Rapid responses are electronic comments to the editor. They enable our users
to debate issues raised in articles published on bmj.com. A rapid response
is first posted online. If you need the URL (web address) of an individual
response, simply click on the response headline and copy the URL from the
browser window. A proportion of responses will, after editing, be published
online and in the print journal as letters, which are indexed in PubMed.
Rapid responses are not indexed in PubMed and they are not journal articles.
The BMJ reserves the right to remove responses which are being
wilfully misrepresented as published articles or when it is brought to our
attention that a response spreads misinformation.
From March 2022, the word limit for rapid responses will be 600 words not
including references and author details. We will no longer post responses
that exceed this limit.
The word limit for letters selected from posted responses remains 300 words.
I am writing in response to Mr Stafford's feature (ref 1) on Streek et al (ref 2).
There are several issues regarding the applicability of their conclusions, throwing doubt on some claims reported in the BMJ article.
1. Study sample selection and its applicability to Germany
The test population were drawn from the town of Gangelt from the following intention:
"Based on the sample size recommendations of the World Health Organisation (WHO) (see below), the aim was to collect data from at least 300 households in Gangelt. To reach this target, a sample of 600 persons aged older than 18 years was drawn from the civil register. Sampling was done randomly under the side condition that all 600 persons had different surnames, as it was assumed that different surnames were likely to indicate different households. After sampling, the 600 selected persons were contacted by mail and were invited to the study acquisition center, which was established at the site of a public school in Gangelt. The letters sent to the 600 selected persons also included invitations for all persons living in the respective households to participate in the study."
As a result:
"For the study, 600 adult persons with different surnames in Gangelt were randomly selected, and all household members were asked to participate in the study. Data and materials were collected over a 7-day period (March 30 to April 6) six weeks after super-spreading event. Of the 1,007 individuals participating in the study, 987 individuals were seen in the local study acquisition center in a community school, and 20 individuals were visited in their homes due to age or limited mobility."
"Complete information from both pharyngeal swabs and blood samples was available for 919 study participants living in 405 households"
Therefore 56% of the study participants are in fact "secondary" participants involved by virtue of the original invitation to one of the participants living in the household.
This is important, since infection risk of a person is expected to be higher in a household in which at least one other person was infected, a fact also proven by this study.
Furthermore, the researchers said:
"The average number of people in household clusters examined in this study was 2.27 (sd = 1.11, range 1-6) compared to Gangelt (2.44 as of 2011), the state NRW (2.02, as of Dec 2018) and Germany (1.99, as of Dec 2018)."
Accordingly, there are more people living in households of 2 or more than the national German average household. The researchers reported 24.2% are single households whereas the national statistics is 41.7% (ref 3).
To confound further, supplementary figure 1 showed over-representation of age groups 40-65 years, and 65 years or older in the study population compared to the town of Gangelt and the national average in Germany.
Therefore there is really very limited applicability of estimated infection prevalence (and infection fatality rate derived by this prevalance) in Gangelt and Germany.
There are well-known regional variation in prevalence (and hence calculated mortality rate based on prevalence) within a country, given examples worldwide. This may be influenced by availability of testing kits, regional testing criteria and community demographics.
Hence, the headline "Number of Germans infected could be 10 times higher than official estimates" is really based on assumptions which can easily be challenged.
2. "Virus transmission within households is relatively low"
The researchers concluded:
"The unexpectedly low secondary infection risk among persons living in the same household has important implications for measures installed to contain the SARS-CoV-2 virus pandemic."
They considered only "a relatively moderate increase of the secondary infection risk which depended on the household cluster size (increase from 15.5% baseline risk by 28% for two people, 20% for three people, 3% for four people)."
This interpretation is challenging since their comparison is based on absolute risk of one person getting COVID-19 from community contact:
"A significant association between household cluster size and the per-person infection risk was found (Fig. 5B, p<0.001). In a two-person household cluster, the estimated risk for the second infection increased from 15.53% to 43.59% [25.26%; 64.60%]; in a three-person household cluster the estimated risk for the second and third persons increased from 15.53% to 35.71% [19.57%; 55.60%] each, and in a four-person household cluster the estimated risk for the second, third and fourth persons increased from 15.53% to 18.33% [9.67%; 28.74%] each. For household clusters with at least one infected child (< 18 years), the estimated per-person risk for the other person to be infected in three-person household clusters increased from 15.53% to 66.67% [21.83%, 100.00%] and in four-person household clusters from 15.53% to 33.33% [9.02%; 71.60%]."
The overall absolute risk (AR) of 35-43% of contracting COVID-19 infection is considered at best "moderate" by the researchers.
Most clinicians are familiar with issues involving "relative risk" (RR) as a measure of risk; the classic examples illustrating dangers of RR as risk estimation uses conditions with very low incidence or prevalence, often less than 5 or 10%. However on the matter of a likely linked association (as household members are by their proximity to the index case in their home) where the prevalence is significantly higher than 10%, a 43% AR of infection by virtue of living in a household of 2 (which represent 45.4% of the study) with the other member infected is by no means a "moderate risk".
As children are under-represented, households of 4 (or more) probably involve 4 adults living in a large home; with increased living space affording more practical isolation measures and reduced physical contact. Conversely the living space for 1- or 2-person household is probably far smaller which explains the dramatic increase in AR and RR.
I would certainly be worried if my AR of getting COVID-19 infection is almost 50% from whatever daily activities and circumstances, including living in a household with an infected person.
And almost 50% of the study population lives in a household of 2.
Hopefully this will help others reconsider the validity of study conclusions and assertions.
Study conclusions were based on flawed assumptions and perception of risks
Dear Editors
I am writing in response to Mr Stafford's feature (ref 1) on Streek et al (ref 2).
There are several issues regarding the applicability of their conclusions, throwing doubt on some claims reported in the BMJ article.
1. Study sample selection and its applicability to Germany
The test population were drawn from the town of Gangelt from the following intention:
"Based on the sample size recommendations of the World Health Organisation (WHO) (see below), the aim was to collect data from at least 300 households in Gangelt. To reach this target, a sample of 600 persons aged older than 18 years was drawn from the civil register. Sampling was done randomly under the side condition that all 600 persons had different surnames, as it was assumed that different surnames were likely to indicate different households. After sampling, the 600 selected persons were contacted by mail and were invited to the study acquisition center, which was established at the site of a public school in Gangelt. The letters sent to the 600 selected persons also included invitations for all persons living in the respective households to participate in the study."
As a result:
"For the study, 600 adult persons with different surnames in Gangelt were randomly selected, and all household members were asked to participate in the study. Data and materials were collected over a 7-day period (March 30 to April 6) six weeks after super-spreading event. Of the 1,007 individuals participating in the study, 987 individuals were seen in the local study acquisition center in a community school, and 20 individuals were visited in their homes due to age or limited mobility."
"Complete information from both pharyngeal swabs and blood samples was available for 919 study participants living in 405 households"
Therefore 56% of the study participants are in fact "secondary" participants involved by virtue of the original invitation to one of the participants living in the household.
This is important, since infection risk of a person is expected to be higher in a household in which at least one other person was infected, a fact also proven by this study.
Furthermore, the researchers said:
"The average number of people in household clusters examined in this study was 2.27 (sd = 1.11, range 1-6) compared to Gangelt (2.44 as of 2011), the state NRW (2.02, as of Dec 2018) and Germany (1.99, as of Dec 2018)."
Accordingly, there are more people living in households of 2 or more than the national German average household. The researchers reported 24.2% are single households whereas the national statistics is 41.7% (ref 3).
To confound further, supplementary figure 1 showed over-representation of age groups 40-65 years, and 65 years or older in the study population compared to the town of Gangelt and the national average in Germany.
Therefore there is really very limited applicability of estimated infection prevalence (and infection fatality rate derived by this prevalance) in Gangelt and Germany.
There are well-known regional variation in prevalence (and hence calculated mortality rate based on prevalence) within a country, given examples worldwide. This may be influenced by availability of testing kits, regional testing criteria and community demographics.
Hence, the headline "Number of Germans infected could be 10 times higher than official estimates" is really based on assumptions which can easily be challenged.
2. "Virus transmission within households is relatively low"
The researchers concluded:
"The unexpectedly low secondary infection risk among persons living in the same household has important implications for measures installed to contain the SARS-CoV-2 virus pandemic."
They considered only "a relatively moderate increase of the secondary infection risk which depended on the household cluster size (increase from 15.5% baseline risk by 28% for two people, 20% for three people, 3% for four people)."
This interpretation is challenging since their comparison is based on absolute risk of one person getting COVID-19 from community contact:
"A significant association between household cluster size and the per-person infection risk was found (Fig. 5B, p<0.001). In a two-person household cluster, the estimated risk for the second infection increased from 15.53% to 43.59% [25.26%; 64.60%]; in a three-person household cluster the estimated risk for the second and third persons increased from 15.53% to 35.71% [19.57%; 55.60%] each, and in a four-person household cluster the estimated risk for the second, third and fourth persons increased from 15.53% to 18.33% [9.67%; 28.74%] each. For household clusters with at least one infected child (< 18 years), the estimated per-person risk for the other person to be infected in three-person household clusters increased from 15.53% to 66.67% [21.83%, 100.00%] and in four-person household clusters from 15.53% to 33.33% [9.02%; 71.60%]."
The overall absolute risk (AR) of 35-43% of contracting COVID-19 infection is considered at best "moderate" by the researchers.
Most clinicians are familiar with issues involving "relative risk" (RR) as a measure of risk; the classic examples illustrating dangers of RR as risk estimation uses conditions with very low incidence or prevalence, often less than 5 or 10%. However on the matter of a likely linked association (as household members are by their proximity to the index case in their home) where the prevalence is significantly higher than 10%, a 43% AR of infection by virtue of living in a household of 2 (which represent 45.4% of the study) with the other member infected is by no means a "moderate risk".
As children are under-represented, households of 4 (or more) probably involve 4 adults living in a large home; with increased living space affording more practical isolation measures and reduced physical contact. Conversely the living space for 1- or 2-person household is probably far smaller which explains the dramatic increase in AR and RR.
I would certainly be worried if my AR of getting COVID-19 infection is almost 50% from whatever daily activities and circumstances, including living in a household with an infected person.
And almost 50% of the study population lives in a household of 2.
Hopefully this will help others reconsider the validity of study conclusions and assertions.
References:
1. https://www.bmj.com/content/bmj/369/bmj.m1862.full.pdf
2. http://www.ukbonn.de/C12582D3002FD21D/vwLookupDownloads/Streeck_et_al_In...
3. https://www.destatis.de/EN/Themes/Society-Environment/Population/Househo...
Competing interests: No competing interests