Immunocompromised children and young people are at no increased risk of severe COVID-19

Cohort

1645 participants consented to the study. Of these, 1527 successfully completed at least one weekly questionnaire. A further 505 were withdrawn during the study period (Fig. 1). The commonest reason for withdrawal was not completing the survey for three successive weeks. At week 32, the study team started contacting participants who had not completed questionnaires for three successive weeks regarding ongoing participation. This led to a sharp rise in withdrawals. Median age was 11 years (range 0–18). 55•1% of participants were female. Baseline characteristics are displayed in Table 1.

1. Download : Download high-res image (491KB)
2. Download : Download full-size image

Fig. 1. Weekly participant numbers and weekly questionnaire response rate (16th March 2020–14th March 2021).

Table 1. Characteristics of participants referred to the ImmunoCOVID study.

	Referred but not participating	Withdrawn participants	Currently active participants	COVID positive(subset)
Sample size	1373	505	1022	38
Female	53•8%	54•5%	55•1%	55•3%
Teenagers	64•4%***	55•6%	55%	68•4%
Age	11•9 + 4•3***	10•8 + 4•4	10•7 + 4•3	12•6 + 4•2**
UK geographySouthMiddleNorth	***9•9%44•0%46•1%	*16•6%41•1%42•3%	21•4%42•1%36•5%	18•4%47•4%34•2%
Treatment teamRheumatologyImmunologyRespiratoryOncologyGastroenterologyNephrologyNeurologyHaematologyNeonatologyDiabetesInfectionDermatologyHeart and Lung	42•8%8•1%7•7%5•3%7•5%11•7%3•1%3•7%0•3%7•3%2•0%0•5%0•1%	48•2%10•4%6•3%2•6%8•3%12%3•9%1•6%0•2%5•1%0•4%0•8%0•2%	48•7%9•3%5•1%4•3%9•7%10•7%2•2%2•3%0•2%5•3%0•4%0•9%0•6%	52•6%13•2%5•3%7•9%2•6%7•9%0%2•6%0%5•3%2•6%0%0%
DiagnosisGastroenterologyNephrologyTransplantsDiabetesImmunodeficiencyOncologyRespiratoryOtherJuvenile Idiopathic ArthritisOther immune condition		10•7%11%7•4%6•4%10•3%5•6%7•2%29•1%37•4%15•5%	13•1%9%7•6%6•5%8•7%7•7%8•5%26•7%35•4%18•5%	0%*10•8%13•5%5•4%18•9%10•8%8•1%29•7%40•5%16•2%
MedicationOther drugsChemotherapyAntibioticsOther immunosuppressantsSystemic corticosteroidsBiologicsMethotrexate		49%0•9%16•6%16•7%*17•5%34•7%27•2%	53•7%0•8%16%21•6%19%35•1%29•3%	58•8%0%16•7%23•7%16•7%25%27•8%

Statistical significance is calculated with Chi-Square statistics. If one or more cells in the contingency table had an expected count of less than five the Fisher’s exact statistic was calculated. Group differences in age were calculated with F test.

 

⁎

significant at 0•05 level.

⁎⁎

significant at 0•01 level.

⁎⁎⁎

significant at 0•001 level.

Serology

Between December 2020 and January 2021 all participants (n = 1073) were invited to take part in an optional serology sub-study. They were provided with information sheets electronically with a link to an online consent form. Consenting participants were sent sampling kits via post between February and March 2021. Participants collected a dried blood sample using a safety lancet and Mitra® sampling device. Validated in-house enzyme-linked immunosorbent assays (ELISA) were carried at the virology laboratories of the University of Nottingham to detect SARS-CoV-2 anti-nucleocapsid and anti-spike IgG antibodies, as previously described.³² All assays were performed on Opentrons OT-2 Precision liquid handling robots (full protocol available on request). Original assay validations were performed on a separate sample set using commercial anti-S1 (Euroimmun) and anti-N assays (Roche) and conducted at the Rare and Imported Pathogens Laboratory at Public Health England using manufacturers’ protocols.

Antibodies were first extracted from dried samples to a 1:15 dilution by shaking for 2 h in 150 µL phosphate buffered saline (PBS) containing 0•1% Triton-x100 and 0•05% sodium azide. Extracted samples were diluted to 1:120 with 3% skimmed milk powder in PBS containing 0•05% Tween 20 and 0•05% sodium azide.

ELISAs were performed by coating 384 well NUNC maxisorp plates (Thermofisher) with either 20 µL/well of 0•5 µg/mL of SARS-CoV-2 spike protein S1 subunit (His-tagged; Sinobiologicals, GmbH) or SARS-CoV-2 Nucleocapsid (His-Tagged; Sinobiologicals, GmbH) in carbonate-bicarbonate buffer (CBC; Merck UK), sealing with foil film and incubating overnight at 4 °C. Wells were then washed three times with 100 µL PBS with 0•05% Tween 20 (PBST) using a Thermofisher Wellwash Versa plate washer with a final aspirate. Wells were immediately filled with 100 µL of 3% skimmed milk powder (w/v) in PBS and 0•05% sodium azide, foil sealed and blocked overnight at 4 °C. Assay plates were then washed three times as before and duplicate wells filled with 20 µL of the 1:120 diluted sample, diluted in 3% skimmed milk powder (w/v) in PBS and 0•05% sodium azide.

Control samples included duplicates of a 1:120 dilution of a pool of sera derived from 32 individuals previously identified as having high nucleocapsid and high spike S1 reactivity. Three negative samples, each derived from a separate pool of 21 pre-covid sera were added to separate wells in duplicate. Wells containing only dilution buffer were used as non-serum blank. After 15 min incubation at room temperature (21 °C), the plate was washed three times in PBST. This was followed by addition to each well of 20 µL of gamma chain-specific anti-human IgG-HRP conjugate (Sigma A0170–1ML) at 1:30,000 dilution in PBST and incubation for 15 min at room temperature. Wells were then washed three times in PBST. 40 µL of One-step ultraTMB substrate solution (Thermofisher) was added to each well. After 20 min incubation at room temperature, 40µL of 2 N H2SO4 was added to each well and the 384 well plate read at 450 nm absorption in a GlowMax Explorer microplate reader (Promega Corp.). Data from absorbance measurements were output to Excel data sheets and subsequently imported into Graphpad Prism for basic analysis. More detailed statistical analysis was then carried out on SPSS.

Seropositivity was indicated according to the ratio method (i.e. ratio of the average sample optical density/average Negative optical density). According to this approach a ratio of > = 1•1 was classified as positive and a ratio of < 1•1 was classified as negative.

Statistical analysis

Data cleaning, formatting, and analyses were performed with SAS 9.4 (SAS Institute INC, Cary NC). Survival analysis techniques were used to explore the risk of SARS-CoV-2 infection, as measured by PCR testing, to participants throughout the first year of the study. The semi-parametric methodology of Cox regression (Cox, 1972) was employed. Factors associated with the risk of SARS-CoV-2 infection were determined. In addition, the symptoms uniquely experienced at time of SARS-CoV-2 infection were determined. For analysis purposes we have treated infection as a non-repeated event. Data was collected on a weekly basis, creating ties in the data. Parameters are therefore estimated using the estimation method developed by Efron (1977). Participant information was only collected from participants during the study period from recruitment to the last data received (when they may have chosen to withdraw formally or informally). Data prior to the first questionnaire and beyond the last questionnaire received per participant is unknown and is described as having been censored. Every individual was considered at risk of SARS-CoV-2 infection when they entered the study until they either were infected or were censored. To control for late entry, participants’ risk for infection was determined between study entry and withdrawal or finish time. We assume that censoring due to withdrawal is non-informative, but receiving “other immunosuppressants” (azathioprine, tacrolimus or mycophenolate mofetil) has been included in the survival model as participants receiving these treatments were less likely to withdraw from the study.

Role of the funding source

The funders had no role in study design, data collection, data analysis, data interpretation, or writing of the manuscript. The corresponding author had full access to all data and had final responsibility for the decision to submit for publication.

PCR confirmed cases

Participants reported a total of 1355 PCR tests through the study period. The majority of participants (919, 60•2%) did not report any PCR tests. 576 (37•8%) reported between 1 and 5 tests and 2% reported between 6 and 20 tests. During the government advised shielding period (March–August 2020) no SARS-CoV-2 infections were reported. The first positive SARS-CoV-2 PCR test was in week 30 (28th September–5th October). Further positive tests were subsequently reported, with peaks in rates in weeks 36 (9th–15th November 2020) and 42 (21st–27th December 2020) with five and six cases, respectively (Fig. 2). There were a total of 38 PCR detected SARS-CoV-2 infections, giving an incidence rate of 3•7% (n = 1022). Of these six were asymptomatic, 28 had symptoms managed at home, and four reported hospital admission. Those reporting hospital admissions are described in Table 2. These participants all had varying presentations and significant underlying co-morbidities and therapies; two were transplant patients, one had acute lymphoblastic leukaemia, and one had X-linked agammaglobulinemia. Clinical presentation was varied but none reported acute severe COVID-19 or paediatric multisystem inflammatory syndrome. None were admitted to paediatric intensive care units or died.

1. Download : Download high-res image (255KB)
2. Download : Download full-size image

Fig. 2. Number of participant SARS-CoV-2 infections by week (16th March 2020–14th March 2021).

Table 2. Description of PCR positive SARS-CoV-2 infection reporting hospital admission.

Week of PCR	Week of admission	Age in years	Sex	Admission Details	Past Medical History	Baseline Medication
42	42	17	Male	Admitted with negative PCR, symptoms of bowel obstruction and underwent laparotomy. Asymptomatically tested positive prior to discharge. Subsequently readmitted with fever and abdominal pain for three days before symptom resolution.	liver transplant, atypical haemolytic uraemic syndrome, chronic kidney disease, and superior vena cava stenosis	prednisolone, sirolimus, mycophenolate mofetil, hydralazine, doxazosin, amlodipine, losartan, atenolol, allopurinol, and desloratadine
				Subsequent serology negative.
42	44	12	Male	Initially tested positive with only symptom being abdominal pain. Subsequently admitted for two days with bacterial chest infection. Put down to post viral immunosuppression. Admission swab negative.	heart transplant	sirolimus, mycophenolate mofetil, pravastatin, and melatonin
				Subsequent serology negative.
43	52	17	Male	Tested positive, initially managed at home with antibiotics. Subsequently had persistent fever leading to investigation with bronchoscopy which showed changes consistent with SARS-CoV-2 infection. Five day admission for remdesivir giving symptom resolution.	X-linked agammaglobulinemia	human normal immunoglobulin
45	45	6	Male	Household tested positive two weeks prior but no test sought at time for participant. Admission with anaemia for two days with positive PCR swab. Received one unit of blood. Tests for leukaemia relapse negative.	acute lymphoblastic leukaemia, asthma, and autism	mercaptopurine, co-trimoxazole, gabapentin, fluticasone with salmeterol, montelukast, methotrexate, macrogol, and esomeprazole

Hazard analysis

Assuming that hazard remained constant between August 2020 and March 2021, an estimate of hazard across the whole cohort is 0•00,148 or 38/25,687 (number of cases/ total number of answered questionnaires). This is equivalent to a positive SARS-CoV-2 PCR test once every 676 weeks. This hazard was in the context of varying disease incidence during the pandemic, and varying approaches of participants to shielding or not. While NHS electronic systems sent text messages telling children and young people designated “at risk” to shield based on adult diagnostic categories and the assumed risk, many specialist clinicians advised school attendance to children with all but the most severe immune problems.¹⁶

The Cox regression model indicates for each additional year of age the hazard of infection goes up 13%. Having an immunodeficiency increases hazard of infection almost three times. Those individuals who reported positive PCR tests were more likely than other participants to report worsening of fever, cough, and sore throat symptoms during the same week.

Serology

26 participants reported receiving a first dose of a COVID vaccine prior to serology. Of these, six (23•1%) were serologically positive with a mean of 23 days between vaccination and sampling (range 14–33). Three of these were positive for only anti-S1 IgG and three positive for both anti-S1 and anti-N IgG. Those who were negative had a mean of 15•65 days between vaccination and sampling (range 1–42). Vaccinated participants were excluded from further analysis.

Of 452 unvaccinated participants, anti-N and/or anti-S antibodies were detected in 41 samples (9•1%). Of those participants reporting a prior positive PCR test 18 took part in the serology sub-study and of these, antibodies were detected in nine (50%). Four of these were positive for only anti-N IgG, one positive for only anti-S1 IgG, and four positive for both anti-N and anti-S1 IgG. In those with no previous positive PCR test, antibodies were detected in 32 of 434 (7•4%). Seventeen of these were positive for only anti-N IgG, six positive for only anti-S1 IgG, and nine positive for both anti-N and anti-S1 IgG. Associations were tested and are shown in Table 4. Detection of antibodies was associated with being female and over ten years old. Associations were also found with having had a transplant and being on immunosuppressants or corticosteroids. In those with prior positive PCR tests no associations were found between serology result and interval between positive test and serology test.

Serology

While 9•1% (41/452) of unvaccinated children and young people tested positive for antibodies reactive to SARS-CoV-2 in the serology sub-study, only 2•5% (38/1527) reported positive SARS-CoV-2 PCR during the survey study. This suggests a proportion of participants had infection which was not confirmed by PCR test. Only 50% (9/18) of participants included in the serology sub-study who reported prior positive SARS-CoV-2 PCR tested positive for antibodies reactive to SARS-CoV-2. Half of PCR positive participants were not serologically positive when tested and some seroconverted with no prior detected infection. This may mean SARS-CoV-2 incidence was higher than reported here with participants having an infection undetected by PCR and serology.

Rates reported in this UK cohort are comparable to those (6•9%, 68/992) reported in a cohort of children of UK healthcare workers during the first wave of the pandemic 2020.³⁹ Other UK paediatric seroepidemiology studies have yet to be formally reported.^26,27

The serology rates are comparable to a cross-sectional study in Madrid from September 2020 to February 2021 in paediatric rheumatology patients where 5•7% (6/105) had PCR positive cases and 20•0% (21/105) had serology positive cases with no overlap between these groups.²² Our results are comparable to initial data from a prospective Belgian cohort study which reported 6•0% (15/250) serology positive participants in paediatric patients with chronic medical conditions. Only one seropositive participant, with a diagnosis of cystic fibrosis, required hospital admission while the rest were either asymptomatic or had mild symptoms.²¹

These rates are higher than observed in a cross sectional study from the USA from April 2020 to May 2020 and June 2020 to July 2020. In immunocompromised paediatric patients there were 1% (5/485) serology positive cases and 0•4% (2/485) PCR positive cases with one of these in both groups.²⁵

The reason for PCR positive cases not having detectable antibodies has not been established in our study and no associations are evident. Some evidence from inflammatory bowel disease patients suggests infliximab and other cytokine inhibitors may reduce seroconversion.^23,40 Of participants with negative serology tests but prior PCR confirmed SARS-CoV-2 infection, five had juvenile idiopathic arthritis, one immunodeficiency, and one type one diabetes. The remaining two are described in Table 2 as they reported admission with SARS-CoV-2 infection.

Limited conclusions can be drawn from serology in those who received a prior dose of a COVID vaccine given the single time point and short interval between vaccination and antibody testing.

Limitations

Limitations of our study include random censoring, due to late entry or withdrawal, which was experienced by 60% of participants. Due to this level of censoring, less confidence should be placed in results than the calculated confidence intervals. Questionnaire data is patient or parent reported and subject to recall bias, although we tried to minimize this by sending the questionnaire every week. Symptom over-reporting is possible, particularly in weeks when participants reported positive PCR tests. Symptom under-reporting is possible, particularly in participants with a baseline of symptoms from chronic disease, although less likely due to levels of anxiety in this cohort.¹⁸ Hazard of infection is likely over-estimated due to self-selection bias with participants presumably responding to surveys more frequently in weeks they reported a positive PCR swab. The shielding precautions taken will have varied both between individuals and through the course of the pandemic, so the individual hazard of SARS-CoV-2 infection will vary. The serology sub-study cohort was not representative of the overall study as described in the results section.

Following the study period reported here, in May 2021 the predominant strain in the UK became the Delta variant. There has been concern about the increased transmissibility and severity found in adults infected with the Delta variant.⁴¹Fortunately, children have been shown to not be at increased risk of severe disease from it⁴². This emphasises the need for ongoing surveillance of the impact of variant strains on the risk to immunocompromised children.