Epidemiological Application of Cycle Threshold Values Under The COVID-19 Pandemic in Korea

Abstract

Backgrounds Cycle Threshold values [Ct values] are the measured values, the determinants of positivity in a COVID-19 diagnosis. Studies on the usefulness of Ct values for forecasting the number of COVID-19 newly confirmed are being actively conducted worldwide because of their features that are inversely proportional to the virus load, and their association with infectious/infectious power (et al., 2021; Rodríguez et al., 2021). On the other side, Ct values have been used only in the diagnostic field in Korea. Forecasting the number of newly confirmed cases and strategies for responding to health and infectious diseases has relied on the time-varying reproduction number [Rt] from the beginning of the COVID-19 outbreak. Rt indicates how many additional confirmed cases one confirmed patient has created. However, Rt has some limitations. It is challenging to reduce Rt estimation to less than a week (Lin et al., 2022), and its definition and calculation formula require identifying the infection route (Yoo, 2021; Jung, 2020). To improve the forecasting model of COVID-19 newly confirmed in Korea, this study conducted the forecasting with Ct values to supplement the limitations of Rt. As a result of a systematic literature review, there were some differences between the results of static estimation and dynamic estimation. The research objectives are as follows. First, Rt and Ct values were compared through the Ct values time series distribution by COVID-19 pandemic periods and visually verified whether Ct values could forecast the actual new confirmation. Next, time series analysis-based forecasting of the COVID-19 confirmation pattern using the Ct values by COVID-19 pandemic periods was conducted to determine the time difference of several days when predictive power is the highest. This study conducted forecasting by using both static and dynamic estimation methods for comparison of each result. Methods The study period included 489,133 specimens collected for SARS CoV-2 diagnostic testing about two years from February 7, 2020, to December 31, 2021. Following domestic RT-PCR protocol, this study used Ct values RdRp as representative Ct values. Based on the Korea Disease Control and Prevention Agency [KDCA] publication, the periods of the COVID-19 pandemic in Korea were classified into four periods. This study presented the descriptive statistics table and graph for each variable of the diagnostic test results to understand the overall diagnostic test results of the samples included in the analysis. For visual verification before forecasting, the distribution pattern of the Ct values by the COVID-19 pandemic periods at the population group level is expressed as a time series graph along with the Rt and the number of new confirmed cases, respectively. After verifying the limitations of Rt and the utility of Ct values (a forecasting index for new confirmation) through the two distribution graphs, a statistical analysis was conducted to confirm the actual predictive power of Ct values. To ascertain the time difference when Ct values show the highest predictive power to newly confirmed cases, this study conducted both static and dynamic estimation methods with time lags. As a static method, a simple linear regression model was used, and as a dynamic method, the Distributed Lag Model [DLM] was used. An alpha of 0.05 was used for all tests, and all graphs and statistical analyses were performed with STATA (version 17). Results As a result of the descriptive statistics, most of the diagnostic tests were conducted at examination institutions, and more than half of the confirmed patients were in their 20s to 50s. More than half of the diagnostic kit products used were Seegene, the same gene amplification equipment was used in almost all tests, and more than half of the samples used were throat swabs (NPS/OPS). Ct values E, RdRp, and N all showed similar distribution forms, and values were mainly distributed between 10-20. In the two Ct values distribution pattern graphs, it was visually ascertained that the change in Rt did not sufficiently explain the period when the number of new confirmed cases increased rapidly compared to Ct values. Also, it was ascertained that the Ct values were first affected before the change in the quarantine policy affects the number of confirmed cases. Through these two time-series distribution graphs, it was assumed that it is possible to forecast the occurrence of confirmed patients according to the distribution of Ct values. Static estimation (simple linear regression model) and dynamic estimation (DLM) were used to perform forecasting of the actual occurrence of new confirmed cases using Ct values. First, as a result of the simple linear regression model, the forecasting was the best in the lag of 5-14 days. For forecasting through the simple linear regression model, the time difference with the highest predictive power differed by COVID-19 pandemic periods (11 days of the total period, 9 days of the first period, 5 days of the second period, 14 days of the third period, and 10 days of the fourth period). However, the results were likely to be false regression because it does not assume the normality of the time series. Also, the number of newly confirmed cases is difficult to forecast simply as a result of simultaneous static estimation, it was necessary to eliminate the effects of temporal fluctuations and make forecasts through dynamic estimation. After removing the effect of time fluctuations through DLM, both the third and fourth periods, which have no autocorrelation of the error term, were the best forecasted until the three-day time difference. As a result of the forecast through Rt in the same analysis method, it was best always forecasted at the 8-day time difference. And this study compared the forecast of the number of newly confirmed cases through Ct values and Rt by specifying the time of Omicron spread. However, the analysis period was short for 30 days and it was hard to get significant results because the dominant species period of Omicron was from February 2022. Conclusions This study confirmed the epidemiological utility of forecasting the newly confirmed cases by Ct values, which was rarely used in Korea. A COVID 19 forecast study was conducted using the domestic Ct values distribution at the population group level, including the time when the mutated virus appeared. It is the first domestic study to perform forecasts through Ct values using both static and dynamic models, and it revealed that Ct values can be used as a short-term (3 days) COVID-19 forecast indicator. Based on the results of this study, if various follow-up studies are conducted by combining clinical and policy indicators, the epidemiological utility of Ct values as well as predicting the number of new confirmed cases will be further expanded to contribute to establishing health policies.