjusted analyses.
We have now clarified on page 8 and

English (United Kingdom) spelling was used since this is a UK-based journal.
Final sentence is grammatically incorrect. I think deleting the "lower" before "hemoglobin"would correct this. Thank you. We have revised as suggested. Please define what "crude" means.
The following language should be adjusted to reduce reliance on p-values for scientific interpretation: "other putative markers of infection (CD8+ T cells, fever, tuberculosis, body temperature, diarrhoea, and cryptococcal antigen test) did not emerge as factors affecting plasma calprotectin levels." Please change this language to specify these were not "statistically significant." This could be a result of low power. The point estimates should also be discussed, especially since this is a small pilot.
We have revised this to indicate P<0.1 was used for selection. We agree with the issue of reliance on P-values but regarded that having many variables in the final model was undesirable in a small pilot.  . This is added to page 6. All analyses are stratified.
Please show or at least describe the following data: "There was no evidence that any of the three randomized interventions affected the change in calprotectin between baseline and 4 weeks (data not shown)." Rephrase to describe lack of statistical significance and also discuss the point estimates.
These estimates have been added on page 11 and discussed on page 13. The following could be clarified: "This pilot study focused on assessing plasma calprotectin mortality among patients with advanced HIV disease determined by a very low CD4+ cell count." Reword to include more detail: "This pilot study focused on assessing the association with pre-ART plasma calprotectin and mortality within 24 weeks among patients with advanced HIV disease defined by a very low CD4+ cell count (<100cells/µl)." This has been reworded as suggested.
The following is confusing: "It appeared that calprotectin levels at ART initiation are likely to be predictive of deaths occurring in the first 4 weeks." Please use more concrete language to describe your main finding. Also, it is unclear if 24 weeks or 4 weeks is your survival end point. Calprotectin levels were associated with both, right? Thank you. The primary analysis is to 24 weeks and this has been clarified along with a more concrete statement. Please describe your limitations specifically. "However, the sample size for this exploratory analysis was limited. This pilot study requires validation in a larger sample and in other clinical settings before being used to guide further investigations or specific interventions." Should be explicit about your limitations, how they may have impacted your results, and thus why next steps are needed. You do this at the end of the discussion section, so confusing why you mention this in two places. This is also the first time the term "exploratory" is used in the manuscript. Please include this in the methods so it's clear what you mean by this term. Thank you. The term 'exploratory' has been removed and we have revised the limitations as suggested.
Please adjust language in the following: "However, markers of infection markers apart from cough were not associated with baseline plasma calprotectin." The phrase "not associated" is not ideal. Better to state: "However, markers of infection markers apart from cough were not statistically significantly associated with baseline plasma calprotectin." And discuss the point estimate. Why do you think cough associated but others not? Is this just an issue of multiple comparisons and a spurious association or do you have insufficient power to detect other infection markers? I think this is worth further discussion.
Thank you, we have made this change. Insufficient power is quite possible and we have included that.
The authors mention "…potential for bias during sample selection as included participants were required to have a full set of sample" but you state earlier that you are confident this was missing at random, so this is assumed not to have biased your results, right? Please include this information here if you are confident in the MAR assumption and it's impact on your analysis. We have remove 'missing at random' since sites were differentially affected and mentioned this under limitations.
Can you please discuss if it will be feasible for the ELISA test to be used in low/middle income locations where it could be of most benefit due to high rates of ART initiation at an advanced HIV stage? Do you think ART clinics have the infrastructure and resources to perform this test to inform patient care or are costs and lab resources potentially another barrier?
We have added this to the conclusions: 'Although capacity for an ELISA-based test may be limited at many clinics, there is at least one quantitative serum calprotectin lateral-flow test available'.
Competing Interests: No competing interests were disclosed.