Long COVID among the first three waves of COVID-19 in Japan: a multicentre cohort study

Abstract

Objectives

Severe acute respiratory syndrome coronavirus 2 significantly impacts Japan with a high number of infections and deaths reported. Long coronavirus disease (COVID) characterised by persistent symptoms after COVID-19 has gained recognition but varies across studies. This study aimed to investigate the differences in long COVID among patients hospitalised during Japan’s first three waves of the pandemic.

Design

Multicentre prospective cohort study.

Setting

26 medical facilities across Japan between February 2020 and February 2021.

Participants

In total, 1066 hospitalised patients diagnosed with COVID-19 were included with 206, 301 and 559 patients in the first, second and third waves, respectively. Data were collected using electronic data capture and patient-reported outcome forms.

Primary and secondary outcome measures

Long COVID was assessed at 3, 6 and 12 months after COVID-19 diagnosis.

Results

Significant differences were observed between the waves in various baseline and clinical characteristics such as age, body mass index (BMI), comorbidities, the severity of COVID-19, complications and treatment during hospitalisation. Long COVID, particularly dyspnoea, was most prevalent in the first wave. Multivariate logistic regression analysis confirmed a significant positive association between the first wave and long COVID including dyspnoea after adjusting for age, sex, BMI, smoking status and COVID-19 severity.

Conclusions

Patients hospitalised during the first wave had a higher risk of experiencing long COVID, especially dyspnoea, than those hospitalised during the other waves. These findings underscore the need for continued monitoring and managing long COVID in COVID-19 survivors, particularly in those hospitalised during the first wave.

Trial registration number

UMIN000042299.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Various risk factors including age, sex, body mass index, smoking status and severity of COVID-19 have been reported in long coronavirus disease (COVID).

WHAT THIS STUDY ADDS

• Patients hospitalised during the first wave had a significantly higher risk of experiencing one or more long COVID symptoms with dyspnoea being the most persistent and prevalent symptom compared with those hospitalised during the second and third waves in Japan even after adjusting for known risk factors for long COVID.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Continued monitoring and management of long COVID symptoms in COVID-19 survivors are needed, particularly in those hospitalised during the first wave.

Introduction

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has had a significant impact in Japan with 33 802 739 infected individuals and 74 669 deaths reported as of 8 May 2023.¹ While the mortality rate has been decreasing owing to the development of new treatments and the widespread use of vaccines,² 10–30% of patients suffer from various symptoms for weeks to months after COVID-19.³ This condition, known as ‘long coronavirus disease (COVID)’,^{4 5} has been increasingly recognised. Although there are increasing reports on long COVID, the backgrounds of patients including the timing of infection, severity during the acute phase, race and other factors vary resulting in different frequencies and types of long COVID.⁶,⁸

The clinical features of acute COVID-19 vary both globally and regionally.⁹,¹¹ Similarly, in Japan, the clinical characteristics of hospitalised patients varied during COVID-19 outbreaks.¹² Recent reports suggest that long COVID differs depending on the causative strain, for example, the Omicron strain causes milder symptoms than conventional strains.¹³,¹⁷ A recent review reported that long COVID is also influenced by various host factors including dysregulation of multiple body systems such as the immune, nervous and endocrine systems as well as gut microbiota, host genetics, sex and mental health.¹⁸ However, no research in Japan has focused on the differences in long COVID symptoms between each wave of the pandemic and there are no published papers analysing long COVID specifically in relation to the early waves (first three waves). Our previous cross-national prospective study in Japan on long-term COVID among hospitalised patients diagnosed with COVID-19 aimed to reveal the prevalence of and risk factors for long COVID.¹⁹ This study included patients from the first three waves caused by conventional strains, distinguishing them from the dominant mutant strains in the fourth wave and beyond.²⁰

The present study, as a reanalysis of our previous study mentioned above,¹⁹ aimed to clarify the differences in long COVID among Japan’s first three COVID-19 waves, deepen the understanding of patients suffering from long COVID and prepare for long COVID in future pandemics.

Methods

Study design and participants

This study aimed to analyse data from our previous prospective cohort study conducted in Japan between February 2020 and February 2021. The detailed protocol for this project has been published previously.²¹ Patients aged≥18 years, diagnosed with COVID-19 via SARS-CoV-2 PCR or antigen tests and hospitalised in 1 of 26 participating medical facilities were included. Patients who had difficulty understanding informed consent or reading and answering questionnaires because of language impairment, cognitive issues or psychiatric disorders were excluded. A total of 1200 patients were registered with 1066 patients analysed after excluding 134 because of incomplete clinical information obtained from each facility using an electric data capture (EDC) system (online supplemental figure 1).

Procedures

Participating facilities sent information sheets to potential participants for confirmation. The participants were informed of their right to withdraw and skip questions. Clinical information was collected using an EDC system which covered 168 clinical survey items including demographics, comorbidities, hospitalisation details, laboratory and imaging findings at admission, complications and prescribed medications. Participants who provided consent were asked to complete an original questionnaire on paper or on a smartphone application 3, 6 and 12 months after their COVID-19 diagnosis. The questionnaire included a comprehensive survey of 24 symptoms occurring after COVID-19 diagnosis (summarised in online supplemental table 1). Additional symptoms are reported in the Optional comments section. Questionnaires were collected with anonymisation and confidentiality. In this study, individuals who reported having symptoms for at least 3 months after the initial onset were classified as having long COVID. Since SARS-CoV-2 vaccines were introduced in Japan on 12 April 2021, during the fourth wave,¹² no participants in this study had received the vaccine prior to COVID-19 infection. Therefore, the potential impact of preinfection vaccination on long COVID was not considered in this study. This study was registered in the UMIN Clinical Trials Registry. Additionally, approval and permission to conduct the study were obtained from the institutional ethical review committee of each participating institution.

Statistical analysis

The patients’ baseline and clinical characteristics including laboratory findings on admission were summarised as means and SD for continuous variables and as numbers and percentages for categorical variables. The baseline and clinical characteristics of the COVID-19 waves were compared using analysis of variance (ANOVA) for continuous variables and χ² tests for categorical variables. The proportions of patients with one or more sequelae symptoms and those of patients with individual sequelae including dyspnoea between COVID-19 waves were compared using χ² tests at 3, 6 and 12 months after the diagnosis of COVID-19. The proportion of patients with one or more sequelae symptoms or dyspnoea at 3, 6 and 12 months was also compared using χ² tests. The average number of long COVID symptoms at 3, 6 and 12 months after diagnosis was compared among the COVID-19 waves using ANOVA. The association between COVID-19 waves and these outcomes was evaluated using multivariate logistic regression analysis using COVID-19 waves, age, sex, body mass index (BMI), smoking status and oxygen requirement as covariates for the presence of one or more sequelae at all time points, dyspnoea at all time points and dyspnoea at each time point of 3, 6 and 12 months. Covariates other than COVID-19 waves were chosen based on their known association with long COVID in previous large-scale observational studies.^{6 8 22 23} Two-sided p values <0.05 were considered statistically significant. The JMP 16 program (SAS Institute Japan, Tokyo, Japan) and SAS statistical software (V.9.4; Cary, North Carolina, USA) were used for all analyses.

Patient and public involvement statement

Patients or the public were not involved in the development of the research question, study design, recruitment, outcome measures, dissemination plans and conduction of the study.

Results

Patient allocation into three groups based on the COVID-19 waves

Temporal changes in patient enrolment and newly diagnosed COVID-19 cases in Japan revealed similar fluctuations corresponding to the three waves (figure 1): The first wave (29 January 2020 to 13 June 2020), the second wave (14 June 2020 to 9 October 2020) and the third wave (10 October 2020 to 8 February 2021),¹² all of which were caused by conventional strains whereas mutant strains were responsible for the fourth and subsequent waves.²⁰ The groups comprised 206, 301 and 559 patients in the first, second and third waves, respectively.

Figure 1. Number of patients in the present study and newly diagnosed COVID-19 cases in Japan in each COVID-19 wave between January 2020 and March 2021. The first case of COVID-19 was reported in Japan on 16 January 2020. The bar graph represents the number of patients in the present study and the line graph shows the number of patients newly diagnosed with COVID-19 in Japan. The first wave spanned from 29 January 2020 to 13 June 2020, the second wave from 14 June 2020 to 9 October 2020 and the third wave from 10 October 2020 to 28 February 2021. The red, blue and green bar graphs indicate patients diagnosed with COVID-19 during the first, second and third waves, respectively.

Comparison of baseline and clinical characteristics among COVID-19 waves

Significant differences in baseline and clinical characteristics were observed among the waves in many categories (table 1) and laboratory findings (online supplemental table 2). Patients in the third wave were older, had higher BMIs and had a higher prevalence of hypertension, diabetes and chronic kidney disease. They also had higher rates of oxygen therapy, intensive care unit admission, thromboembolism and renal damage during hospitalisation. Conversely, patients in the second wave were younger and had lower rates of hypertension, diabetes, chronic kidney disease, oxygen therapy, intensive care treatment and various complications during hospitalisation. The second wave also included a lower proportion of patients with ground-glass opacities or consolidations on chest imaging. The first wave had the highest proportion of patients requiring intubation with ground-glass opacities, consolidation, bacterial infections, heart failure and liver damage. Treatment patterns showed decreased usage rates of ciclesonide, hydroxychloroquine sulphate and lopinavir/ritonavir in subsequent waves whereas that of remdesivir and steroids increased.

Table 1. Clinical characteristics of COVID-19 patients by infection waves.

	Total(N=1066)	First wave(N=206)	Second wave(N=301)	Third wave(N=559)	P value
Age in years	56.1 (±16.8)	53.4 (±17.4)	52.0 (±16.9)	59.3 (±15.9)	<0.0001
Age group					<0.0001
Young (40≤)	215 (20.2)	56 (27.5)	81 (26.9)	78 (14.0)
Middle-aged (41–64)	489 (46.0)	89 (43.6)	144 (47.8)	256 (45.9)
Older adult (≥65)	359 (33.8)	59 (28.9)	76 (25.3)	224 (40.1)
Sex					0.3747
Male	679 (63.7)	124 (60.2)	189 (62.8)	366 (65.5)
Female	387 (36.3)	82 (39.8)	112 (37.2)	193 (34.5)
BMI	24.2 (±4.1)	23.8 (±4.0)	24.0 (±4.0)	24.5 (±4.1)	0.0415
Smoking status					0.197
Never smoked	504 (47.3)	101 (49.0)	142 (47.2)	261 (46.7)
Ex-smoker	264 (24.8)	50 (24.3)	68 (22.6)	146 (26.1)
Current smoker	118 (11.1)	22 (10.7)	44 (14.6)	52 (9.3)
Comorbidities
Hypertension	344 (32.5)	54 (26.3)	72 (24.0)	218 (39.4)	<0.0001
Diabetes	178 (16.9)	27 (13.2)	36 (12.1)	115 (20.8)	0.0015
Cardiovascular disease	68 (6.4)	10 (4.9)	17 (5.7)	41 (7.4)	0.3768
Malignancy	71 (6.7)	13 (6.4)	14 (4.7)	44 (8.0)	0.1796
Chronic obstructive pulmonary disease	32 (3.0)	6 (2.9)	6 (2.0)	20 (3.6)	0.4185
Asthma	55 (5.2)	14 (6.9)	15 (5.1)	26 (4.7)	0.4867
Hyperuremic disease	108 (10.2)	15 (7.4)	25 (8.4)	68 (12.3)	0.061
Chronic liver disease	35 (3.3)	4 (2.0)	12 (4.0)	19 (3.5)	0.4393
Chronic kidney disease	46 (4.4)	2 (1.0)	12 (4.0)	32 (5.9)	0.0139
Oxygen requirement					<0.0001
With oxygen requirement	343 (32.2)	61 (29.6)	62 (20.6)	220 (39.4)
Without oxygen requirement	697 (65.4)	141 (68.4)	231 (76.7)	325 (58.1)
Intensive care treatment
Admission to ICU	104 (9.9)	19 (9.4)	13 (4.4)	72 (13.2)	0.0002
Intubation	47 (4.5)	18 (8.9)	8 (2.7)	21 (3.9)	0.0027
ECMO	2 (0.2)	1 (0.5)	0 (0)	1 (0.18)	0.4607
Imaging findings on admission
Chest radiography ground-glass opacities	469 (84.4)	102 (96.2)	103 (68.7)	264 (88.0)	<0.0001
Chest radiography consolidation	172 (44.3)	43 (81.1)	30 (26.6)	99 (44.6)	<0.0001
Chest CT ground-glass opacities	650 (92.6)	114 (98.3)	158 (83.6)	378 (95.2)	<0.0001
Chest CT consolidation	272 (61.0)	41 (93.2)	33 (47.1)	119 (63.6)	<0.0001
Complications during the hospital stay
Bacterial infection	62 (5.9)	22 (11.1)	14 (4.7)	26 (4.8)	0.0031
Fungal infection	4 (0.4)	2 (1.0)	1 (0.3)	1 (0.2)	0.2813
Heart failure	10 (1.0)	5 (2.5)	2 (0.7)	3 (0.6)	0.0477
Myocardial infarction	3 (0.3)	1 (0.5)	1 (0.3)	1 (0.2)	0.7631
Thromboembolism	24 (2.3)	3 (1.5)	2 (0.7)	19 (3.5)	0.0241
Liver damage	342 (32.9)	75 (37.3)	68 (23.1)	199 (36.5)	0.0001
Renal damage	158 (15.3)	22 (10.9)	19 (6.5)	117 (21.7)	<0.0001
Macrophage activation syndrome	12 (1.2)	4 (2.0)	3 (1.0)	5 (1.0)	0.4952
Medications prescribed to patients
Ciclesonide	132 (12.5)	56 (27.7)	40 (13.3)	36 (6.5)	<0.0001
Favipiravir	276 (26.2)	58 (28.4)	71 (23.7)	147 (26.7)	0.4522
Hydroxychloroquine sulphate	5 (0.5)	5 (2.5)	0 (0.0)	0 (0.0)	<0.0001
Lopinavir, ritonavir	10 (1.0)	10 (4.9)	0 (0.0)	0 (0.0)	<0.0001
Remdesivir	111 (10.6)	3 (1.5)	28 (9.4)	80 (14.6)	<0.0001
Nafamostat mesilate	60 (5.7)	7 (3.5)	17 (5.7)	36 (6.6)	0.2638
Anti-IL-6R antibody	12 (1.1)	3 (1.5)	4 (1.4)	5 (0.9)	0.7548
Steroid	328 (31.3)	19 (9.4)	61 (20.5)	248 (45.2)	<0.0001

Data are presented as mean±standard deviationSD or n (%).

BMI, body mass index; ECMO, extracorporeal membrane oxygenationICU, intensive care unit

Comparison of long COVID among COVID-19 waves

There were no significant differences in the proportion of patients with one or more long COVID symptoms between waves at 3, 6 and 12 months after diagnosis; however, the first wave exhibited a higher trend (first wave: 52.7%, 44.9% and 35.2%; second wave: 43.0%, 39.0% and 29.4%; third wave: 45.8%, 39.7% and 34.1% at 3, 6 and 12 months, respectively). The first wave had the highest proportion of patients who had one or more long COVID symptoms across all time points (first wave, 31.3%; second wave, 19.5%; third wave, 22.9%) (figure 2A, online supplemental figure 2A). Among the 24 long COVID symptoms assessed, the average number of symptoms at 3, 6 and 12 months after diagnosis was compared between waves but no significant differences were observed (online supplemental table 3). When comparing the prevalence of each of the 24 symptoms at each time point between the waves, dyspnoea, alopecia, sleeping disorders and headaches were more prevalent at 3 months; dyspnoea and numbness at 6 months; and dyspnoea at 12 months in the first wave (figure 3). Dyspnoea had the highest prevalence at all time points in the first wave (first wave: 21.6%, 17.1% and 13.8%; second wave: 10.0%, 8.3% and 6.4%; third wave: 13.0%, 9.0% and 7.7% at 3, 6 and 12 months, respectively). Notably, the first wave also had the highest rate of dyspnoea across all time points (first wave, 10.4%; second wave, 1.6%; third wave, 2.7%) (figure 2B, online supplemental figure 2B).

Figure 2. Comparison of the prevalence of long COVID and dyspnoea in patients in each COVID-19 wave. (A) Comparison of the proportion of patients in each COVID-19 wave with one or more symptoms of long COVID at 3, 6 and 12 months, respectively, and across all three time points after diagnosis of COVID-19. No significant differences were observed at each time point. However, a significant difference was found in the proportion of patients with one or more symptoms of long COVID across all time points with the highest proportion observed during the first wave. The p value was calculated using the χ² test. (B) Comparison of the proportion of patients in each COVID-19 wave with dyspnoea at 3, 6 and 12 months, respectively, and across all three time points after diagnosis of COVID-19. Significant differences were observed at every time point and across all time points with the highest proportion observed during the first wave in all instances. P values were calculated using the χ² test. Long COVID, long coronavirus disease.

Figure 3. Comparison of the proportion of patients with each symptom of long COVID among infection waves at 3, 6 and 12 months. Only the symptoms with a prevalence of 5% or higher at any one of the three time points in at least one of the three waves are shown. Significant differences were observed in the prevalence of dyspnoea, alopecia, sleeping disorders and headaches at 3 months; dyspnoea and numbness at 6 months; and dyspnoea at 12 months with the first wave exhibiting the highest prevalence. Long COVID, long coronavirus disease.

Association between COVID-19 waves and long COVID

In light of the above results, multivariate logistic regression analysis was performed to assess the association between COVID-19 waves and long COVID considering age, sex, BMI, smoking status and COVID-19 severity as covariates which have been reported as risk factors for long COVID in large observational studies.^{6 8 22 23} The first wave compared with the second and the third wave showed a significant positive association with one or more long COVID symptoms across all time points (adjusted OR (aOR) 2.13, 95% CI 1.21 to 3.75 with the second wave as reference, aOR 1.70, 95% CI 1.05 to 2.75 with the third wave as reference), dyspnoea across all time points (aOR 10.99, 95% CI 2.34 to 50.00 with the second wave as reference, aOR 6.02, 95% CI 2.25 to 16.13 with the third wave as reference) and dyspnoea at 3, 6 and 12 months after diagnosis, respectively, (aOR 2.35, 2.16, 3.24, 95% CI 1.29 to 4.31, 1.07 to 4.37, 1.35 to 7.75 with the second wave as reference; aOR 1.90, 2.40, 2.20, 95% CI 1.15 to 3.15, 1.31 to 4.41, 1.11 to 4.37 with the third wave as reference) (figure 4).

Figure 4. Risk of prolonged long COVID, presence of dyspnoea and prolonged dyspnoea according to the first wave compared with the second and third waves. Forest plot of adjusted ORs and 95% CIs according to the first wave compared with the second and third waves using multivariate logistic regression analysis. Outcomes were adjusted for the COVID-19 wave, age, sex, BMI, smoking status and severity of COVID-19. Compared with the second and third waves, patients in the first wave exhibited a significantly higher risk of having one or more symptoms of long COVID at all time points as well as experiencing dyspnoea at each or all time points. Long COVID, long coronavirus disease.

Discussion

Our study examined the differences in long COVID symptoms among COVID-19 patients hospitalised during Japan’s first three waves. Patients hospitalised during the first wave had a significantly higher risk of one or more long COVID symptoms across all three time points (3, 6 and 12 months) than did those in the second and third waves. Dyspnoea was the most persistent and prevalent symptom with the highest prevalence observed in patients hospitalised during the first wave. The higher occurrence of long COVID symptoms, particularly dyspnoea, in patients hospitalised during the first wave may be attributed to several factors including differences in baseline and clinical characteristics among the different waves.

One possible factor was the high proportion of patients with ground-glass opacities or consolidations observed on chest imaging during the first wave. This indicates a higher incidence of COVID-19 pneumonia during the first wave which could lead to lung tissue damage and subsequent pulmonary fibrosis resulting in decreased lung function and a potential link to dyspnoea as a long COVID symptom.²⁴ It has been reported that COVID-19 pneumonia is one of the complications associated with an increased risk of long COVID.⁵ Additionally, secondary pulmonary infections have been reported to increase the risk of pulmonary fibrosis²⁴ and the first wave yielded a higher incidence of bacterial infections including bacterial pneumonia than did the other waves suggesting a potential association with cases of pulmonary fibrosis in the first wave. Previous reports comparing the clinical characteristics of hospitalised COVID-19 patients during the first three waves also reported a higher incidence of bacterial pneumonia and bacterial infections in patients during the first wave than during the second and third waves.^{12 25} However, this study did not collect data from follow-up chest CT scans after discharge.

Second, a higher prevalence of patients required intubation and invasive mechanical ventilation (IMV) during the first wave. As shown in table 1, the proportion of patients requiring oxygen therapy and ICU admission was highest during the third wave. However, the proportion of patients requiring intubation was lower in the third wave and higher in the first. Several reasons could account for this difference such as an improved understanding of the disease leading to the avoidance of invasive ventilation through the use of high-flow oxygen therapy, the increased number of COVID-19 patients who placed a strain on medical institutions compared with that observed in the first wave and the fact that the third wave had the highest proportion of older patients with the highest reported mortality rate without IMV¹² which may have led to a policy of not using IMV even when respiratory failure was severe and ventilator management was necessary. Patients in the first wave had the longest average length of hospital stay, exceeding 30 days, compared with those in the second and third waves.¹² It is conceivable that during the first wave more patients receiving IMV may have experienced respiratory muscle weakness and reduced lower extremity muscle strength because of prolonged hospitalisation leading to decreased exercise tolerance and increased dyspnoea.

As a third potential factor, the difference in medications administered across the three waves may have played a role. As shown in table 1, there were significant differences in the medications administered during each wave. Patients in the first wave were significantly more likely to be treated with ciclesonide, hydroxychloroquine sulfate and lopinavir/ritonavir compared with those in the second and third waves. Therefore, we investigated whether these administered medications were associated with long COVID symptoms. We divided the first wave patients into two groups: Those who received ciclesonide, hydroxychloroquine sulfate or lopinavir/ritonavir and those who did not. We then compared the clinical characteristics known to be risk factors for long COVID and the prevalence of the long COVID symptoms that were significantly higher in the first wave. The group that received these medications had a significantly higher proportion of middle-aged and older adults and a higher rate of severe cases requiring oxygen therapy. Our previous studies have already reported that middle age and oxygen requirement are risk factors for long COVID.¹⁹ However, despite these differences, there were no significant between-group differences in the proportion of patients with one or more symptoms or dyspnoea throughout 3, 6 and 12 months or in the prevalence of dyspnoea at 3, 6 and 12 months (online supplemental table 4). This suggests that the medications themselves had a minimal impact on long COVID symptoms during the first wave.

A fourth potential factor is that the viral strain responsible for the first wave had a higher risk of causing long COVID than did strains from other waves. Our analysis demonstrated a significant positive correlation between the first wave and long-term COVID symptoms at 3, 6 and 12 months after diagnosis including dyspnoea even after adjusting for known risk factors for long COVID such as age, sex, BMI, smoking status and disease severity of COVID-19.^{6 8 22 23} Various comorbidities have also been reported as risk factors for long COVID.^{6 22} The highest prevalence of comorbidities including hypertension, diabetes and chronic kidney disease was observed during the third wave which had a higher proportion of older individuals. These findings suggest that specific factors related to the first wave, beyond the known risk factors, contributed to a higher occurrence of long COVID symptoms including dyspnoea with one potential factor being the variation in viral strains. The clinical features of acute COVID-19 symptoms varied depending on the different waves of the outbreak in Japan.¹²²⁵,²⁷ Previous studies have suggested that mutant strains or specific variants of SARS-CoV-2 such as the Omicron strain may result in milder long COVID symptoms compared with those caused by conventional strains.¹³,¹⁷ In our study, all three waves were caused by the conventional strains whereas the mutant strains were responsible for the fourth and subsequent waves.²⁰ As mentioned above, there were slight variations in the conventional strains during the first three waves. The first COVID-19 case in Japan was detected on 16 January 2020 in a man with a history of travel to Wuhan, China.²⁸ Subsequently, cases of infection with the Wuhan strain which is the Phylogenetic Assignment of Named Global Outbreak Lineages (PANGOLIN) A or Wuhan-Hu-1 were reported in Japan including the outbreak on a cruise ship reported in February which was almost identical to that caused by Wuhan-Hu-1.²⁹ In mid-March 2020, an epidemic caused by the European strain (PANGOLIN B.1.1.114), believed to have been imported from Europe began and the presence of Wuhan-Hu-1 gradually diminished. As the first wave subsided in June, a lineage derived from B.1.1.114 with six nucleotide mutations (PANGOLIN B.1.1.284) suddenly emerged triggering a second epidemic wave. During the third wave which began in October, another distinct lineage (PANGOLIN B.1.1.214) became predominant. Most genomic sequences classified as B.1.1.284 and B.1.1.214 in the Global Initiative on Sharing All Influenza Data database are registered in Japan.³⁰ It is believed that both lineages originated from B.1.1.114 within the country. Although there have been no previous reports comparing the risks and severity of long COVID among these four viral strains, our study suggests that a potentially higher risk of long-term COVID symptoms, particularly dyspnoea, was associated with Wuhan-Hu-1 and B.1.1.114 which triggered the first wave compared with the aforementioned viral strains responsible for the second and third waves. Further studies focusing on viral genome analysis and its correlation with long COVID could provide valuable insights into this possibility.

Our study emphasises the need for continued follow-up and support for hospitalised patients during the first wave as they are at a higher risk of experiencing persistent symptoms. Dyspnoea was the most prevalent and persistent symptom in these patients. Exercise training, rehabilitation or nutritional management for individuals with long COVID may effectively reduce dyspnoea or other long COVID symptoms and improve their quality of life.^{31 32} Thus, our findings suggest that implementing such interventions in the posthospitalisation phase could alleviate the burden of persistent symptoms in this vulnerable population.

Our study had several limitations. First, the study only included patients who were hospitalised in the participating medical facilities which may introduce a selection bias towards more severe cases. Additionally, the size of the study population was limited for a nationwide study in Japan which may also introduce selection bias. Therefore, the findings may not be representative of all COVID-19 patients in Japan. However, to the best of our knowledge, this study represents the largest prospective cohort study on long COVID in Japan. When compared with other nationwide studies investigating differences across epidemic waves of COVID-19 in Japan, the clinical characteristics of the patient population in our study as described in the Results section closely align with those reported in these studies (online supplemental table 5).^{12 25} This suggests that our study population likely maintains a certain level of representativeness. Second, the study relied on self-reported symptoms from patients which may have been subject to recall bias or misinterpretation. While many previous studies have also used questionnaire-based methods and share similar limitations, the accuracy and consistency of reporting long COVID symptoms may vary among individuals. Finally, our study focused on patients from the first three waves of the pandemic in Japan. These findings may not apply to patients in other countries or future waves of pandemics with different variants.

Conclusion

Our study contributes to the growing body of knowledge on long COVID by highlighting the differences in the occurrence of symptoms among hospitalised COVID-19 patients during the first three waves in Japan. Patients hospitalised during the first wave had a significantly higher risk of experiencing one or more long COVID symptoms with dyspnoea being the most persistent and prevalent symptom. These findings underscore the need for continued monitoring and management of long COVID symptoms in COVID-19 survivors, particularly in those hospitalised during the first wave. Further research is needed to explore these differences’ underlying mechanisms and long-term implications.

Data availability statement

Data are available upon reasonable request.