Trajectory patterns of SARS-CoV-2 neutralising antibody response in convalescent COVID-19 patients.

Background: The adaptive immune responses of COVID-19 patients contributes to virus clearance, restoration of health and protection from re-infection. The patterns of and the associated characteristics with longitudinal neutralising antibody (NAb) response following SARS-CoV-2 infection are important in their potential association with the population risks of re-infection.
Methods: This is a longitudinal study with blood samples and clinical data collected in adults aged 18 or above following diagnosis of SARS-CoV-2 infection. NAb levels were measured by the SARS-CoV-2 surrogate virus neutralisation test (sVNT). Anonymous clinical and laboratory data were matched with surveillance data for each subject for enabling analyses and applying latent class mixed models for trajectory delineation. Logistic regression models were performed to compare the characteristics between the identified classes.
Results: In 2020-2021, 368 convalescent patients in Hong Kong are tested for NAb. Their seroconversion occur within 3 months in 97% symptomatic patients, the level of which are maintained at 97% after 9 months. The NAb trajectories of 200 symptomatic patients are classified by the initial response and subsequent trend into high-persistent and waning classes in latent class mixed models. High-persistent (15.5%) class patients are older and most have chronic illnesses. Waning class patients (84.5%) are largely young adults who are mildly symptomatic including 2 who serorevert after 10 months. Conclusions: Characteristic sub-class variabilities in clinical pattern are noted especially among patients with waning NAb. The heterogeneity of the NAb trajectory patterns and their clinical association can be important for informing vaccination strategy to prevent re-infection.

Introduction

SARS-CoV-2, the causative virus of COVID-19, is an extremely contagious pathogen that has continued to cause major outbreaks globally ever since its emergence in late 2019 (https://www.who.int/emergencies/diseases/novel-coronavirus-2019). Before the widespread use of vaccine and in the absence of definitive curative therapy in most places, recovery of SARS-CoV-2 infected patients depends almost exclusively on the functioning of host immunity[1]. Studies have shown that both humoral and cell-mediated immunity contributed to the clearance of the virus and restoration of health[2-4]. With the increasing availability of validated serological tests, antibody responses have become widely investigated for clinical evaluation. The profiling of antibody responses covers different antigen and immunoglobulin classes. Whereas different components of humoral immunity exhibited distinct kinetics[5], antibody against receptor binding domain (RBD) or spike protein normally rises to a peak and becomes maintained for over 5 months, corresponding with virus neutralisation capacity[6]. Similar pattern was observed in another study on seropositive healthcare workers with antibody monitoring for 8 months after symptom onset[7].

An evaluation of the dynamics of neutralising antibody responses after acute SARS-CoV-2 infection is crucial for understanding not just the pace of health recovery but also the risk of re-infection. A study reported that neutralising antibody progressively falls after 5–8 weeks after symptomatic infection though it remains detectable by 8 months[8]. Asymptomatic and mildly symptomatic infections are associated with lower neutralising antibody levels[9,10], while older patients may have higher level[10]. Overall, heterogeneity of neutralising antibody responses within 3 months after diagnoses has been shown in some studies[10,11]. A more recent study with sampling up to 6 months has identified 5 different patterns of neutralising antibody dynamics[12]. With vaccination playing a central role in protecting individuals from SARS-CoV-2 infection, delineation of the dynamics of neutralising antibody response over longer interval would be important in assessing the vaccination needs of convalescent patients. It is against this background that we undertook to examine the trajectory of neutralising antibody (NAb) responses and factors associated with its pattern in SARS-CoV-2 patients who had recovered from COVID-19 of different severity. In this study, we classified convalescent symptomatic patients by their differences in the intensity of NAb responses initially and overtime, showing that high-persistent responses were associated with older age and the presence of chronic illnesses.

Methods

Study design and data sources

This is a longitudinal study with the collection of blood samples and clinical data in COVID-19 patients between February 2020 and February 2021 in Hong Kong, where stringent isolation policy is in place with all confirmed cases are reported and hospitalised irrespective of symptoms, travel, and contact history. COVID-19 vaccination programme started from 22 February 2021 while the reporting date of the last recruited patient was 20 February 2021. Blood sampling was performed at multiple time points during hospitalisation and follow-ups after discharge. Written informed consent was obtained from each participant. Separately, anonymous surveillance data of COVID-19 reported cases in Hong Kong were collected from the Centre for Health Protection, Hong Kong Special Administrative Region Government. The data included socio-demographics, history of chronic illnesses at diagnosis, travel history, and epidemiological linkages[13]. Anonymous clinical and laboratory records of all reported cases with case identifiers matched with surveillance data were retrieved from the Hospital Authority, Hong Kong. Ethical approval of the Joint Chinese University of Hong Kong—New Territories East Cluster Clinical Research Ethics Committee (CREC Reference Number: 2020.218) was obtained. This study is registered on ClinicalTrials.gov (NCT05028881).

Participants and setting

Patients were recruited from the Prince of Wales Hospital, a tertiary hospital in the public service with catchment for some 1.8 million population in the New Territories East Region of Hong Kong. The inclusion criteria included adults of age 18 or above admitted to the hospital with a confirmed diagnosis of SARS-CoV-2 infection, with the detection of SARS-CoV-2 nucleic acid from respiratory specimens. Diagnosis of SARS-CoV-2 infection was made by reverse transcription polymerase chain reaction (RT-PCR) assays on respiratory specimens at the hospital laboratory followed by confirmation by the Public Health Laboratory of Centre for Health Protection.

COVID-19 serology

The humoral response against SARS-CoV-2 including IgG and neutralising antibodies were examined by using commercially available ELISAs. The analysis of SARS-CoV-2 antibodies were performed using the SARS-CoV-2 NP IgG ELISA (nucleocapsid protein-based antigen; ImmunoDiagnostics) for IgG NP, and anti-SARS-CoV-2 IgG ELISA (S1 subunit of spike protein-based antigen; EUROIMMUN) for IgG spike. NAb were measured by the SARS-CoV-2 surrogate virus neutralisation test (sVNT), based on antibody-mediated blockage of ACE2-spike RBD interaction (RBD-targeting NAbs; GenScript)[14]. A positive NAb result with sVNT was defined as a level above 20% inhibition based on manufacturer’s recommendation. The tested specimens were heparinised blood which were separated and stored in aliquots of 0.5 mL at −80 °C until use. They were retrieved and heat-inactivated at 56 °C for 30 mins before testing. All assays were performed in accordance with the manufacturers’ instructions.

Data processing and analysis

Patients with both IgG NP and NAb measured between day 15 and 90 from onset date for symptomatic patients or reporting date for asymptomatic patients were included in the study. The main outcome was NAb level, and the secondary outcomes were IgG spike and IgG NP detection. We defined severe/critical condition when there was diagnosis of pneumonia and requirement of supplemental oxygen, or organ support, or admission to intensive care unit (ICU). The sociodemographic variables including gender, age, onset date, reporting date, hospital admission and discharge dates, travel history (e.g. origin and mode of travel), and epidemiological linkage collected in this study were used to link the surveillance and hospital data. Records of patients that could not be successfully linked with other data sources were included in the analyses, with the corresponding variables marked as missing values. Descriptive statistics were used to display the characteristics of the data in the study.

To delineate the trajectories of NAb (ranging between 0% and 100%) since the onset date, symptomatic participants with at least 2 sVNT measurements in separate months and who had been followed-up for at least 3 months from the onset date were analysed. The time variable in the analysis was the interval (months) between the onset date and the blood sample collection date. During hospitalisation, the first record, and records with the maximum and the minimum values of repeat laboratory measurements were selected as variables. Laboratory markers included albumin, CKD-EPI eGFR[15], Ct-value of SARS-CoV-2 RT-PCR from upper respiratory tract specimens, globulin, haemoglobulin, lymphocyte, neutrophil, and white blood cell counts.

Latent class mixed model (LCMM) was applied for NAb trajectory delineation. LCMM is a combination of latent class model and mixed model for repeated measurements, which have been applied in classifying the anti-spike antibody response to SARS-CoV-2 vaccines[16], and natural SARS-CoV-2 infection[17]. In this study, we performed LCMM in lcmm R package[18]. Both linear and non-linear functions (beta, 3-quant-splines, 5-quant-splines, and custom splines) were performed and compared. Model with lower Bayesian information criterion (BIC) value was selected. Based on posterior probability of ≥70%, patients were assigned to the class with the highest probability. To compare the characteristics of patients between identified classes, bivariable logistic regression models were performed in SPSS 25[19]. In light of the potential association of age and chronic illnesses, we included the variable of age 60 or above as a confounder in multivariable logistic regression models. Multivariable multinominal logistic regression models were performed to compare more than two classes. Complete-case analyses were performed.