A reasonable worst-case scenario could see protection fall to below 50 per cent for the elderly and healthcare workers by winter, and expert has said.
The Pfizer jab was 88 per cent effective at preventing Covid-19 infection a month after the second dose.
But after five to six months the protection decreased to 74 per cent, suggesting protection fell 14 percentage points in four months, latest analysis from the Zoe Covid study indicates.
With the AstraZeneca vaccine, there was a protection against infection of 77 per cent one month after the second dose.
After four to five months protection decreased to 67 per cent, suggesting protection fell by 10 percentage points over three months.
The study drew on more than 1.2 million test results and participants.
An initial 96.2 per cent risk reduction in infection - up to two months after the second dose - was found by Pfizer’s mid-term efficacy trial.
There was an 83.7 per cent reduction more than four months after the second dose, a 12.5 percentage point risk reduction.
Real world analysis would be expected to show less protection than clinical trials, and the vaccines were not trialled against the now dominant delta variant of the virus.
The Zoe Covid Study launched an app feature in December 2020 to enable logging of Covid-19 vaccines and monitor real-world side-effects and effectiveness in its cohort of more than a million active users.
Zoe used data from vaccines which were logged from December last year to July 2021 and from infections which occurred between May this year when the delta variant became dominant, and July.
The results have been adjusted to give an average risk of infection reduction across the population.
While protection appears to decrease steadily, individual risk may vary due to individual variation in antibody duration, researchers say.
Across the UK, vaccines were rolled out among the older and the most vulnerable in society along with health workers before rolling out vaccines to younger age groups across the UK.
This means the majority of people who had their second dose five to six months ago will be older or considered vulnerable due to other health reasons.
These people are now likely to be at increased risk of Covid-19 compared to those vaccinated more recently, it suggests.
Researchers say that in order to confidently show how vaccine effectiveness changes over time in different age groups, more data is needed over a longer period of time.
Professor Tim Spector, lead scientist on the Zoe Covid Study app, said: “In my opinion, a reasonable worst-case scenario could see protection below 50% for the elderly and healthcare workers by winter.
“If high levels of infection in the UK, driven by loosened social restrictions and a highly transmissible variant, this scenario could mean increased hospitalisations and deaths.
“We urgently need to make plans for vaccine boosters, and based on vaccine resources, decide if a strategy to vaccinate children is sensible if our aim is to reduce deaths and hospital admissions.
“Waning protection is to be expected and is not a reason to not get vaccinated.
“Vaccines still provide high levels of protection for the majority of the population, especially against the Delta variant, so we still need as many people as possible to get fully vaccinated.”