Independent report

TCRP modelling group findings: risk of severe COVID-19 outcomes

Published 31 March 2023

Applies to England

Background

In winter 2022, the COVID-19 Antivirals and Therapeutics Taskforce was planning for the potential wider deployment of antivirals against COVID-19, should there be new evidence of the drug’s effectiveness in a wider cohort. As part of this work the Therapeutics Clinical Review Panel (TCRP) was required to consider whether there are further high-risk patient groups that could benefit from antiviral treatment - in particular, whether there are groups that have a risk that is at least as high as those that are already eligible for treatment.

The groups that are currently being targeted include patients with:

• chromosomal disorders affecting the immune system, including Down’s syndrome
• certain types of cancer or who have received treatment for certain types of cancer
• sickle cell disease
• certain conditions affecting their blood
• chronic kidney disease (CKD) stage 4 or 5
• severe liver disease
• an organ transplant
• certain autoimmune or inflammatory conditions (such as rheumatoid arthritis or inflammatory bowel disease)
• HIV or AIDS who have a weakened immune system
• inherited or acquired conditions affecting their immune system
• rare neurological conditions: multiple sclerosis, motor neurone disease, Huntington’s disease or myasthenia gravis

This note is intended as a guide to whether there are other groups in the population who have an equal or higher risk of severe COVID-19 outcomes than those groups listed above.

The methods used are crude and any groups identified through this process would require closer scrutiny to better understand their risk and to what extent this might be modified by improved access to antivirals and therapeutics.

Methods

The modelling group met to consider estimates of the risk of severe disease (hospitalisation and/or death) in different patient groups, using data from recent, large, representative UK studies. Recent in this sense meant studies performed after the widespread use of immunisation in the UK. Three studies were identified:

1. A retrospective cohort study of linked primary care and death data by the OpenSAFELY consortium.^{[footnote 1]} This study considered the first 3 waves of COVID-19 in the UK, though the last of these (28 May to 14 December 2021) which was primarily caused by the Delta virus is considered here (n=18.7 million adults).

2. Agrawal and others (2022)^{[footnote 2]} undertook a prospective cohort study constructed via electronic linkage of primary care, PCR testing, vaccination, hospitalisation and mortality data in the UK (n=30 million). The study was conducted over the period 20 December 2021 to 28 February 2022 encompassing the first wave of the Omicron variant.

3. Hippisley-Cox and others (2022)^{[footnote 3]} constructed a cohort of adults in England using the QResearch database. They were linked to individual-level national data on vaccination, high-risk patients prioritised for COVID-19 therapeutics, SARS-CoV testing results, hospitalisation data, cancer registries, cancer treatment and radiotherapy records as well as national deaths registries. There were 1.3 million patients with a positive SARS-CoV-2 test that occurred between 11 December 2021 and 31 March 2022. These would have been mostly infected by the Omicron variant.

To determine a threshold for risk we assessed the lowest high-risk condition from those given by the TCRP (above) then compared the risks of severe COVID-19 disease associated with other comorbid conditions from these cohort studies to see whether there were other groups with risks that were at least as high.

Results

Detailed results are given in the respective papers.

Relative risks

Those conditions that appear to have support across at least 2 studies indicating that the risk of severe disease in this group is at least as high as in the groups already identified by the TCRP are marked either side with an asterisk (*).

All 3 studies provide a measure of the relative risks of severe COVID-19 by risk groups. There are, however, significant differences between the studies, in:

• outcomes (COVID-19 deaths in the OpenSAFELY consortium)
• severe disease (COVID-19 related hospital admission or death in the case of Agrawal and others (2022))
• hospitalisation and death (separately in Hippisley-Cox and others (2022))
• timing and therefore dominant strain (Delta in OpenSAFELY consortium, Omicron in Agrawal and others (2022) and Hippisley-Cox and others (2022))
• variables adjusted for in analysis, including vaccination status (which was not adjusted for in OpenSAFELY consortium)

Unsurprisingly, there are differences in the results of the papers with, for instance, the OpenSAFELY consortium reporting significant increases in risk of death from COVID-19 in Asian and black ethnic groups persisting into the third (Delta) wave of COVID-19, but Agrawal and others finding no such elevated risks of severe disease in these ethnic groups during the fourth (Omicron) wave.

Hippisley-Cox and others also find no elevated risk of death by ethic group during the Omicron wave, though some increases in the risk of hospitalisation, particularly in Pakistani and Bangladeshi ethnic groups.

This example serves to illustrate some of the differences between the studies. There are also, however, clear similarities in the results of the 3 papers, with all 3 suggesting that the risk of severe disease or death from COVID-19 is strongly related to *increasing age* and *obesity* as well as *male gender*.

Figures 1a to 4b show the adjusted rate ratio and/or relative hazard of severe COVID-19 and/or death from these 3 studies where the comparison group is not having the condition if the reference group is not given.

The lowest high-risk condition from within the TCRP groups appears to be certain autoimmune or inflammatory conditions (such as rheumatoid arthritis or inflammatory bowel disease). This category of patients roughly coincides with rheumatoid arthritis, lupus, psoriasis (OpenSAFELY) and rheumatoid arthritis or systematic lupus erythematosus (SLE) in both Agrawal and others and Hippisley-Cox and others although the latter study assesses the risk of severe COVID-19 in patients with inflammatory bowel disease separately and finds this to be lower than in rheumatoid arthritis patients (see Figures 3a, 3b, 4a and 4b).

In all 3 studies these rheumatoid arthritis-related groups are at moderately increased risk of severe disease or death. To aid comparison with the other risk categories the estimated 95% confidence interval (CI) on the adjusted rate and/or hazard ratio for severe COVID-19 in these patient groups are highlighted in Figures 1a to 4b with a blue shaded bar. This therefore indicates the lower end of the elevated risk categories given by the TCRP.

Of conditions not listed by the TCRP as high risk, the OpenSAFELY consortium suggests that in wave 3, the hazard ratio of death from COVID-19 was higher for patients who:

• were obese: 35 to 39.9kg per m² (2.25 (95% CI 2.06 to 2.47)); 40 plus kg per m² (4.77 (95% CI 4.34 to 5.25))
• had chronic kidney disease stage 3b (2.56 (95% CI 2.35 to 2.78))
• had dementia (2.81 (95% CI 2.46 to 3.23)), other neurological disease (2.3 (95% CI 2.03 to 2.59)), learning difficulties (3.95 (95% CIs 3.09 to 5.05)), and severe mental illness (2.36 (95% CI 2.02 to 2.74))

than it was for patients with rheumatoid arthritis, lupus or psoriasis (1.72 (95% CI 1.59 to 1.85)).

There is support from the other studies that patients with *dementia* or *obesity* are at least as high a risk as the rheumatoid arthritis containing group. There is no corresponding support from the other studies for the other higher-risk groups identified here, though it should be mentioned that the other 2 studies do not subdivide CKD stage 3 into stage 3b and 3a.

Analysis of the results of Agrawal and others suggests that of those conditions not explicitly listed by the TCRP, pulmonary hypertension and dementia appear to have higher relative risks of severe disease than rheumatoid arthritis or SLE patients. Specifically, in those who had received a booster dose of an mRNA derived vaccine, the estimated adjusted rate ratio for severe COVID-19 for those with dementia was 2.65 (95% CI 2.54 to 2.77) and 2.84 (95% CI 2.63 to 3.07) for patients with pulmonary hypertension, compared with a rate ratio of 2.32 (95% CI 2.20 to 2.45) for those with rheumatoid arthritis or SLE (all comparisons are to those without the condition and after adjusting for non-clinical factors).

As mentioned above the other studies also identify *dementia* patients as being at significantly higher risk. There is a weak signal from the other studies that pulmonary hypertension may be associated with elevated risk with Hippisley-Cox and others identifying that women with pulmonary hypertension and/or fibrosis may be at a similar or higher risk of hospitalisation and death than women with rheumatoid arthritis or SLE (Figures 3a, 3b, 4a and 4b). This association does not appear to hold for men. Although the OpenSAFELY study does not look at pulmonary hypertension specifically, patients with high blood pressure or diagnosed hypertension were not found to be at higher risk than rheumatoid arthritis, lupus or psoriasis patients in the OpenSAFELY analysis (Figures 1a and 1b).

Hippisley-Cox and others suggest that both male and female patients with *type 1 diabetes* are significantly more likely to be hospitalised or die following SARS-CoV-2 infection than those who have rheumatoid arthritis or SLE (Figures 3a, 3b, 4a and 4b). There is some support from the other studies that diabetics may be at increased risk, but it is not consistent. Agrawal and others suggests that type 1 diabetes patients are at a similar risk to those with rheumatoid arthritis (Figure 2).

The OpenSAFELY consortium does not distinguish between types 1 and 2 diabetes but does find that all groups of *diabetes* (controlled, uncontrolled, without recent Hb1ac measure) were at similar or elevated risk of severe disease in the third wave compared with the reference group (those with rheumatoid arthritis, lupus or psoriasis).

The results of the Hippisley-Cox and others study also suggests that patients taking oral steroids are at significantly increased risk of death or hospitalisation than those with rheumatoid arthritis or SLE (Figures 3a, 3b, 4a and 4b). Agrawal and others do not find this group to be at greater risk (Figure 2), though the OpenSAFELY consortium found that asthma patients on *oral steroids* had a similar risk of death during the third wave as the rheumatoid arthritis containing group (Figures 1a and 1b). There is therefore weak support from these studies that patient on oral steroids may be at increased risk.

Hippisley-Cox and others also find some inconsistent patterns by gender and by outcome that may require further investigation. For instance, they find that women with congestive heart failure or thromboembolism appear to be at significantly greater risk of death (Figures 3a and 3b) but not hospitalisation (Figures 4a and 4b) than women with rheumatoid arthritis or SLE, whereas men with these comorbid conditions do not appear to have an elevated risk of either hospitalisation or death compared with those with rheumatoid arthritis or SLE. Agrawal and others lend some support to these observations and find that patients with *heart failure* are at a similar risk of severe COVID-19 as patients with the rheumatoid arthritis or SLE (Figure 2).

Absolute risk of severe disease

The above analysis looks at risk of severe COVID-19 given the presence or absence of individual clinical risk factors. However, to put the risk of these conditions in context and assess whether there are other groups in the population (for example, age groups) who may be at higher risk, it is necessary to compare the absolute risk of severe disease or death across all these different groups. The OpenSAFELY consortium does just this, the results of which are presented in Figure 5, where for ease of comparison the estimated 95% CI of the absolute risk of COVID-19 death associated with rheumatoid arthritis, lupus or psoriasis is shown as a shaded blue bar.

The figure demonstrates that during the third (Delta) wave, the absolute risk of death from COVID-19 from individuals who were over 70 years of age without other comorbidities was significantly higher than that of a typical (age standardised) patient with rheumatoid arthritis, lupus or psoriasis, namely:

• 1.48 (95% CIs 1.41 to 1.55) per 1,000 person-years for individuals aged 70 to 79 years
• 4.54 per 1,000 person-years (95% CI 4.37 to 4.7) for 80 plus years of age
• 1.04 per 1,000 person-years (95% CI 0.96 to 1.12) for rheumatoid arthritis, lupus or psoriasis patients

Note also that many of these risk factors interact so that, for example, a patient in their 20s who has diabetes may be at elevated risk compared to others in their age group without diabetes, but their risk is still low. To take these interactions into account requires an individual risk-prediction tool. Hippisley-Cox and others provide such a tool.

Discussion

The risk groups as defined by the TCRP and the 3 studies are not identical. In addition, there are methodological, outcomes and clinical coding differences between the studies and although all 3 are large and UK-based the timing of the studies differ and the variables controlled for in the analysis differ too. There are also likely some differences in the patient populations. Some variation in results would therefore be expected, as well as some difficulties in reading across to the high-risk groups defined by the TCRP. These difficulties are illustrated with the findings related to diabetes, in which all 3 studies point to an increase in risk for some categories of diabetic patients, but as the subcategories differ between the studies it is difficult to clearly ascertain which of these groups are at greatest risk and why.

The findings presented here should be regarded as a crude indicator of which patient groups warrant further, in depth assessments of their risk of severe COVID-19 disease.

It is also clear that there have been changes in risk over the course of the pandemic OpenSAFELY consortium. Further changes may have occurred since these studies concluded (in late 2021 or early 2022) or may occur into the future. Indeed, the OpenSAFELY analysis has recently been updated to include data from the Omicron wave, but as these results are not yet in the public domain, we restrict the analyses here to the latest publicly available data.

All 3 studies suggest that there are some clinical risk groups who may have similar or greater relative risk of severe disease than those listed by the TCRP (a higher risk than that associated with rheumatoid arthritis and similar autoimmune conditions). The patient groups with support from at least 2 independent analyses include:

• those who are obese
• certain classes of diabetes patients
• patients with dementia
• possibly, those with heart failure
• possibly, those who are on oral steroids

There is also a longer list of groups of patients which were identified as being high risk in single studies only. All of these categories of patients should be investigated further to ascertain:

• whether they do indeed have an elevated risk
• whether there are subgroups within them who have higher or lower risks than the average
• what may be the underlying mechanism driving these changes in risk
• to what extent this may be ameliorated by improved access to novel COVID-19 antivirals or therapeutics

Apart from clinical risk factors, the studies also highlight the continuing importance of other demographic and social differences in risk, with increasing age in particular persisting as a major independent risk factor for severe COVID-19 outcomes.

Figure 1a: OpenSAFELY hazard ratio of death by demographic and socio-economic variables by COVID-19 wave

Note: adapted from OpenSAFELY consortium.

Figure 1b: OpenSAFELY hazard ratio of death by clinical conditions by COVID-19 wave. The 95% CI on the hazard for rheumatoid arthritis, lupus or psoriasis in the third wave is highlighted with a blue shaded bar

Note: adapted from OpenSAFELY consortium.

Figure 2: adjusted rate ratios (95% CIs) for specific clinical risk factors associated with COVID-19 hospitalisation or death among individuals who received booster doses MRNA-1273 or BNT162b2

Note: the 95% CI for the risk ratio associated with rheumatoid arthritis or SLE is highlighted with a blue shaded bar. Adapted from Agrawal and others (2022).

Figure 3a: adjusted hazard ratio of deaths by clinical risk groups for males (a). Adjustment for all variables including age and BMI (body mass index)

Note: adapted from Hippisley-Cox and others (2022).

Figure 3b: adjusted hazard ratio of deaths by clinical risk groups for females (b). Adjustment for all variables including age and BMI

Note: adapted from Hippisley-Cox and others (2022).

Figure 4a: adjusted hazard ratio of hospitalisation by clinical risk groups for males (a). Adjustment for all variables including age and BMI

Note: adapted from Hippisley-Cox and others (2022).

Figure 4b: adjusted hazard ratio of hospitalisation by clinical risk groups for females (b). Adjustment for all variables including age and BMI

Note: adapted from Hippisley-Cox and others (2022).

Figure 5: sex and age-standardised COVID-19-related death rates (IR) and 95% CIs per 1,000 person-years in OpenSAFELY in the 3 pandemic waves

Note: models were standardised for age and sex using the European standard population except for the death rates by age group (not standardised) and death rates by sex (standardised by age). The 95% CI for the risk of COVID-19 death from rheumatoid arthritis, lupus or psoriasis in the third wave is highlighted with a blue shaded bar. Adapted from OpenSAFELY consortium.

Report authors

• John Edmunds (London School of Hygiene and Tropical Medicine)
• Aziz Sheikh (University of Edinburgh)
• Ben Goldacre (University of Oxford)
• Julia Hippisley-Cox (University of Oxford)
• Laurie Tomlinson (London School of Hygiene and Tropical Medicine)
• Elizabeth Williamson (London School of Hygiene and Tropical Medicine)

References

1. The OpenSAFELY Collaborative: Changes in COVID-19-related mortality across key demographic and clinical subgroups: an observational cohort study using the OpenSAFELY platform on 18 million adults in England. MedRXiv (2022). ↩

2. Agrawal U, and others: Severe COVID-19 outcomes after full vaccination of primary schedule and initial boosters: pooled analysis of national prospective cohort studies of 30 million individuals in England, Northern Ireland, Scotland, and Wales (2022), Lancet 400(10260):1305-20. ↩

3. Hippisley-Cox J, Khunti K, Sheikh A, Nguyen-Van-Tam JS, Coupland CAC: QCovid 4 - Predicting risk of death or hospitalisation from COVID-19 in adults testing positive for SARSCoV- 2 infection during the Omicron wave in England. MedRXiv (2022). ↩