Four distinct subtypes of long COVID defined in machine learning study


Using machine learning to track symptom clusters in around 35,000 COVID patients, researchers have identified four distinct types of long COVID. The findings suggest long COVID is a diverse disease with a wide variety of clinical manifestations.

The new research looked at two large cohorts of patients, all with at least one persistent symptom lasting between 30 and 180 days following a SARS-CoV-2 infection. A machine learning algorithm sorted through the mass of data, covering around 137 different lingering symptoms.

In a new study published in Nature Medicine, the researchers describe “four subphenotypes dominated by new conditions of the cardiac and renal systems (Subphenotype 1); respiratory system, sleep and anxiety problems (Subphenotype 2); musculoskeletal and nervous systems (Subphenotype 3); and digestive and respiratory systems (Subphenotype 4).”

The first subtype was found to be the most common, accounting for 34% of long COVID patients in the dataset. This subtype included patients with heart and kidney problems, anemia and circulatory disorders.

This manifestation of long COVID was more common in older patients (with an average age of 65) and those suffering from severe COVID. Interestingly, this subtype of long COVID was most prominent in those infected during the very first wave of disease across the first half of 2020.

The second subtype identified, almost as common as the first, accounted for 33% of all cases. This appearance of long COVID was dominated by lingering respiratory symptoms, chest pain, anxiety, headache and insomnia.

Unlike the first subtype, this second type of long COVID was associated with a more mild acute disease. It also appeared more common in patients infected later in the pandemic (between November 2020 and November 2021).

The third subtype (23% of patients) was mostly linked with musculoskeletal and nervous system disorders, including nerve pain and headache. This subtype was most commonly seen in patients with preexisting autoimmune conditions such as rheumatoid arthritis and asthma.

The final subtype was the rarest, only seen in 10% of patients. It was dominated by gastrointestinal disorders, including stomach pain, nausea and gut problems. This final subtype was linked to the most mild acute disease.

Rainu Kaushal, from Department of Population Health Sciences at Weill Cornell Medicine and co-senior author on the new study, said better understanding the variety of clinical manifestations of long COVID not only helps doctors effectively treat patients but guides researchers working to develop new treatments for this chronic condition.

“Understanding the epidemiology of long COVID allows clinicians to help patients understand their symptoms and prognoses and facilitates multispecialty treatment for patients,” said Kaushal. “Electronic health records offer a window into this condition, allowing us to better characterize long COVID symptoms, informing other types of research including foundational discoveries and clinical trials.”

The study isn’t the first to attempt to break long COVID down into distinct subtypes. A large UK study published in 2022 suggested the condition can be divided into three symptom clusters (cognitive, respiratory, and everything else).

However, these new findings offer the most robust and comprehensive categorization of long COVID subtypes to date. And the next step for the research team will be to better identify the risk factors for each long COVID subtype, and work out whether some treatments are more effective for certain subtypes than others.

The new study was published in Nature Medicine.

Source: Weill Cornell Medicine