In this paper we summarise the extant literature on the interaction of HIV with COVID-19. We used an analytic model separating risk of infection from risk of severe disease. In terms of risk of infection, the definitive answer would come from asymptomatic screening, but HIV status is unlikely to be collected as part of such exercises. Hence, diagnosis rates are the next-best surrogate marker, even though they are profoundly influenced by the occurrence of symptoms or hospitalization. We identified 27 studies to date which address infection risk as a function of HIV status. Most studies assessed HIV prevalence in COVID-19 diagnosed individuals, although some followed COVID-19 incidence in HIV clinic cohorts. (7, 13-17) The consensus that has emerged provides reassurance that PLWH appear no more likely to become infected than their non-infected counterparts; in our model, βHIV≈β0. The caveat to this conclusion is that PLWH may have been shielding or otherwise reducing their exposure (reducing the factor, α) during the pandemic, in which case, true infection risk may be higher than estimated. In terms of severity, early studies were reassuring. PLWH did not appear over-represented among those experiencing complications with COVID-19. However, it is now apparent that, when adjusted for other risk factors, even well-controlled HIV infection does appear associated with increased mortality risk. Indeed, much of the disparity between the findings of different studies can be accounted for by the way in which authors have accounted (or not) for co-factors and confounders in their models. Almost all studies that have considered comorbidities have found that the same factors operate within HIV-positive COVID-19-infected cohorts as have been identified in the general population (age, cardiovascular disease, obesity etc), as might be expected. Taking co-factors into account is not possible in small studies, but larger studies allow adjustment which may be crucial for interpretation. So, for example, in Geretti et al (21) and Bhaskaran et al, (20) unadjusted mortality rates in PLWH are no higher than background mortality rates. However, in both studies, once data are adjusted for age etc, the independent adverse impact of HIV infection is evident. This increased mortality risk with HIV infection manifests itself in different parts of the patient pathway in different settings. Thus in the South African study cited above, the increased mortality risk was most apparent in non-hospitalised individuals; (19) in the NYC study, the big difference was in diagnosis and hospitalisation rates - once admitted mortality was similar to non-HIV-infected individuals; (12) in the UK study, in which hospitalisation was an entry criterion, mortality rates were found to be significantly higher in PLWH after admission. (21) It does therefore appear reasonable to take HIV status into account in developing policies and practice, but do we treat all PLWH the same? What about the impact of CD4 count and viral load? Here, the data is far less clear-cut. We can postulate with some confidence that the additional risk of HIV infection for mortality seems to apply even to those with well-controlled HIV. Several studies observe an independent effect of HIV infection across all strata; for example Tesoriero et al found an increased hospitalisation risk in individuals with CD4≥500 and viral suppression. (12) Furthermore, the two UK studies discussed above were performed in settings where 94% of subjects would be expected to be well-controlled. Whether those with low CD4 counts and/or viraemia are at higher risk remains unclear. Several studies show a relationship between lower CD4 counts and worse outcomes, (12-13,19), but some studies suggest low CD4 counts are protective. (7, 22) Although it might appear counter-intuitive that “immunosuppression” would worsen “immunopathology”, it should be remembered that advanced HIV is characterised by immune dysfunction with heightened levels of immune activation as much as “immunosuppression”. Recommendations based upon CD4 and/or viral load status thus appear to have a weak evidence-base, but may be justified in a pragmatic risk-stratified model. Current UK guidelines differentiate those with a CD4 count <50 cells/uL for special consideration, (3) but there is little evidence supporting where to make this cut-off. In the NYC study, the authors separated out a cohort with a CD4 count <200 cells/uL or <14% (“Stage 3”) and found they had a higher HIV-attributable mortality risk, (12) but they did not explore lower cut-offs. At the other end of the scale, it does not appear possible to define a “safe” level of CD4 count from the studies published to date; indeed, one can argue from the mortality impact of HIV even in settings where almost everyone is on effective treatment, (20-21) that no such “safe” level exists. One might however pragmatically define a level where recommendations for PLWH match those for the general population. Such stratified advice is likely to vary on a regional basis. Currently, the countries with most COVID-19 cases are relatively low HIV seroprevalence countries. Conversely, countries with high HIV seroprevalence may have limited COVID-19 testing resources and incidence rates may be underestimated. This is why studies such as Boulle et al (19) are particularly valuable as they address COVID-19 risks in a high HIV burden setting (as well as making intriguing observations about TB/COVID interactions). Although their conclusions may not be generalisable to other settings, the consistency in message, if not magnitude, of the effect of HIV on COVID-19 mortality across all the larger studies cited is notable. This overview has attempted to be comprehensive, but we may have missed publications and have not captured those in languages other than English. Of course, in such a rapidly-changing situation new studies are being published every month but this study summarises the current state-of-play and constitutes a comprehensive basis for future literature reviews. We have deliberately chosen not to focus on whether some antiretroviral drugs impact COVID-19 outcomes directly as the data are conflicting and such questions are best addressed in prospective clinical trials. We considered a meta-analysis, but after reviewing the disparate nature of studies with their different data collection strategies, and estimating the number of studies that would need to be excluded, felt that a comprehensive descriptive analysis was more helpful. Understanding the impact of HIV on COVID-19 is important in order to inform guidance and policy. “Shielding” has costs in terms of emotional and psychological wellbeing, finances and quality of life. If PLWH are to be asked to take extra precautions, this guidance needs to be based on good data. Additionally, as vaccines become available, prioritisation for vaccination is likely, in which case we need data to establish where PLWH stand in any risk-stratification. Despite an extensive body of literature, more research is needed to clarify risk in different settings and particularly to clarify how guidance should be shaped by measures such as CD4 count and viral load, but the new data from large studies is consistent: even well-controlled HIV appears associated with an increased COVID-19 mortality risk.