Two UK studies find that HIV infection may be a risk factor for dying from COVID-19

People with HIV without co-morbidities were not at increased risk
Image: engin akyurt/Unsplash. Image is for illustrative purposes only.

Two studies relating COVID-19 mortality in the UK to HIV status have both concluded that having HIV raises the risk of dying from COVID-19, after adjusting for age and some other factors.

The first (Bhaskaran) is a population survey of mortality risks, which relates death from COVID-19, as listed on death certificates, to HIV status recorded in National Health Service (NHS) primary care records.

The other (Geretti) is a prospective cohort study of mortality in patients who have been hospitalised due to COVID-19 and compares mortality in patients with and without HIV.

Glossary

comorbidity

The presence of one or more additional health conditions at the same time as a primary condition (such as HIV).

morbidity

Illness.

statistical significance

Statistical tests are used to judge whether the results of a study could be due to chance and would not be confirmed if the study was repeated. If result is probably not due to chance, the results are ‘statistically significant’. 

peer review

The process of subjecting a scientist’s research to the scrutiny of other scientists working in the same field. Studies published in medical journals are usually peer reviewed, whereas conference presentations are not.

diabetes

A group of diseases characterized by high levels of blood sugar (glucose). Type 1 diabetes occurs when the body fails to produce insulin, which is a hormone that regulates blood sugar. Type 2 diabetes occurs when the body either does not produce enough insulin or does not use insulin normally (insulin resistance). Common symptoms of diabetes include frequent urination, unusual thirst and extreme hunger. Some antiretroviral drugs may increase the risk of type 2 diabetes.

The first study finds that, since 1 February and up to 22 June, people with HIV had a 130% raised risk (i.e. 2.3 times the risk) of dying from COVID-19 compared with the general population (a risk similar to that seen in a recent study in South Africa’s Western Cape province, presented at AIDS 2020: Virtual in July).

The second study finds a 63% raised risk of dying of COVID-19 among the HIV-positive members in its database of hospitalised patients, once age and state of health on admission are taken into account. The database consists of the UK patient data from ISARIC, an international research consortium. It covers a slightly different time frame to the first study; from 18 January, when COVID-19 PCR testing first became available to UK hospitals, to 18 June.

Both studies face the problem that in the UK, where HIV is a lot rarer than in South Africa, they are working with small numbers of COVID-19 deaths in people with HIV.  The first study found only 25 people whose HIV status was recorded in GP records and whose death certificates recorded death from COVID-19. The second study found 115 people hospitalised for COVID-19 who were recorded as having HIV, and virtually the same number of deaths, 26. Many but not all of these will be the same people. These small numbers make it very hard to show that their results are statistically significant and not just due to chance, and can lead to different results.

For instance, the first study found that the raised risk of death in people with HIV seemed even larger in people who self-described as Black, as opposed to any other ethnicity; the second did not find this association.

Interestingly, the first study found that, although this was not statistically significant owing to the low numbers, the higher risk of death was most evident during the first 60 days of the pandemic. The authors speculate that this may reflect less social distancing and/or greater vulnerability to infection during the early weeks, before people with HIV were advised to shield. Beyond day 90 of the pandemic (from 2 May) the increased risk due to HIV was no longer apparent.

The findings of both studies have been published as pre-print articles, which means they have not yet been peer-reviewed. In a statement issued in response to the studies, the British HIV Association (BHIVA) and allied organisations urge that the findings should be interpreted with caution, especially as due to limited figures and under-recording, the influence of other risk factors for COVID-19 mortality could be under-estimated.

The OpenSAFELY study

The first study comes from OpenSAFELY, a national database established in reaction to COVID-19, which uses electronic medical records data from 17.2 million NHS England patients, and national death registrations, to calculate the risk of COVID-19 according to various risk factors. They have previously published an analysis of other risk factors and confirmed that old age was by far the strongest one; people over 80 were at least 20 times more likely to die from COVID-19 than people aged 50-59.

The OpenSAFELY dataset included 27,480 adults living with HIV and 17,255,425 without HIV. This is about 30% of the 55-million English population, and the same proportion of the approximately 90,600 people living with HIV.

People living with HIV were more likely to be male and to be Black than the general population. Thirty-one per cent of people with HIV lived in neighbourhoods classified as 'most deprived' compared to 19% of the general population.

Although the median age of patients in this vast database was similar (48 years), around 60% of people living with HIV were clustered in the 40-59 age band compared to 30% of the general population. They were less likely to be aged 60 or over.

There was no difference between people living with HIV and the general population in the prevalence of obesity or any co-morbidity except for chronic liver disease and hepatitis C infection.

The study logged 14,882 deaths where COVID-19 appeared on the death certificate up to 22 June 2020. Twenty-five of these deaths were of people living with HIV and 14,857 in people without HIV. The cumulative mortality rate was 0.087% in people living with HIV (one death in every 1149 people) and 0.038% in people without HIV (one death in every 2639 people).

People living with HIV were 2.9 times more likely to die after adjusting for age and sex, to account for them being more likely to be male and younger. After adjusting for ethnicity and the index of deprivation, the risk diminished slightly and further diminished when obesity, smoking and co-morbidities were taken into account. Nevertheless, even after adjusting for all the known risk factors, people living with HIV were 2.3 times more likely to die of COVID-19 than people without HIV.

Black people living with HIV were at highest risk of dying from COVID-19. They were 3.8 times more likely to die than Black people without HIV. In contrast, White, South Asian and other ethnic groups living with HIV had a 1.64-fold increased risk of death compared with HIV-negative people of the same ethnic groups.

People with HIV without co-morbidities were not at increased risk of death. But the presence of at least one co-morbidity raised the risk of death. Twenty-three of the 25 people with HIV who died had at least one co-morbidity, most commonly hypertension or diabetes. This does however mean that because there were only two deaths in people without co-morbidities, it is hard to say whether they were at increased risk of death compared with HIV-negative people who also did not have co-morbidities.

This study was not able to look at the impact of viral suppression on COVID-19 death risk, because it drew data from general practitioners, not HIV clinic records. But the investigators point out that 94% of people with HIV in the UK are on antiretroviral treatment and 97% have suppressed viral load, so the increased risk associated with HIV is highly unlikely to be explained by unsuppressed viral load.

The risk of death in people living with HIV was highest during the first 60 days of the pandemic (from 1 February to 2 April, largely before lockdown). During this period people with HIV were four times more likely to die than HIV-negative people. After this, the difference due to HIV became much smaller and beyond day 90 (after 2 May), people living with HIV no longer had an increased risk of death.

The study investigators suggest that the higher death rate in people with HIV during the first 60 days could reflect a higher infection rate at the beginning of the pandemic, with the subsequent decline in deaths explained by strong adherence to social distancing, and shielding by the most vulnerable.

But they say that this hypothesis needs to be tested further. It could be that more people with HIV were working in jobs with more exposure to COVID-19, or because London, where more people with HIV live, experienced an earlier peak of cases and deaths than other parts of the country.

It’s important to note that this study can’t show that people with HIV who contract SARS-CoV-2 are more likely to die than others. This is because it can’t show whether people with HIV are more likely to catch the virus; it might be that an increased risk of catching the virus may contribute to the increased risk of death.

The ISARIC Study

The second study also cannot say whether people with HIV are more likely to catch SARS-CoV-2, or become ill if they do.

But it does suggest that once the important factor of age is controlled for, a higher proportion of people with HIV, if sufficiently symptomatic to be hospitalised, may die from COVID-19 infection. It found that whereas a higher proportion of HIV-negative people hospitalised for COVID-19 died, this was because they were older; the risk was reversed in younger people, with more people who had HIV and hospitalised for COVID dying than people without HIV, if they were under 60.

The data were taken from ISARIC (the International Severe Acute Respiratory and emerging Infections Consortium), an international database of people hospitalised due to COVID-19. Findings from ISARIC have been the subject of other studies, including one of COVID-19 and ethnicity in the UK.

ISARIC analysed characteristics and outcomes of 47,539 patients hospitalised with COVID-19, of whom 115 (about 1 in 400) had HIV.

Of the people with HIV, nearly 90% were on record as being on antiretroviral therapy (ART), which is slightly fewer than OpenSAFELY study’s assumption that 94% were on ART and virally controlled. In fact, if 97% of those on ART were virally controlled, as UK figures suggest, then this suggests that 84% of the ISARIC HIV-positive patients were virally controlled.

The most striking difference between the people with HIV and those without is age: the HIV-positive group had a median age of 55 and the HIV-negative a median age of 74; 75% of the HIV-negative group were aged over 60, compared with 30% of the HIV-positive one.

Although there was no sign that HIV-positive people were being admitted earlier or later than other patients, they had longer duration of symptoms on average (five days versus three) and were more likely to have systemic symptoms such as fever and raised heart rate.

Mortality was measured as the proportion of people who died 28 days or less after hospital admission. By this time half of the entire patient group had been discharged alive, but 29% had died (the rest remained in hospital, had transferred to other care, or their outcomes were unknown).

Mortality, unadjusted for other factors, was higher in the HIV-negative people: mortality by 28 days was 32% in them and 25% in the HIV-positive patients, a 26% lower rate. After adjusting for sex and ethnicity this hardly changed, with HIV-positive people 23% less likely to die – a contrasting finding from the OpenSAFELY study. 

However, age made a big difference. Because the HIV-negative group was on average so much older, once age was controlled for, it meant that the HIV-positive group became 39% more likely to die, relative to HIV-negative persons of the same age. Controlling for date of admission and time of symptom acquisition increased this risk difference, and controlling again for co-morbidities increased this raised risk further, to a 63% higher risk of dying if you had HIV once other characteristics were excluded.

When the HIV-positive people who died were compared to the ones who survived, then the factors that were statistically significantly related to dying were: badly controlled diabetes with complications; obesity (33% of the HIV-positive patients who died were obese compared with 14% of the ones who did not); signs of respiratory distress on admission; and signs of inflammatory disease such as high white blood cell and C-reactive protein counts.

The positive people who died were older than the ones who survived but only slightly so, and this was not quite statistically significant (age 58 versus 54.5, p = 0.07).

Implications and reaction

The two studies taken together suggest that HIV infection is to some extent associated with a raised risk of dying from COVID-19 symptoms if you are infected and get symptoms. They only look at one outcome: death. It’s important to note that neither study tells us whether people with HIV are at higher risk than others of catching SARS-CoV-2, of developing symptoms, or of needing hospitalisation if they do.

It is also important to understand how these two studies, conducted on what must be overlapping groups of patients, produce different results. While both agree that HIV increases the risk of COVID-19 death, the estimate of how much it contributes is more than twice as great in OpenSAFELY as it is in ISARIC (132% increase in risk compared with 63%). Why the second study finds no additional increase in risk for people with HIV who are Black also needs investigation. And the finding from OpenSAFELY that the relative mortality risk of having HIV appeared to decline over time is also important to explain.

The British HIV Association released a statement, signed also by the National AIDS Trust, Terrence Higgins Trust and NAM aidsmap urging caution towards the findings. The statement notes that, as pre-prints that have not been peer-reviewed, both studies are “new medical research that has yet to be evaluated and so should not be used to guide clinical practice.”

They comment in particular that the OpenSAFELY study was only able to look at the records of 27,500 of the 88,000 people with HIV in the UK, and that people who do not have HIV recorded in their primary care records may differ from ones who do.

They also question whether assuming, for instance, that if smoking status and body weight were not recorded, it should be assumed that people were not smokers or obese. And they note that neither study had information on CD4 count or HIV viral suppression, and that this could fail to detect whether poorly-controlled HIV infection had a part to play.

In addition, as with all COVID-19 outcome studies, results could be biased if there are different testing rates. If people with HIV were more likely to test for SARS-CoV2, this could overestimate the proportion of HIV-positive people who died of illness ascribed to it.

The BHIVA statement also notes the importance of emphasising that very few people who did not have co-morbidities died from COVID-19.

And they add: “We encourage groups producing data with potentially significant conclusions to work with HIV societies and community organisations to develop appropriate messaging for people providing HIV care and, most importantly, for people living with HIV.”

Nonetheless, the OpenSAFELY investigators conclude: “Our findings suggest that people living with HIV may be a higher-risk group for COVID-19 death...and may also need priority consideration if and when a vaccine against SARS-CoV2 becomes available".

References

Bhaskaran K et al. HIV infection and COVID-19 death: population-based cohort analysis of UK primary care data and linked national death registrations within the OpenSAFELY platform. Pre-print, see https://www.medrxiv.org/content/10.1101/2020.08.07.20169490v1.

Geretti AM et al. Outcomes of COVID-19 related hospitalisation among people with HIV in the ISARIC WHO clinical characterisation protocol UK protocol: prospective observational study. Pre-print, see https://www.medrxiv.org/content/10.1101/2020.08.07.20170449v1.