What do the UK data tell us about the Delta variant? Are the vaccines working?

Summary points:

• The Delta variant has taken over the UK

• The cases are predominantly young people, especially unvaccinated ones.

• Nearly all older adults in the UK have been vaccinated, but few young people.

• Rigorous analyses of the rich UK data demonstrate vaccines retain high efficacy in preventing symptomatic infections as well as hospitalized infections.

• Most hospitalized infected are young, unvaccinated individuals.

• There have been very few COVID-19 deaths from Delta to date.

• The fact that most are from breakthrough infections are driven by the high vaccination rate in the older population and low risk of death in young people.

• Delta is unlikely to produce a massive surge in the USA, but will likely produce local upticks in communities with low rates of vaccination and previous infection.

• Unvaccinated countries will likely experience a more severe Delta surge than the UK.

There is a lot of talk in the media about how the Delta variant is bringing a third wave to many countries in the world. This variant has proven the most transmissible yet, spreading a rate more than double the original SARS-CoV-2 wild type variant, and it is quickly becoming the dominant strain in many places, already comprising nearly all cases in the UK and now comprising near half of the cases in the USA.

From these data, two claims have emerged and are spreading around social media:

1. The vaccines don't work, since >1/2 of COVID-19 deaths in the UK are from those who had been vaccinated.

2. The Delta variant is mild and inconsequential, since the raw hospitalization and death counts are low, that we have no need to be concerned about it at all -- it is as docile as a common cold and the media is just creating hype and alarm by discussing it. Ivor Cummins (whose past claims I've critiqued here) is pushing this argument, calling it the "Delta Deceit"

To investigate these claims, we need to unravel information about cases, hospitalizations and deaths, and consider how they vary by age and vaccination status, and also consider how vaccination rates vary across age groups. We need to aggregate information across all these factors to characterize the Delta variant and vaccine efficacy against it. Fortunately, the UK with its centralized healthcare system and rigorous electronic medical records (EMR) has extensive high-quality medical data on the entire population, and it has also ramped up the most extensive SARS-CoV-2 genetic sequencing efforts in the world, sequencing more than half of confirmed cases. This gives the UK the best variant-specific data in the world. The UK also has an excellent scientific team that is keeping an eye on these data and writing frequent reports of key observations, including technical briefings for variants of concern (VOC) that they share with the public on accruing knowledge on the emerging variants. In this blog post, I will unravel what the UK data tell us about the Delta variant and vaccines, relying heavily on information provided in the UK COV technical briefing #17 written on June 25, 2021, and also integrating in other data and data summaries I have encountered that I have found useful. In summary, the data make it clear that:

• The UK has an extensive sequencing program that enables population-level analysis by variants that is not possible elsewhere.

• Delta has taken over the UK, quickly dominating all other variants.

• Delta infections are happening almost exclusively in the young, with only about 10% in adults >50 years old.

• Nearly all adults >50 years old are fully vaccinated, while only <30% of adults <50 have been vaccinated to date.

• A rigorous study of these data using a case-negative design show that the vaccines are dramatically reducing the number of delta cases that would have been seen had there been no vaccination. Both vaccines have high efficacy vs. symptomatic infection with Delta after two doses, but much lower efficacy after just one dose.

• Unlike in previous waves in which COVID-19 hospitalizations were dominated by older adults, with Delta we see a vast majority of the hospitalized being younger adults, and specifically the unvaccinated younger adults.

• A study of the hospitalization data rigorously shows that the vaccines further reduce the risk of being hospitalized if you are infected, and as a result both vaccines have remarkably high efficacy in preventing hospitalized cases.

• The seemingly paradoxical results of >50% of COVID-19 deaths being among the vaccinated can be explained by the fact that the vaccine is not perfect, most unvaccinated are young, and older people are in general at a much higher risk of death of infection.

• Unvaccinated countries are likely to experience much higher transmission and death rates than what has been experienced in the UK, so this Delta variant should not be taken lightly just because the vaccines have taken the edge off it in the UK.

After presenting these data, I will briefly discuss what this information means we can expect in the USA as the Delta variant takes over this month.

Why is the UK data so useful for studying the effect of variants?

It is difficulty to do population-level studies of specific SARS-CoV-2 variants since a given viral sample must be genetically sequenced to figure out which variant it is. Only a small proportion of cases in the USA are sequenced, but in the UK in 2021 they have ramped up sequencing efforts from <10% of samples earlier in the year to nearly 70% of all samples sequenced by June. This is remarkably useful, and will enable the UK to uncover new variants quickly, and enable population-level analysis by variants that is not possible elsewhere. This is why I think a deep dive to understand the UK data as it pertains to Delta and the vaccines is worthwhile.

The Delta variant has taken over the UK

This near-population level sequencing has enabled the UK to get accurate estimates of the shifting variant proportions in its population.

As can be seen in these data, the Alpha variant (aka B.1.1.7 or UK variant) completely dominated UK cases from the beginning of 2021 until April, when the Delta variant was introduced from India. Since its introduction to the UK, Delta has completely taken over the UK, by June comprising >95% of all new sequenced cases of SARS-CoV-2 infections.

Side note on Beta variant: See also how the Beta variant (aka B.1.351 or South African variant) was also introduced into the UK in April, but did not really take off, and was eventually displaced completely by the Delta variant. Recall the Beta variant is the one showing the strongest immune escape potential to date, with the ability to reduce efficacy from vaccination or previous infection the most, and as I will show later in this blog, Delta really does not demonstrate this type of immune escape potential. This suggests that as of April 2021 in the UK at least, the transmissibility of a variant is much more strongly predictive of its fitness to take over by natural selection than its immune escape potential.

Unique characteristics allow the UK to infer variant status even without sequencing

Some unique circumstances right now make it possible for the UK scientists to accurately determine variant status from a simple PCR test without even having to do genetic sequencing. The Alpha variant has a specific mutation that causes a false negative result from one of the probes, the S-probe, of the SARS-CoV-2 PCR test, while the Delta does not. Given that the genetic sequencing data show that the vast majority of cases this year in the UK have been either the Alpha or Delta variants, this means that if we look at the status of the S-probe on the PCR test, we can reliably determine whether a given case was Alpha (S negative) or Delta (S positive), which they refer to as genotyping. From this, they infer the population level cases as Alpha or Delta for non-sequenced samples to use in the analyses. That's what is meant by "genotyped" in the figure above.

Delta cases are dominated by the young.

Following is a plot of the UK cases by age decades.

As can be seen by this plot, the delta cases are strongly dominated by young people, with only about 10% of cases in those above 50 years of age. This is a stark contrast to earlier in the pandemic. For example, here is a plot of the age distribution of UK cases by age as of July 30, 2020, in which more than 50% of cases where in those over 50 years of age:

The fact that the current Delta surge is being driven by the young age groups can also be clearly seen in this plot of data from Scotland as seen in a tweet by @PaulStenton1

The vaccination rates also strongly vary by age groups in the UK.

Distribution coverage in UK by age groups

Like most places, the UK's vaccine distribution strategy involved first vaccinating the oldest, most vulnerable members of the population before moving the younger ones who were at far less risk of serious disease or death from COVID-19.

The UK government has a vaccine surveillance website in which they track vaccinations over time. Here are the age-specific data for the week 26 report, through June 30, 2021 for dose 1 and dose 2:

From this, we can see that by the end of April (week 17), by which time the Delta variant was well established in the UK, nearly 90% of >70yr population was already fully vaccinated, while the <50 population had <10% even given the first dose, and even now, we have >90% of the >50yr but <30% of the <50yr old population that has been fully vaccinated. Following are the summaries by age group as of July 1, 2021:

This suggest the hypothesis that the Delta cases are dominated by the younger age groups because they are largely unvaccinated, and the older age groups are not getting infected with Delta because they are protected from the vaccines.

This is supported by the raw data, given that not only is the vast majority of Delta cases in the <50yr old group, but in the unvaccinated among that group. Even though >25% of individuals <50yr are fully vaccinated, we have only 5.5% of the <50yr old cases among those with linked vaccine and case specimen data are from fully vaccinated. Also, in the >50yr group, even though >90% of them are fully vaccinated, 57.7% of the cases with linked vaccine and case specimen data are from the small fraction who are not fully vaccinated.

While this data is suggestive that the vaccines are proving effective against the Delta variant, a more rigorous analyses is required to formally test this hypothesis.

Vaccine efficacy vs. Delta variant

The extensive UK data including PCR testing data split out by variant and vaccine status enables the UK scientific team to rigorously test this hypothesis using a test negative design. With this type of design, it is possible to get an unbiased estimate of the relative vaccine efficacy from observational data for a given reference population, in this case individuals reporting symptoms and getting a SARS-CoV-2 PCR test, without requiring a randomized design, assuming uniform testing practices.

As described in this medRXiv paper, currently under peer review for publication, they used this design to estimate the effect of vaccination on risk of a positive test after adjusting for age, ethnicity, history of travel, clinical vulnerability, and week of test by UK region. This enabled them to get valid efficacy measures for each vaccine, dose, and variant while adjusting for the seasonal and regional drift of the pandemic over time, making the results quite scientifically rigorous.

The following table presents the efficacy estimates from this paper, with the number in parentheses the 95% confidence intervals.

This shows that while the vaccine efficacy is slightly lower for Delta than Alpha, the drop-off is very slight, so the vaccines retain most of their efficacy in preventing symptomatic disease for the Delta variant. This shows that Delta does not have strong immune escape against vaccine-based immune protection.

Note how much higher the two-dose efficacy is than single-dose efficacy. This property raises urgency to encourage people who have received one dose to come back for their second dose.

Note also the difference between vaccines. Pfizer retains 88% efficacy vs. symptomatic disease for Delta, while AstraZeneca only shows 67% efficacy. This suggests that the breakthrough infection rate after vaccination is 2.75 times greater for AstraZeneca than it is for Pfizer { (1-0.67)/(1-0.88) = 2.75} . This has implications when projecting what to expect from countries like the USA or Israel who are vaccinating only with the higher efficacy mRNA vaccines.

In spite of the vast data provided by the UK government, it has been exceedingly difficult to get raw numbers on the overall proportion of UK vaccinated residents who were given AstraZeneca vs. Pfizer, much less split out by age. I have not been able to find such numbers anywhere. However, if you look at how much closer the "any vaccine" numbers are to AstraZeneca than Pfizer, this suggests that a very large number of UK residents have gotten AstraZeneca rather than Pfizer.

Nonetheless, these data make it clear that the vaccines maintain near-full efficacy for the Delta variant in protecting against symptomatic disease.

Hospitalizations dominated by younger unvaccinated

The UK report also reported hospitalization data for SARS-CoV-2 infected individuals by age and vaccine status after merging SARS-CoV-2 PCR test, vaccination status, and emergency room admission records, with hospitalization measured by measured by

1. number of ER visits

2. number of serious ER visits (resulting in overnight stay)

And they reported data two ways:

1. Including all with positive PCR tests within 2 weeks of hospitalization

2. Excluding those whose positive PCR test occurred on day of hospital admission.

Here are the data:

We see that unlike previous surges in the pandemic, the vast majority of COVID-related hospitalizations are in young people (<50yr), which is not surprising given that >90% of SARS-CoV-2 cases are occurring in this younger age group.

While these hospitalization counts are relatively low, we are relatively early in the Delta surge in the UK as indicated by the sharply increasing exponential growth in cases. If we look at a plot of new COVID-19 hospital admissions over time from OurWorldInData, we can see that while the absolute numbers are not too high yet, the growth in daily new hospitalizations is steep like previous surges, and well on its way to reaching the high numbers seen in previous surges, in spite of the fact that this surge is being driven by unvaccinated young people with low marginal risk of hospitalization. Thus, any claim that this Delta variant is docile and not of concern is premature.

One key question is whether the vaccines seem to be protecting against hospitalization. To evaluate this, once again we need to look at a rigorous statistical analyses that can test this hypothesis while accounting for inherent differences in risk of hospitalization based on age and clinical risk factors.

Vaccine efficacy vs. hospitalized disease for Delta variant

The UK scientific team has released a technical report looking at vaccine efficacy vs. hospitalization for Alpha and Delta variants by linking all symptomatic SARS-CoV-2 cases between 4/12/21 and 6/4/21 with emergency care data for all hospital admissions, considering any hospitalizations within 14 days of a positive SARS-CoV-2 PCR test, excepting injuries, as a COVID-related hospitalization.

To compute the efficacy of vaccine (VE) they fit two models:

1. Model to assess reduction of risk of symptomatic disease (OR_sympt)

2. Model to assess reduction of risk of hospitalization given symptomatic disease (HR_hosp).

They then computed the vaccine efficacy by the following equation:

VE=100% x {1-(OR_sympt x HR_hosp)}

which provides a legitimate estimate of the effectiveness of vaccine in preventing COVID-related hosptalizations that would have occurred sans vaccination.

The first model (1) to estimate OR_sympt was the logistic regression model described above that adjusted for age, ethnicity, history of travel, clinical vulnerability, and week of test by UK region.

The second model (2) to estimate HR_hosp was a Cox regression model for time to hospitalization by vaccine status adjusting for age, clinical risk, ethnicity, and test week to adjust for potential confounding factors that are known to affect the general risk of hospitalization so the effect of vaccination can be effectively isolated.

The following Table 1 from the UK report shows the results for 1 or 2 doses for each vaccine for both the Alpha and Delta variants:

To interpret this table, let's look at the AstraZeneca/Dose 1/Alpha variant results.

• The OR_symp=0.51 suggests the single dose of AstraZeneca reduces the risk of symptomatic infection by 100% x (1 - 0.51) = 49%

• The HR_hosp=0.48 suggests that the single dose of AstraZeneca reduces the risk of hospitalization for those who have already tested positive by SARS-CoV-2 PCR test by 100% x (1 - 0.48) = 52%

• The VE = 100% x {1 - (0.51)(0.48)} = 100% x {1 - 0.24} = 76% then combines these results to suggest that 76% of hospitalized SARS-CoV-2 infections from the Alpha variant that would have occurred sans vaccination were prevented by the single dose of AstraZeneca vaccine.

Based on these results, we can see that the vaccines demonstrate strong efficacy vs. hospitalized diseased from the Delta variant, with those fully vaccinated by Pfizer being protected 96%, AstraZeneca 92%, and even those with just a single dose of vaccine protected by 94% or 71% vs. hospitalized disease for Pfizer and AstraZeneca, respectively.

Given the relatively small sample sizes, there is some uncertainty in these estimates, as reflected in the numbers in parentheses in the chart that show the 95% confidence intervals, but these results provide strong rigorous evidence that the both vaccines are strongly protecting vs. vaccinated disease, and showing that the unvaccinated have substantially higher risk of hospitalized disease.

Explaining deaths from COVID-19 for Delta variant

Given all of this strong evidence that the vaccines are protecting against infections and hospitalizations for the Delta variant, where do these claims the "vaccines are not working" because "most COVID-19 deaths from Delta are coming from vaccinated people?"

This comes from the following table in the UK Delta technical report, that indeed shows that of the 117 reported COVID-19 deaths from Delta infections, 50/117 (43%) are from fully vaccinated people and 19/117 (16%) are from partially vaccinated people, so indeed almost 60% of deaths are from people who have been vaccinated, and only 44/117 (38%) are in those verified to be unvaccinated.

This appears to be a paradoxical result, but can be explained by the following factors:

1. While demonstratively very effective in preventing symptomatic infections, the vaccines are not perfect and some breakthrough infections still happen.

2. The very old people are nearly 100% vaccinated, so any SARS-CoV-2 cases that occur will be breakthrough infections after vaccinations.

3. The very old people are at much higher risk of death from any viral infection, including SARS-CoV-2, so even a small number of breakthrough infections in this age group can produce a substantial number of deaths (without vaccination, the number of cases in this group and thus the number of deaths would have been MUCH higher).

4. The young people are mostly unvaccinated, so the vast majority of SARS-CoV-2 infections that occur will be in the unvaccinated, and not be breakthrough infections.

5. Young people are at very low risk of death from any viral infection, including SARS-CoV-2, so even a very large number of infections will result in very few deaths.

These factors together explain the seemingly paradoxical result.

A key factor to keep in mind is the exceptionally low death rate we are seeing from the Delta variant -- just 117 deaths so far from >93k cases so far. Of course, it is still relatively early in the Delta surge, with the exponential growth not showing evidence of slowing yet and nearly 90% of Delta cases occurring in the past 4 weeks, which given the lag time from infection to death means many of the deaths that will be produced by these cases have not been seen yet. So it is premature to claim the Delta variant is docile and unconcerning.

Conclusions, and what to expect from Delta in the USA and other countries

From these extensive UK data, we have learned a great deal about the Delta surge in the UK:

• It is primarily driven by the young unvaccinated people

• While there are some breakthrough infections after vaccination, there are not many

• It is clear that a vast majority of infections that would have occurred sans vaccination were prevented in the vaccinated individuals.

• Rigorous analyses demonstrate that the vaccines retain near full efficacy against symptomatic or hospitalized infections for the Delta variant as they did for the Alpha variant and the wild type variant in the phase 3 clinical trials. The vaccines work very well against this variant; it is not an immune escape variant.

• The strong age difference in vaccination explains some of the unusual characteristics of the data, including the fact that most deaths have occurred in vaccinated people (because they are much older) and that the overall rate of hospitalization and death is low relative to other variants (because the Delta infected are very young).

So what do these results mean for other countries?

First, for unvaccinated countries, these results have very little direct relevance, for the following reasons:

1. The growth of cases is strongly suppressed by the substantial proportion vaccinated in the population. Without vaccination, the exponential growth would occur at a much faster rate, potentially 35-40% faster based on vaccination rates and efficacy estimates, than in the UK. As can be seen by simulating exponential growth, this type of extra growth would produce 2x, 5x, and 10x the number of Delta cases compared to what the UK experienced over a 1, 2, and 3 month period, respectively. This extra growth is enough to overwhelm hospitals and increase death rates, as seen in India.

2. The low death rate is driven by the fact that the vast majority of cases in the UK are among the young, primarily because the older adults at greater risk of death are strongly protected by the vaccines. In an unvaccinated society, a higher proportion of Delta cases would be from more vulnerable populations and thus would produce a higher death rate.

These factors explain why India experienced a much faster case growth rate and much higher death rate than the UK. Unvaccinated countries might experience Delta more like India and less like the UK, as we are starting to see in certain countries around the world.

Second, even for countries with high level of vaccination, the vaccine manufacturers used might make a big difference in what a Delta surge would look like in that country. Countries using predominantly lower efficacy vaccines would naturally have less protection against Delta's rapid transmission, while countries solely using vaccines with higher efficiency would have more protection. As mentioned above, a substantial proportion of the UK vaccinated population, perhaps bit more than half, have been vaccinated with the AstraZeneca vaccine whose lower efficacy numbers for Delta suggest 2.75x higher breakthrough infection rate than what would be experience with Pfizer, so the level of breakthrough infections may be much lower in countries like the USA and Israel that almost exclusively are vaccinating with the more efficacious mRNA vaccines.

For the USA, some reasonable expectations about our Delta surge are:

• The Delta surge will hit harder in unvaccinated communities, primarily in rural communities, and more heavy in the southeast and western regions of the country.

• We might see a higher death rate in some of these unvaccinated regions since a more substantial proportion of the >65 population at higher risk of death from infection is unvaccinated (10-20% unvaccinated) than in the UK (<5% unvaccinated)

• We might see a slower Delta transmission rate in the population given that: (1)The proportion previously infected is 50% higher in the USA than the UK, (2) the USA is using higher efficacy vaccines so will have fewer breakthrough infections and (3) While the USA and UK have similar proportion of population fully vaccinated (~50%), the USA has slightly higher percentages of fully vaccinated young people who are more likely to expose others to the virus (31% for 16-17, 36% 18-24, and 41% for 25-39 in the USA vs. <20% for the UK in the <40 age group.)

Putting all of this together, I expect that Delta should not ignite a national surge reminiscent of past surges, but will instead pronounced upticks in various local communities around the country. Think many local brush fires but none of them leading to an out of control forest fire. I expect the rate of hospitalizations and deaths overall to remain relatively low as in the UK, although in the unvaccinated individuals their individual risks could rival the rates seen in previous surges.

As we can see in this tweet by Tom Wenseleers, the Delta variant is already becoming the predominant strain in much of the USA so this Delta surge has now begun.

Missed opportunities with the UK data.

Given the richness of these UK data, there are several analyses that could have been done but were not:

1. Breakthrough infections by vaccine type: As mentioned above, it would be instructive to see breakthrough infections and their associated hospitalizations and deaths split out by vaccine type.

2. Reinfection rate for previously infected people who have not been vaccinated: It is clearly established in the literature that previous infection provides substantial protection against reinfection, even among the unvaccinated, including the UK's SIREN study. Given their electronic medical records should have evidence of past confirmed SARS-CoV-2 infection, they could look at the level of protection offered by previous infection against reinfection with the Delta variant. This is important since, while the data above establish the Delta variant does not have immune escape capabilities against immune protection offered by vaccination, there is very little data on whether Delta has immune escape potential against immune protection offered from previous infection. There is a dearth of information on reinfection risk for the Delta variant, and this could have provided such data to see whether we can expect the Delta variant to infect many who were previously infected.

3. Vaccine efficacy in previously infected after 1 or 2 doses: The same data would enable an estimation of additional vaccine efficacy provided by vaccination for those who were previously infected. There is strong literature showing that blood immune markers are greatly improved by the first vaccine dose among those who were previously infected, but not the second, suggesting perhaps previously infected people should be vaccinated with a single dose. However, there is practically no epidemiological data on how these laboratory measurements translate in terms of protection against reinfection. These UK data could provide a rigorous estimate of this quantity, and this estimate could be used to potentially revise the vaccination protocol to possibly suggest just a single dose for those previously infected, or even no vaccination for some subgroups depending on the individual cost-benefit analysis.

These analyses could provide further useful results that would help project how other countries with different levels of previous infection and different brands of vaccine might experience their Delta surge, and would also provide some much-needed epidemiological data to inform the best public health policies in terms of vaccination schedules. Hopefully the UK scientists with access to these data will consider performing these analyses.

UPDATE: By the way, these results from the UK data strongly agree with data included in a similar medRXiv paper from Canadian data. Following are the vaccine efficacies vs. symptomatic infection and hospitalization by variant and number of doses (Canada also has data on Moderna, and also on Beta/Gamma variants) :