Our study is a cross-sectional survey-based observational to determine the difference in demographic and sport-specific characteristics, disease severity, duration, and symptoms, self-reported approach to RTS and exercise capacity, and fatigue score between athletes who had returned to their pre-illness level of sport and those who had not. The impetus for our study was the threat of COVID-19 consequences in athletes over the age of 35 years, doing a high volume of cardiovascular exercise. Participants were recruited by snowball sampling from athletic clubs in South Africa between December 2021 and July 2022. Emailers were sent to 64 athletic clubs containing the invitation to participate in the online survey and a request to forward it on to their members. The invitation was also disseminated via social media to 18 WhatsApp groups and 25 Facebook pages. Athletic clubs approved the post on their social media pages and chats. Invitations could then be forwarded by the recipients in a snowball fashion to fellow athletes.

Eligible participants were adult (age 18+) endurance athletes of both sexes who ran for three or more hours per week or cycled for three or more hours per week (McKinney et al. 2019). Those who participated in additional sports over and above were also included and described as multisport athletes. The athletes had either been diagnosed with COVID-19 through a positive polymerase chain reaction (PCR) test or exhibited the known symptoms of COVID-19. Athletes who were admitted to the intensive care unit (ICU) were excluded as ICU-acquired weakness occurs from 15–29 days, and necessitates prolonged rehabilitation to return to normal function (Van Aerde et al. 2020). Other exclusion criteria for athletes were not being proficient in the English language, residing outside of South Africa, and a lack of informed consent. The sample size estimation equalled 178. This was based on the change in aerobic capacity of the athletic population infected with COVID-19 compared to a control group (10.0% vs. 3.5%) (Crameri et al. 2020), with an effect size of 0.63, a confidence interval (CI) of 0.05, power of 0.9, and adding 10% for each confounding variable that is age, sex, sport, stage of disease. To achieve this sample size, we assumed a 30% response rate and 50 members per athletics club. At the time, the prevalence of COVID-19 infections in the city was 5.4%, thus, the questionnaire needed to be accessed by 3200 athletes at 64 clubs.

A self-administered web-based questionnaire was developed by the research team using the Research Electronic Data Capture (REDCap) platform and was based on current literature. Sociodemographic data including sex, age, and sport-specific data constituted the first 10 questions, followed by 10 questions on COVID-19 status, prevalence of comorbidities, and post-COVID sequelae. The third section included questions regarding the management of their acute illness based on the World Health Organization (WHO) Severity Scale (Marshall et al. 2020) and symptoms classified according to the National Institutes of Health (NIH) clinical spectrum (COVID-19 Treatment Guidelines, 2023). Symptom duration and time to return to their pre-illness level of sport were categorised to align with the NICE definitions (NICE 2021). The change in exercise tolerance was based on self-reported pace and power values.

Fatigue was scored using the standardised Multidimensional Fatigue Inventory (MDFI) tool (Smets et al. 1995). The MDFI consists of 20 questions that assess five categories of fatigue, namely: general fatigue, physical fatigue, reduced activity, reduced motivation, and mental fatigue. The scales used are freely available, thus no permissions were required. The MDFI has good internal consistency (Cronbach’s alpha = 0.84), and construct and convergent validity are reported (Smets et al. 1995). A professional panel in the field of sports medicine and respiratory health validated our study questionnaire. Construct validity of the questionnaire as a research tool was performed by sending a test questionnaire to 10 COVID-19 convalescent athletes, 4 of whom were medical professionals. Test-retest reliability was performed 2 weeks apart to avoid a learning effect. The null hypothesis of this study was that there was no difference in the effect of COVID-19 on the group that had returned to their pre-illness level of sports participation (y-RTS) versus those that had not (n-RTS). The primary outcome was a return to the pre-illness level of sport. Therefore, in the statistical analysis, we abbreviated the subgroup of athletes who selected the option ‘still not returned to pre-illness level of sport’ in the questionnaire as the n-RTS group. All others were combined into the y-RTS group.

A lack of consent, answering ‘No’ to contracting COVID-19, or incomplete symptom data resulted in exclusion from data analysis. There was minimal missing data as radio button questions were marked *required. Data were extracted from REDCap into Microsoft Excel and imported into SPSS Statistics for Mac 28.0.1.1.(14) (IBM, Chicago, IL, United States [US]), which was used for all analyses. Descriptive characteristics were calculated using means and standard deviations for continuous variables, and percentages and frequencies for categorical variables for the total sample (All), the group that had returned to pre-illness level of sport (y-RTS), and the group that had not (n-RTS). The normality of distribution of continuous variables was assessed with the Shapiro–Wilk test. Independent t-tests were used to compare normally distributed variables and Mann–Whitney U test for parametric continuous data. Chi-square or Fisher’s exact tests were used to compare categorical data, with Cochran trend for ‘symptom duration’. Univariate and multivariate logistic regression was performed to evaluate the risk of not returning to pre-illness level of sport associated with the following variables: sex, age, general fatigue, physical fatigue, reduced activity, mental fatigue, reduced motivation, and exertional dyspnoea. Because of a strong association of general fatigue, physical fatigue and dyspnoea, only physical fatigue was included in the multivariate model. Model assumptions including goodness of fit, removal of residuals and collinearity were met.

Unconditional ethical approval was provided by the University of the Witwatersrand’s Human Research Ethics Committee (Medical) (Certificate no. M210657). Our study procedures adhered to the principles set forth in the Declaration of Helsinki and the South African Guidelines for Good Clinical Practice and the Medical Research Council. An information sheet explaining our study purpose and data sharing according to the FAIR principles (GO FAIR, FAIR principles 2016) was linked to the survey and ensured confidentiality and anonymity and permission to withdraw from our study at any point. The participants were asked if they consented to the anonymised data being used for research purposes. After electronic consent was obtained, the full questionnaire opened. The completion of the questions implied that consent was given throughout. Only the research team had access to the REDCap platform on which the information sheet and survey was developed and data were stored. The exported data were stored one password-protected computer and backed up to Apple iCloud Drive.

After the surveys were assessed for eligibility, 308 participants were included in data analysis (Figure 1). Table 1 summarises our study population characteristics in three groups: all 308 participants, 211 (68.5%) who returned to pre-illness level of sport (y-RTS), and 97 (31.5%) who had not yet returned (n-RTS). The mean age was 44.9 + 10.2 years, and 170 (55.2%) were male. There is no significant difference in the demographic and sport-specific characteristics of y-RTS versus n-RTS.

Most athletes did not have coexisting comorbidities (n = 253, 82.1%). However, comorbidities that were reported include asthma (n = 18, 5.8%), hypertension (n = 12, 3.8%), and cardiac disease (n = 4, 1.3%). The majority (94%) of athletes (n = 289) experienced one or more of the known COVID-19 symptoms. The only significant difference in COVID-19 characteristics of the y-RTS group compared to the n-RTS was the severity of the disease according to the WHO COVID-19 Severity Scale, indicating that more athletes consulted healthcare professionals and/or were admitted to hospital in the n-RTS group (p = 0.032).

Table 2 summarises symptoms at the time of illness. There was no significant difference in asymptomatic disease experienced by the y-RTS versus n-RTS groups. A total of 21 participants (6.8%) tested negative, but their symptom presentation implied false negative tests. Fatigue was the only mild symptom more prevalent in the n-RTS group (p = 0.044). We found a higher proportion of moderate disease in n-RTS group (p = 0.006). There were eight athletes in total with severe illness requiring supplemental oxygen, with no significant difference between the groups (p = 0.710). No participants cited critical illness. Post-COVID-19 sequelae included cardiac (n = 15, 4.9%) and pulmonary (n = 12, 3.9%) consequences, and were more prevalent in the n-RTS group (p = 0.031).

The time frame in which athletes in the whole group returned to pre-illness level of sport is displayed in Figure 2, with n = 97 (31.5%) not yet at their previous level and constituted the n-RTS group. Table 3 displays the presence and severity of symptoms while exercising and other self-reported factors limiting their RTS. The athletes approached their resumption of training with similar consideration of most factors, with only pace being monitored more closely in the n-RTS group (p = 0.035). There was a distinctly higher proportion (n = 27, 27.8%) of athletes still experiencing symptoms at rest in the n-RTS group. The most frequently reported exertional symptoms by the whole group were fatigue (n = 216, 70.1%), and elevated heart rate (n = 216, 70.1%), followed by dyspnoea (n = 174, 56.5%). All symptoms except musculoskeletal pain occurred significantly more frequently in the n-RTS group. All self-reported limitations to exercise capacity were more common in the n-RTS group (p < 0.001). Training volume (p < 0.001), running pace (p < 0.001), and cycling threshold power (p = 0.024) were lower in the n-RTS group compared to the y-RTS group.

The MDFI (Table 4) assesses five categories of fatigue (general fatigue, physical fatigue, reduced activity, reduced motivation and mental fatigue). Athletes who had not yet returned to pre-illness level scored higher in all fatigue categories, indicating considerably worse overall fatigue compared to the y-RTS group (p < 0.001). The results of the univariate regression analysis in Table 5 suggest all variables excluding age and sex are associated with RTS. The multivariate regression analysis adjusted for dyspnoea and general fatigue, showed that an increase in physical fatigue is associated with decreased odds of return to sport (OR 0.673, p < 0.001). An increase in mental fatigue is associated with increased odds of return to sport (OR 1.0361, p = 0.047).

The main strength of this study is the sample size and statistical power. Athletes were recruited throughout South Africa, which enhances generalisability to the South African athletic population, although generalisation may not be translated to other countries. The focus of our study was endurance athletes and therefore mostly older than 35 years of age, fulfilling a gap in the literature. Quantifying subjective descriptions of fatigue by using the MDFI is a strength of our study as to our knowledge, it has not been used in other studies on athletic populations. Future studies using the MDFI in athletes is recommended. This is one of few studies that has associated exertional symptoms with RTS status.

As is customary for surveys, certain limitations exist. Snowball sampling will not allow a response rate to be calculated and recorded, thus reducing the power of our study. The participant’s responses rely on retrospective recall (recall bias) and the period from illness to the time of our study will vary between participants. The inclusion criteria did not specify the presence of COVID-19 symptoms, thus selection bias caused underrepresentation of the number of asymptomatic cases. Vaccine status was unknown; however, by deducing from the time of illness, most athletes would not have had the opportunity to be vaccinated. Athletes in our study would have been infected by a mix of viral strains. We did not ask whether home oxygen was used, which may have placed some participants not admitted to the hospital in the ‘severe illness’ category. These limitations were unavoidable with the current knowledge at the time of our study. No direct cause-effect relationship can be determined from our observational study.

These findings may provide valuable insights into the acute and long-term manifestations of this disease in a masters-age athlete population. Symptoms that may indicate cardiopulmonary consequences in an athlete population that does a high volume of cardiovascular exercise should be investigated. An interdisciplinary approach to management is vital to facilitate return to sport and the important mental and physical benefits thereof. This will augment outcomes after respiratory tract infections in endurance athletes.