Longitudinal profile of antibody response to SARS-CoV-2 in patients with COVID-19 in a setting from Sub–Saharan Africa: A prospective longitudinal study

Background Serological testing for SARS-CoV-2 plays an important role for epidemiological studies, in aiding the diagnosis of COVID-19, and assess vaccine responses. Little is known on dynamics of SARS-CoV-2 serology in African settings. Here, we aimed to characterize the longitudinal antibody response profile to SARS-CoV-2 in Ethiopia. Methods In this prospective study, a total of 102 PCR-confirmed COVID-19 patients were enrolled. We obtained 802 plasma samples collected serially. SARS-CoV-2 antibodies were determined using four lateral flow immune-assays (LFIAs), and an electrochemiluminescent immunoassay. We determined longitudinal antibody response to SARS-CoV-2 as well as seroconversion dynamics. Results Serological positivity rate ranged between 12%-91%, depending on timing after symptom onset. There was no difference in positivity rate between severe and non-severe COVID-19 cases. The specificity ranged between 90%-97%. Agreement between different assays ranged between 84%-92%. The estimated positive predictive value (PPV) for IgM or IgG in a scenario with seroprevalence at 5% varies from 33% to 58%. Nonetheless, when the population seroprevalence increases to 25% and 50%, there is a corresponding increases in the estimated PPVs. The estimated negative-predictive value (NPV) in a low seroprevalence scenario (5%) is high (>99%). However, the estimated NPV in a high seroprevalence scenario (50%) for IgM or IgG is reduced significantly to 80% to 85%. Overall, 28/102 (27.5%) seroconverted by one or more assays tested, within a median time of 11 (IQR: 9–15) days post symptom onset. The median seroconversion time among symptomatic cases tended to be shorter when compared to asymptomatic patients [9 (IQR: 6–11) vs. 15 (IQR: 13–21) days; p = 0.002]. Overall, seroconversion reached 100% 5.5 weeks after the onset of symptoms. Notably, of the remaining 74 COVID-19 patients included in the cohort, 64 (62.8%) were positive for antibody at the time of enrollment, and 10 (9.8%) patients failed to mount a detectable antibody response by any of the assays tested during follow-up. Conclusions Longitudinal assessment of antibody response in African COVID-19 patients revealed heterogeneous responses. This underscores the need for a comprehensive evaluation of seroassays before implementation. Factors associated with failure to seroconvert needs further research.

Introduction A new coronavirus, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), has recently emerged leading to a global pandemic [1]. Although molecular tests remain the mainstay diagnostic tool for identifying SARS-CoV-2 infected patients [2,3], their use is limited due to multitude of factors. Molecular tests require complex laboratories and highly trained expertise, creating a challenge for their use in resource-constrained settings. In addition, the sensitivity of molecular assays for SARS-CoV-2 is variable, depending on sample type, disease severity and sampling time after onset of infection [3]. To overcome some of these obstacles, serological tests, including rapid diagnostic tests (RDTs) for use at Point-of-Care (PoC), have emerged as an important alternative. Serological testing for SARS-CoV-2 plays an important role for epidemiological studies and in aiding the diagnosis of (previous) COVID- 19 in PCRnegative patients [4][5][6][7]. In addition, serology testing may be important in the assessment of convalescent plasma, and also in the identification, and quantification of vaccine responses [8,9]. Several reports have shown that median seroconversion time is 7-14 days after symptom onset [7,[10][11][12]. In addition, it has been demonstrated that antibody levels may correlate with COVID-19 severity [11,12].
Most of the available data on the longitudinal profile of antibody responses are from high income countries (HICs), while little is known from low and medium income countries (LMICs), in particular from Sub-Saharan Africa (SSA) [13,14]. In addition, applied SARS--CoV-2 serological tests evaluated in HICs might experience significantly higher cross-reactivity in SSA that may impact on the specificity of the assays, as background infectious diseases are entirely different [15,16]. Therefore, there is a need for quality assessment of antibody responses to SARS-CoV-2 in the context of SSA. Such tests are needed urgently to scale-up SARS-CoV-2 surveillance, and diagnosis in LMICs, including in SSA.
In this study, we aimed to determine the longitudinal antibody response of SARS-CoV-2 using different assays in SSA context. In addition, we assessed seroconversion dynamics, and characteristics of different anti-SARS-CoV-2 antibodies. To the best of our knowledge, this study is the first prospective study to evaluate the kinetics of anti-SARS-CoV-2 antibody response, as well as seroconversion patterns among COVID-19 patients in SSA.

Study population
The study is part of the ongoing Profile-Cov study, a prospective observational cohort study in Ethiopia with rapid diagnostic profiling of SARS-CoV-2 in the context of persistent immune activation (Clinicaltrials.gov: NCT04473365). Consecutive patients with confirmed real-time polymerase chain reaction (RT-PCR) test result were enrolled. They were recruited from Mekelle University College of Health Sciences (Kuyha COVID-19 Isolation and Treatment Center), Mekelle City, Northern Ethiopia, in a prospective fashion between July 15 and October 28,2020. Individuals presenting to the Isolation and Treatment Center were screened for SARS-CoV-2 infection with a naso-pharyngeal swab and a RT-PCR. Following the declaration by the WHO that COVID-19 has become pandemic, the Ethiopian Ministry of Health implemented a mass screening of all travelers, people who had come in contact with COVID-19 patients, those from high-risk settings (e.g. health-care workers), as well as those with symptoms and signs suggestive of SARS-CoV-2 infection [17]. All cases with a confirmed SARS--CoV-2 infection were admitted to dedicated COVID-19 Isolation and Treatment Centers. Patients were admitted irrespective of clinical severity status. Whenever patients progress to severe or critical COVID-19, they were admitted to the intensive care unit (ICU).
Sociodemographic, clinical data and laboratory data were collected using standardized Case Record Forms (CRFs) adapted from the International Severe Acute Respiratory and Emerging Infection Consortium's (ISARIC) CRFs for emerging severe acute respiratory infections [18]. Patient's clinical status was stratified following WHO criteria as asymptomatic, mild/moderate, severe (with dyspnea, respiratory rate � 30 breaths per minute, O 2 saturation � 93%, lung infiltrates � 50% of the lung fields within 24-48 hours), and critical (with respiratory failure, septic shock, and/or multiple organ failure) [19]. For this study, asymptomatic and mild/moderate cases were considered as non-severe cases and both severe and critical were considered as severe cases. All data were then entered onto electronic medical records.
For determining specificity, RT-PCR negative specimens (n = 100) were obtained during COVID-19 pandemic (August and September 2020) from individuals not suspected of SARS--CoV-2infection. In addition, pre-pandemic specimens (n = 50) collected in 2017 from patients with other infections were included as negative controls in the analysis of specificity.

Sample size
The sample size required to assess kinetics of antibody response to SARS-CoV-2 using serology assay, based on an expected sensitivity (and specificity) of 80% and a desired error margin of ±10% and an alpha level of 5%, was estimated to be 62 COVID-19 cases and 62 non COVID-19 controls. To take into account prevalence variability (e.g. 80%), we aimed to recruit a minimum of 150 (75 per group) individuals.

Laboratory assays
SARS-CoV-2infection was confirmed by RT-PCR on samples obtained from nasopharyngeal swabs, as described previously [17]. The swabs collected were put in virus transport media and sent immediately to the central laboratory. RNA was extracted using the MagNA Pure 96 (Roche, Germany). RT-PCR for the detection of SARS-CoV-2infection was based on the LightMix1 Modular SARS and Wuhan CoV E-gene kit (TIB Molbiol, Berlin, Germany). Briefly, a 25 μL reaction contains 5 μL of RNA, 12.5 μL of 2 × reaction buffer provided with the Superscript III one step RT-PCR system with Platinum Taq Polymerase (LightCycler Multiplex RNA Virus Master (Roche) containing 0.4 mM of each deoxyribonucleotide triphosphates (dNTP) and 3.2 mM magnesium sulphate), 1 μL of reverse transcriptase/Taq mixture from the kit, 0.4 μL of a 50 mM magnesium sulphate solution (TaqMan Fast Virus 1-Step Master Mix (Thermo Fisher)), and 1 μg of nonacetylated bovine serum albumin (Roche). Thermal cycling was performed at 55˚C for 10 min for reverse transcription, followed by 95˚C for 3 min and then 45 cycles of 95˚C for 15s, 58˚C for 30s. We used the thermocycler from Applied Biosystems ViiA7 thermocycler (Applied Biosystems, Hong Kong, China).
Five different commercial kits were studied (S1 Table). These included four lateral flow immune assays (LFIAs), i.e. Canea, Cellex, Innovita, and Vivachek that target the nuclear (N) and spike (S) proteins of SARS-CoV-2, and the Roche elecsys electrochemiluminescent immunoassay (ECLIA) that targets the nucleocapsid protein (NP). Whereas the LFIAs are designed to detect IgM and IgG against SARS-CoV-2, the Roche elecsys assay detects anti-SARS-CoV-2 total antibody levels. The assays were selected based on their availability in the local market as well as performance data. Testing was undertaken following manufacturer's instructions for each assay. For the LFIAs, testing involved the addition of 10 μL of plasma to the sample well, and 80-100 μL of buffer to an adjacent well, followed by 15 minutes incubation at room temperature. The result was based on the appearance of colored bands, interpreted as positive when control and test bands were visible, as negative when only the control band was visible, or as invalid when the control band was not visible. All results were read by two technologists independently and photographed in real time. In the event of a discrepant report, reading by a third technologist served as a tie breaker. The Roche elecsys ECLIA was performed using the Roche Cobas1 e411 immunoassay analyzer. Anti-SARS-CoV-2 total antibody levels were expressed as cut-off index (COI), and a value of � 1.0 was considered as a positive result as per the manufacturer's instructions. The individuals running the serology assays were blinded as to the origin of the sample, clinical data of the study participants, or to the reference SARS--CoV-2 results.
Seroconversion was ascertained if a patient with a negative antibody test at time of enrollment becomes positive with any of the assays tested during follow-up. For estimating the day of seroconversion, the days post symptom onset were determined for symptomatic cases. For asymptomatic patients, we added 6 days (median time between symptom onset and date of positive PCR testing among symptomatic cases) to the PCR testing date.

Statistical analysis
Baseline characteristics for continuous variables were expressed as the median with IQR, and for categorical variables as proportions. Whereas categorical variables were compared using Fisher's exact test or χ 2 test, continuous variables were compared by Mann-Whitney U or Kruskal-Wallis tests as appropriate. The primary outcome for this study was the proportion of positivity rates (sensitivity) of the various serology tests, stratified by Ig isotypes (IgM, or IgG, or total Ig), and according to sampling time (days after symptom onset) among specimens derived from RT-PCR confirmed COVID-19 cases. Specificity was determined among RT-PCR negative, or pre-COVID specimens. Positivity and specificity rates were presented as proportion and 95% confidence interval (CI) limits. Secondary analysis included performance characteristics of the assays stratified by the patient's clinical severity status. Positive predictive values (PPV) and negative predictive values (NPV) were determined taking into consideration a population prevalence of 5%, 25%, and 50% SARS-CoV-2 infection. Agreement between different assays was determined by Kappa statistics. P values <0.05 were considered statistically significant. Reporting of the diagnostic performance of the assays in this study was according to the STARD (Standards for Reporting of Diagnostic Accuracy Studies) guidelines. Data were analyzed using STATA (Statistical package v. 14.0, StataCorp, Texas, USA).

Ethical considerations
Participants provided written informed consent to participate in the Profile-Cov study. The study protocol was reviewed and approved by the Health Research Ethics Review Committee (HRERC) of Mekelle University College of Health Sciences (#ERC 1769/2020). Written informed consent was obtained from all study participants. However, specimens used as controls from pre-pandemic period were specimens from biobank and individual study participant consent was not obtained. Nonetheless, all personal identifiers were de-linked from the original sources, and the HRERC has reviewed and approved it.

Characteristics of study participants
We initially enrolled a total of 120 patients. However, 18 were lost to follow-up, the majority were lost after their SARS-CoV-2 PCR test results, and few had only � 1 sample, and we were not able to draw sufficient longitudinal serial specimens. Thus a total of 102 patients were eventually included, and detailed clinical characteristics of the enrolled participants is provided in Table 1. The majority were male (76.5%), with a median age of 33 (IQR: 27-42) years old, and 12.8% were aged � 60 years. There was no difference in age between male and female study participants. Whereas 32/102 (31.4%) were asymptomatic, the remaining 70/102 (68.6%) were symptomatic (36 had mild/moderate and 34 had severe/critical clinical presentation). COVID-19 patients presenting with severe clinical status were older [median age 50 years (IQR: 38-66) vs. 29 years (IQR: 26-37); p<0.00001]. In addition, severe COVID-19 patients manifested with a significantly higher frequency of symptoms when compared to those with non-severe COVID-19. Patients with severe disease symptoms when compared to non-severe cases were more febrile, dyspneic, had more cough, chest pain, and head ache. The frequency of sore throat, nasal congestion, loss of smell and/or taste, diarrhea and myalgia were not different between both patient groups. Patients with severe COVID-19 were significantly more likely to be tachypneic and tachycardic when compared to those with non-severe disease. In addition, lymphocyte and hematocrit counts were significantly lower among severe COVID-19 patients compared to those with nonsevere clinical manifestations. Comorbidities, including non-communicable diseases, were significantly higher among severe COVID-19 cases when compared to the non-severe. In addition, all severe cases of COVID-19 patients were admitted to the intensive care unit, received supplemental oxygen therapy, or mechanical ventilation. Finally, mortality was significantly higher among severe cases of COVID-19 patients.

Longitudinal antibody response to SARS-CoV-2 and performance of serological assays
The kinetics profile of antibody response is summarized in Figs 1-3. Serology testing was determined longitudinally in 102 patients, who provided a total of 802 plasma specimens  (Fig 2A). During the period 2-to 3-weeks after the onset of symptoms, the positivity rate of all the assays was increased, ranging between 67.3% and 81.3%. Between 3 and 6 weeks after the onset of symptoms, the positivity rate continued to increase for all the assays, ranging from 84.6% to 90.9%. The positivity rate for IgM, IgG as well as combined IgM or IgG was not different for all of the assays assessed. Overall, longitudinal analysis revealed a heterogenous antibody responses profile to all assays tested (Figs  1 and 2B and S2 Table). In addition, we noted a trend towards a higher positivity rate of almost all assays among patients with symptomatic vs. asymptomatic (Fig 3), or severe COVID-19 compared to those with non-severe clinical presentation during the early phase of the infection (S1 Fig). However, the differences were not statistically significant. Notably, very few of the asymptomatic SARS-CoV-2 patients were positive to IgM by Cellex immunoassay (Fig 3).
The specificity of the assays was evaluated using samples from RT-PCR negative individuals (n = 100) derived during the same enrollment period of the COVID-19 patients. In addition, we included pre-COVID-19 samples (n = 50). The results are summarized in   Table. https://doi.org/10.1371/journal.pone.0263627.g001

PLOS ONE
Longitudinal antibody response to SARS-CoV-2 in Africa 100.0% (95CI: 92.75-100.00) for IgM, 100.0% (95CI: 92.75-100.00) for IgG, and from 96.0% (95% CI: 84.72-99.05) to 100.0% (95CI: 92.75-100.00) for IgM or IgG antibodies for samples collected during pre-pandemic period. For the Roche elecsys total Ig assay, the specificity was 94% (95% CI: 82.39-98.13) and 100.0% (95% CI: 92.75-100.00), for RT-PCR negative, and pre-COVID specimens, respectively ( Table 2 and Fig 2C). Overall, 17/150 (11.3%) RT-PCR negative samples were found to be reactive in either one or more of the assays tested. Whereas 9 specimens reacted with only one assay, 6 were reactive with 4 different assays, 1 with 2 different assays, and 1 was reactive with 3 different assays. Only 2 samples appeared positive from those samples collected during the pre-COVID-19 pandemic period. The agreement among the different serology assays evaluated, and considering the results of combined IgM and IgG, was very good, ranging between 83.8% and 92.4% (Fig 4A). In addition, there was a direct correlation between the LFIAs test results and the Roche's elecsys anti-NP total Ig titers (Fig 4B). Given that PPVs and NPVs are influenced by population prevalence of SARS-CoV-2 infection in a given time [20], we estimated the predictive values of five different assays, taking into consideration of population seroprevalences of 5%, 25% and 50% SARS-CoV-2infection (Table 3). Overall, the estimated PPV for IgM or IgG, or anti-NP total Ig in a scenario with low seroprevalence at 5% varies from 32.9% to 57.9%. Nonetheless, when the population seroprevalence increases to 25% and 50%, there is a corresponding increases in the estimated PPVs. The estimated NPV for IgM or IgG, or anti-NP total Ig in a low seroprevalence scenario (5%) is high, ranging between 98.7% and 99.1%. However, the estimated NPV in a high seroprevalence scenario (50%) is reduced significantly to 79.7% to 84.8%. Examples of a hypothetical scenario of the testing outcomes for 1000 people tested using the different assays we evaluated in a setting with a population seroprevalence of SARS-CoV-2 ranging from 5% to 50% is summarized in S3 Table. For example, in a population with a hypothetical seroprevalence of 5%, and using Cellex-an assay with a sensitivity of 83% would have 8 results per 1000 tested classified incorrectly as false negative. In addition, at a seroprevalence of 5%, this same assay with a specificity of 97% would have 28 out of 1000 tested classified incorrectly as false positive. However, at a seroprevalence of 50%, the same assay with the same sensitivity of 83% would provide 85 results for every 1000 cases tested classified incorrectly as false negative, and the same assay with the same specificity of 97% would result in only 15 cases for every 1000 tested that would be classified incorrectly as false positive. Overall, with increased population seroprevalence, PPV and false negative increase, and NPV and false positive decrease [20].

Seroconversion of SARS-CoV-2 antibody
Patients who were antibody-negative at time of hospital admission, but who became antibody-positive during subsequent follow-up were considered as sero-converters. Overall, a total of 28 (27.5%) sero-converters were identified, 64 (62.8%) had already a positive antibody test result at time of admission, and 10 (9.8%) failed to mount a detectable antibody response by any of the assays tested throughout the course of follow-up. Seroconversion pattern as determined by different assays using various antibody isotypes are shown in Fig 5. Patients started to seroconvert as early as 4 to 6 days after onset of symptoms, reaching between 18.2% and 45.5% for the LIFAs IgM or IgG, and 14.3% for the Roche elecsys total Ig assay. The rate of seroconversion within the first 15 days after the onset of symptoms increased to 53.8% and 69.2% for the LFIAs IgM or IgG, and to 47.1% for the Roche elecsys assay. In addition, the seroconversion rate of all assays increased further during 2-to 3-weeks after onset of symptoms, ranging between 58.8% and 79.2%. The seroconversion rate continued to increase for all the assays, reaching 100% after 5.5 weeks after onset of symptoms. Nonetheless, seroconversion rate for Cellex IgM remained low throughout the course of follow-up. Overall, the median time for seroconversion was 11 (IQR: 9-15) days post symptom onset. The median time of seroconversion in the patients with severe COVID-19 tended to be shorter [10 (IQR: 8-11) days] when compared to non-severe COVID-19 patients [13 (IQR: 9-16) days]. However, this difference was not statistically significant (p = 0.251). Of the 28 patients who seroconverted, 11 (39.3%) were asymptomatic and 17 (60.7%) symptomatic. When stratified by symptoms, the median seroconversion time among symptomatic cases tended to be significantly shorter compared to asymptomatic patients [9 (IQR: 6-11) vs. 15 (IQR: 13-21) days; p = 0.002] (Fig 2D).
Of the remaining 74 COVID-19 patients included in the cohort, 10 (9.8%) had no detectable antibody response by each of the assay tested. These patients who failed to seroconvert included 5 mild/moderate, and 5 severe cases, with a median of 28 (IQR: 16-33) days followup after symptom onset (detailed information available in S4 Table).

Discussion
Evaluation of serological tests for SARS-CoV-2 infection has already been described by several reports from HICs [7][8][9][10][11][12][13]. Nonetheless, reports on the evaluation of serology tests on specimens derived from longitudinally enrolled patients from LMICs is scarce [14]. Such evaluations are important since blood samples from LMICs are markedly different from HICs with different immune activation status, genetic backgrounds, and pathogen diversity, and load as compared to HICs. In this work, we assessed serology assays using specimens obtained in a LMIC setting (Ethiopia) from consecutively enrolled COVID-1 9 patients. Thus, we were able to determine the kinetics profile of antibody responses of individual patients, as well as seroconversion patterns. In addition, we were able to determine the performance of different serological assays in LMIC context.
The findings revealed that antibody responses were highly heterogeneous between individuals and assays, consistent with previous reports [21][22][23][24][25][26][27][28]. Comparable trends in antibody response kinetics observed between the different assays suggest that the different antigen targets (NP, N and S) were similarly immunogenic. Longitudinal kinetics assessment of the various serological assays revealed that positivity rate increased with increasing time after the onset of symptoms. Around 80% positivity rate was reached after 3 weeks of symptom onset. This finding concurs with the only other report existing, from Cameroon [14]. A positivity rate that is sufficient to identify all COVID-19 patients may require longer period of time after the onset of disease. This suggests that the mainstay of diagnosis of COVID-19 remains the detection of virus or virus products [2,3]. There are conflicting reports regarding differences in positivity rate between non-severe and severe COVID-19 cases. Whereas some reports revealed that positivity rate among those with severe COVID-19 disease is significantly higher compared to those with non-severe disease [11,12], others have demonstrated that there is no significant difference between the two groups [26]. Interestingly, it was shown that COVID-19 patients with more severe clinical presentation indeed exhibit a delayed production of antibodies compared to non-severe cases [22,27]. In the current study, though there was a tendency towards increased positivity rate among severe compared to non-severe cases, we did not observe a significantly different pattern. In a previous report, we noted that being male is associated with COVID-19 severity [17]. However, in the current study, we did not observe differences in COVID-19 severity neither response to SARS-CoV-2 by gender, probably due to the smaller sample size. Though typically IgM appears earlier than IgG in other infectious events, this pattern is not typical in SARS-CoV-2 infection, as shown in the current study, as well as by previous investigators [26,29], perhaps due to the fact that SARS-CoV-2 targets the lung epithelium initially. Furthermore, positivity rate for the assay Cellex IgM was found to be very low in this study, which is similar to previous report [30]. In contrast, others have reported higher positivity rate, ranging between 55.4% and 87.5% [22,31]. The differences between the studies might be attributed to differences in sampling time, clinical severity status and study population immune responses.
The findings of this study revealed specificity of various serological assays tested > 90% for LFIAs and 97% for Roche elecsys. The findings that samples were reactive in particular to more than one assay may indicate the presence of cross-reacting antibodies, as has been reported in other African countries [15,16]. Moreover, the possibility of false-negative RT-PCR test result should be considered, in particular for those samples collected during the COVID-19 pandemic. This notion is strengthened by the fact that in our cohort 16/17 samples that become reactive to serology were collected during the COVID-19 pandemic period. Thus, future in-depth evaluation of serological assays needs to consider such possibilities.
The positivity rate among the different assays that we tested was variable. The estimated PPV in a population with a seroprevalence of 5% was between 33% and 58%. Hence, an assay with high specificity is required to ascertain an acceptable PPV in a population with low seroprevalence [20]. Nonetheless, the estimated PPV increases significantly to over 90% in a population with seroprevalence of 50%. In such scenario, an assay that is needed is the one with a high sensitivity [20]. Overall, estimated PPV and NPV of each of the assay in a setting with a diverse population seroprevalence scenarios would certainly impact on the diagnostic testing outcomes. This needs to be considered during serosurveys [20].
Several reports have documented that the median time for IgM and IgG seroconversion is 7-14 days after the onset of symptoms [7,[10][11][12]. This is consistent with our finding showing a median seroconversion time of 11 days after onset of symptoms. Overall, these findings reflect the lower sensitivity of serological assays during the acute phase of the disease. In the current study, though there was a tendency towards earlier seroconversion rate among severe compared to non-severe cases, we did not observe a significantly different pattern. Nonetheless, when stratified by symptoms, those with a symptomatic clinical presentation seroconverted earlier when compared with those with asymptomatic presentation. Notably, we identified a total of 10 (9.8%) RT-PCR confirmed COVID-19 patients from our cohort who failed to seroconvert by any of the assays we tested. All these patients were symptomatic, and it is unlikely that these patients may have false-positive RT-PCR results. Except one case with HIV-1, none of the patients had immunosuppression. The number of days after onset of symptoms in two cases was < 5 days, and these patients might have seroconverted later. Our findings are somewhat higher compared to a previous report that showed that approximately 5% of symptomatic PCR-positive patients remain antibody negative [27], but lower than the one reported by Long et al showing lack of seroconversion in > 20% of cases [7]. The fact that some patients fail to mount an adequate level of antibody response indicates that certain serology tests may have suboptimal sensitivity in assessing past infections.
Strengths of the current study include evaluation of longitudinal serial samples from same individuals allowing us to determine the profile of antibody kinetics on individual patients, including exact seroconversion dates in some. In addition, a strength was the inclusion of asymptomatic cases, the analysis by COVID-19 disease severity, and the strict definition of controls (RT-PCR negative specimens obtained during the same period as well as samples collected before COVID-19 pandemic). Limitations associated with the study were the relatively small sample size, lack of evaluation of the neutralizing potential of the anti-SARS-CoV-2 antibodies, difficulty in selection of kits for evaluation, and the use of plasma rather than whole blood specimens to evaluate applicability of LFIAs as PoC tests for field use. In addition, we did not determine long-term outcome of SARS-CoV-2 antibody response several months after the onset of symptoms in the current study. In addition, antibody titer in this study was determined using elecysys total Ig platform that may have precluded the specific determination of antibody response against the spike protein of SARS-CoV-2.
SARS-CoV-2 seroprevalence studies done in African countries with commercial tests validated in Europe, the USA, or Asia is faced with several challenges [32]. Therefore, studies such as the one here contribute to the understanding of the characteristics and pattern of antibody response of COVID-19 patients in the African context. To the best of our knowledge, this study is the first prospective study to evaluate the kinetics of anti-SARS-CoV-2 antibody response evaluating seroconversion pattern among SARS-CoV-2 infected patients from Africa. The study underscores the need for a comprehensive evaluation of commercial assay before use for surveillance or in aiding the diagnostic work up of COVID-19 cases. In addition, our results, together with previous evidence conducted in Africa [33], suggest that combining different serology tests with antigen test may have a role in aiding the diagnosis of COVID-19 in resource-constrained settings. Furthermore, factors associated with failure to mount adequate SARS-CoV-2 antibody response need further research.

Institutional review board statement
The study was reviewed and approved by the Health Research Ethics Review Committee of Mekelle University College of Health Sciences (#ERC 1769/2020).

Informed consent statement
Written informed consent was obtained from all study participants. However, specimens used as controls from pre-pandemic period were specimens from biobank and individual study participant consent was not obtained. Nonetheless, all personal identifiers were de-linked from the original sources.