Evaluating the reactivation of herpesviruses and inflammation as cardiovascular and cerebrovascular risk factors in antiretroviral therapy initiators in an African HIV-infected population (RHICCA): a protocol for a longitudinal cohort study

Introduction In Sub-Saharan Africa, the rising rates of cerebrovascular and cardiovascular diseases (CBD/CVD) are intersecting with an ageing HIV-infected population. The widespread use of antiretroviral therapy (ART) may confer an additive risk and may not completely suppress the risk associated with HIV infection. High-quality prospective studies are needed to determine if HIV-infected patients in Africa are at increased risk of CBD/CVD and to identify factors associated with this risk. This study will test the hypothesis that immune activation and dysfunction, driven by HIV and reactivation of latent herpesvirus infections, lead to increased CBD/CVD risk in Malawian adults aged ≥35 years. Methods and analysis We will conduct a single-centre, 36-month, prospective cohort study in 800 HIV-infected patients initiating ART and 190 HIV-uninfected controls in Blantyre, Malawi. Patients and controls will be recruited from government ART clinics and the community, respectively, and will be frequency-matched by 5-year age band and sex. At baseline and follow-up visits, we will measure carotid intima-media thickness and pulse wave velocity as surrogate markers of vasculopathy, and will be used to estimate CBD/CVD risk. Our primary exposures of interest are cytomegalovirus and varicella zoster reactivation, changes in HIV plasma viral load, and markers of systemic inflammation and endothelial function. Multivariable regression models will be developed to assess the study’s primary hypothesis. The occurrence of clinical CBD/CVD will be assessed as secondary study endpoints. Ethics and dissemination The University of Malawi College of Medicine and Liverpool School of Tropical Medicine research ethics committees approved this work. Our goal is to understand the pathogenesis of CBD/CVD among HIV cohorts on ART, in Sub-Saharan Africa, and provide data to inform future interventional clinical trials. This study runs between May 2017 and August 2020. Results of the main trial will be submitted for publication in a peer-reviewed journal. Trial registration number ISRCTN42862937.


GENERAL COMMENTS
This is an important and well-designed study to examine the associations of HIV infection, immune activation and reactivation of latent herpesvirus infection with risk of cerebrovascular and cardiovascular disease. I have the following comments and suggestions with regards to the statistical analysis plans. 1. For the sample size calculations: a. Please clarify if the expected rate of clinically significant vasculopathy of 18.4% is among the HIV positive group or the control group. b. It is noted that 18.4% of study participants have abnormal PWV at baseline. Will these patients be excluded from the analyses? 2. Please clarify if participants will be withdrawn from the study if they become pregnant during the study period. 3. It is my understanding that the investigators are using the term "interim analysis" to refer to an analysis to be conducted of data up to an interim follow-up visit, rather than of the complete data. Since they have not specified otherwise, I assume this analysis will take place after all participants have reached the interim time point. It may be less confusing for readers if the investigators refer to these analyses as "6-month analyses" or "12-month analyses", for example, to distinguish them from interim analyses conducted in a clinical trial context, which would be accompanied by specific stopping rules, etc. 4. Will missing data be imputed? 5. The investigators may wish to consider the use of pattern mixture models to assess the affect of differential dropout by HIV status, reactivation of herpes viral infection and HIV viral antigenemia on the estimated effects of these characteristics on risk of cerebrovascular and cardiovascular disease. 6. Will the exclusion of study participants who die from analyses depend on whether study participants die from CBD or CVD? 7. In the study design section, it is indicated that the study schedule is described in " Figure 2". Please correct this to "

GENERAL COMMENTS
Great initiative authors; Just a small comment: On page 15, you state that in a subset of participants some tests will be done but you make no mention how these participants will be selected. Same for page 5, the small number that will have an MRI.

VERSION 1 -AUTHOR RESPONSE
Reviewers' Comments to Author: Reviewer: 1 Reviewer Name: Janet Raboud Institution and Country: University of Toronto, Canada Please state any competing interests or state 'None declared': None declared This is an important and well-designed study to examine the associations of HIV infection, immune activation and reactivation of latent herpesvirus infection with risk of cerebrovascular and cardiovascular disease.
RESPONSE: We thank the reviewer for these comments.
I have the following comments and suggestions with regards to the statistical analysis plans.
1. For the sample size calculations: a. Please clarify if the expected rate of clinically significant vasculopathy of 18.4% is among the HIV positive group or the control group.
RESPONSE: This is the expected rate in the HIV positive participants. We have clarified this in the sample size paragraph.
b. It is noted that 18.4% of study participants have abnormal PWV at baseline. Will these patients be excluded from the analyses?
RESPONSE: We will be conducting two types of analysis. The first will examine the progression of surrogate markers over time from whatever baseline value -participants with abnormal PWV at baseline will not be excluded from this analysis. The second will examine the development of newonset vasculopathy -participants with abnormal PWV at baseline will be excluded from this analysis.
We have clarified this point in the manuscript.
2. Please clarify if participants will be withdrawn from the study if they become pregnant during the study period.
RESPONSE: We confirm that if the study participant becomes pregnant after recruitment, they will be withdrawn. Given the age structure of our population, we do not think that this will have a substantial impact on our lost to follow-up numbers. We have clarified this in the manuscript.
3. It is my understanding that the investigators are using the term "interim analysis" to refer to an analysis to be conducted of data up to an interim follow-up visit, rather than of the complete data. Since they have not specified otherwise, I assume this analysis will take place after all participants have reached the interim time point. It may be less confusing for readers if the investigators refer to these analyses as "6-month analyses" or "12-month analyses", for example, to distinguish them from interim analyses conducted in a clinical trial context, which would be accompanied by specific stopping rules, etc.
RESPONSE: We thank the reviewer for pointing out this somewhat confusing notation and we have adopted instead of the notation suggested by the reviewer. Further we have clarified what we mean with each analysis timepoint. We also explicitly state that there will be descriptive analyses performed once baseline data is available on all participants. As part of the development of the statistical analysis plan, we have also limited the scope of the 6-month analysis considerably and this is now reflected in the updated analysis section of the manuscript.

Will missing data be imputed?
RESPONSE: This is included in the statistical analysis plan we have developed but was left out of the brief summary for this protocol paper. We expect a sizeable number of drop-outs and will potentially also have other missing values in the data. For the primary analyses we will use multiple imputation (valid under missing-at-random) as this can be used across all analyses we are planning (unadjusted group comparisons, GLMs, LMMs and GEEs) but plan to perform all-available-cases analyses (requiring the stronger assumption of missing-completely-at-random), direct likelihood and fully Bayesian analyses where applicable as sensitivity analyses. Mixed models will anyway be direct likelihood models and we can also use weighted GEE as part of the sensitivity analyses for the GEE models.
We have added a short paragraph to the statistical analysis section of the protocol paper.
5. The investigators may wish to consider the use of pattern mixture models to assess the affect of differential dropout by HIV status, reactivation of herpes viral infection and HIV viral antigenemia on the estimated effects of these characteristics on risk of cerebrovascular and cardiovascular disease.
RESPONSE: We thank the reviewer for this suggestion. While we will focus on the selection modelling framework, we do hope to use pattern mixture models to gain further insights. However, this will only be feasible if there are sufficiently few distinct missingness patterns. Most of the missing data should be due to drop-out, which should limit the number of missing data patterns. This is also mentioned in the paragraph on missing data we have added to the manuscript.
6. Will the exclusion of study participants who die from analyses depend on whether study participants die from CBD or CVD?
RESPONSE: We will conduct several types of analysis of these complex longitudinal data. We will use GEE models to examine progression of surrogate markers over time. These models can accommodate the occurrence of missing data at a given time point. For example, if a participant died after the 9-month time point, all their data up to 9 months would be included in the model, regardless of their cause of death.
7. In the study design section, it is indicated that the study schedule is described in " Figure 2". Please correct this to " Table 2". RESPONSE: This has been amended, many thanks for highlighting.
8. It is indicated that enrollment for the study started in May 2017 and will continue until August 2020. Will this be sufficient time to enroll 990 participants and follow them for 36 months each?
RESPONSE: The minimum follow-up requirement is 24 months which all participant will achieve.
Reviewer: 2 Reviewer Name: ISAAC SSINABULYA Institution and Country: Makerere University College of Health Science, Uganda