1/ A thread to explain our recent work, and for the love of tweetorials 😄
2/ Pakistan- fifth most populous country in the world, first among LMICs to see the pandemic in late February 2020 . Now >269000 confirmed cases with over 5500 deaths, the largest number seen in Karachi, its largest city. Here is what Karachi is like in relation to Pakistan 
3/ Testing rates have been low in some areas and under reporting of cases due to fear and stigma of the disease and omission of mild symptomatic and asymptomatic cases. telegraph.co.uk/global-health/…
4/ This makes the case for population level sero-surveys like the ones from Spain or Geneva. But our resources are finite. So we opted to adapt the WHO-UNITY protocol. who.int/publications/i…
5/ In April as Phase 1, existing resources were mobilized to do a household survey for antibodies to SARS-CoV-2. We decide to do the survey in 2 areas - hot spots of district East and an area of District Malir where no positive PCR tests were reported.
6/ We saved backup samples at -20 degrees and -80 degrees, and used USA based IgG and IgM kits to analyse. At that point no state of the art testing was available. These kits yielded erroneous results ( IgM positivity of 13%! in April) and were discarded.
7/ The pandemic went full speed ahead in Pakistan. Here is the rolling average from our high transmission district.
8/ We repeated round one testing using chemiluminescence (CLI) and results made more sense with both areas having low test positivity <1% in phase 1
9/ There was a super spreader event (Eid) in Pakistan - May 12/3 2020. We began phase two of study in June (2nd half) about 4 weeks after Eid. Survey methodology in both phases was exactly alike. We used random sampling methods but they were slightly different by site.
10/In high transmission hot spots where we had a line listing of cases, we selected our reference point randomly from a list of cases.The household containing the case wasn't enrolled, an interval was determined through the second digit of a bank note to select the next household
11/ In low transmission area June, 4/164 PCR were + so households selected via simple random sampling. Some suggested this may overestimate the hotspot seroprevalence &underestimate low transmission seroprevalence. But serial nature allowed sites to be compared independently.
12/ Our raw seropositivity rates were 20% for phase 2 in District East(high transmission) and 13% for District Malir (low transmission). But we saw high clustering at household level (not a surprise given that a highly contagious infection is unlikely to be an independent event)
13/ This was confirmed through estimation of the household conditional risk of infection (second phase);41% &31% respectively. CRI can be calculated with aggregate data and we think its use is really underestimated. Here's some nice work that explains it.medrxiv.org/content/10.110…
14/ So we calculated age, gender stratified seroprevalence rates accounting for demographic makeup, household clustering and test inaccuracy. Notice the wide uncertainty in estimates.
15/ In conclusion: This is the first serial seroprevalence study on estimates of antibodies against SARS-CoV-2 from Pakistan using random sampling methods enrolling more than 2000 participants across all age groups and comparing between low-transmission and high-transmission area
16/ Highlights a rise in seroprevalence even in low transmission areas, high heterogeneity with urban slums being less affected. Also the first study to report the conditional risk of infection in households of urban slums showing a high conditional risk of infection in household
17/ Remarkable effort by 8 teams& two supervisors who led the survey. We achieved high individual participation rate- of over 75% given that this was a random selection to study a disease stigmatized in Pakistan. Should mention @imran_nisar @baileyfosdick @DanLarremore
@threadreaderapp unroll please
@threadreaderapp unroll please
