Unmasking Obesity Hypoventilation Syndrome Trajectories: Clustering Patient Journeys in Real-World Noninvasive Ventilation Use

Obesity hypoventilation syndrome (OHS) represents a severe form of obesity-related respiratory compromise that is defined by the combination of obesity (body mass index ⩾30 kg/m²), sleep-disordered breathing (often coexistent with obstructive sleep apnea [OSA]), and hypercapnia (awake resting arterial CO₂ tension/pressure ⩾45 mm Hg) after excluding other causes of hypoventilation (1). OHS can have a variety of clinical presentations and prognoses. Unfortunately, a significant portion of patients are not identified until their presentation with acute-on-chronic hypercapnic respiratory failure, requiring hospitalization and/or admission to the intensive care unit (ICU), and the condition is associated with poor outcomes (2). Although OHS mortality rates have improved with the use of positive airway pressure (PAP) therapy (3), the overall prognosis is poor, which has been attributed to the high rates of cardiometabolic disease in these patients (4). Although clinical trials and the recent American Thoracic Society panel offer guidance on OHS therapy, real-life clinical practice may deviate from guideline-driven pathways (5).

In this issue of AnnalsATS, the study by Pépin and colleagues (pp. 1554–1566) using a large national French healthcare database further contributes to the OHS field with the provision of real-world evidence on health trajectory (6). This observational study identified 8,842 individuals who had noninvasive ventilation (NIV) reimbursement for OHS (based on International Classification of Diseases, 10th Revision criteria) and performed cluster analysis to identify four distinct clusters. The study demonstrated that their clusters of patients with obesity-associated hypoventilation had markedly different outcomes in terms of mortality, respiratory-related hospitalization, cardiovascular hospitalization, and therapy termination rates.

Cluster 1 comprised patients who began NIV in a stable outpatient setting, presumably newly diagnosed, and were younger and predominantly female. Approximately one fourth of this cluster had NIV initiated around bariatric surgery, suggesting its use for perioperative support. There was a high discontinuation rate (52%) without transition to continuous PAP (CPAP), which raises the question whether these patients truly had OHS to start with. Alternatively, NIV termination rates may be related to a large proportion undergoing bariatric surgery. If true, this highlights the reversibility of OHS and the importance of incorporating weight loss management.

In contrast, cluster 2 is a representation of the sickest patients, who commenced NIV after ICU admission for acute-on-chronic respiratory failure. They had higher rates of cardiovascular and metabolic comorbidities, potentially indicative of stage IV OHS (7). As expected, this cluster had the highest mortality and hospitalization rates. During the 24-month lead-in period, a portion of cluster 2 patients also had prior ICU admissions and/or hospitalizations. This demonstrates a missed opportunity to diagnose and intervene early, thereby improving these patients’ prognosis and preventing future acute respiratory failure. Cluster 2 data emphasize the prognostic importance of delayed diagnosis and late-stage intervention.

In theory, clusters 3 and 4 reflect more guideline-concordant PAP therapy. Cluster 3 patients were initially treated with CPAP before escalating to NIV following ambulatory reassessment, likely indicative of CPAP failure. Cluster 4 included patients who were undergoing long-term CPAP therapy before transitioning to NIV, presumably because of the development of hypercapnic respiratory failure. The lower mortality and hospitalization rates observed in these two clusters suggest that regular follow-up and timely treatment escalation may avert OHS-related complications.

One of the study’s strengths is the use of a large and comprehensive database. The French National Healthcare Data System is Europe’s largest database, covering more than 65 million lives and approximately 99% of the French population (8). With close to 9,000 study participants, this is the largest population-based study to date focusing on individuals with OHS treated with NIV. The broad coverage across public and private hospitals in outpatient and inpatient settings reduces biases associated with single-center or device-linked datasets. Because existing randomized clinical trials have stringent eligibility criteria, with the exclusion of participants with OHS with more comorbidities, this study has more real-world generalizability. The study data also predated the coronavirus disease (COVID-19) pandemic and avoided contamination from pandemic-related influences on healthcare delivery, obesity-associated mortality, PAP use, and diagnostic coding.

Another strength is the use of trajectory-based clustering to capture data on clinically important outcomes and therapy transitions and discontinuations over a relatively long period of follow-up (median, 3.1 yr). The study has identified additional clinical phenotypes based on care pathways with differing outcomes. The trajectory figures allow visualization of the dynamic nature of OHS therapy and disease journey, which is rarely captured in existing clinical trials.

The study also offers insight into real-life treatment pathways and hints at its deviation from current guideline recommendations. Despite consensus that CPAP should be the first-line treatment in patients with OHS and coexistent severe OSA (sOSA), cluster 1 showed that NIV is directly initiated in many patients who are in stable condition. Some of the discrepancy may be due to the timing of the dataset and the publication of the American Thoracic Society guidelines (5). Nonetheless, the underuse of CPAP raises concerns about the acceptance, implementation, and practicality of the current treatment ladder for OHS with sOSA.

There are several important limitations to this study. There is an absence of key clinical parameters such as body mass index, gas exchange (from arterial blood gas analysis), pulmonary function, and sleep polysomnographic results that are fundamental in the diagnosis and phenotyping of patients with OHS. The study is unable to distinguish between sOSA and non-sOSA phenotypes or assess the severity of OHS based on physiological data. This impacts the generalizability and interpretability of the study.

Although the authors used the relevant codes to exclude other potential causes of hypoventilation such as chronic obstructive pulmonary disease and some neuromuscular diseases, other important contributors were not excluded (chest wall disorders, medication-related causes, and other cardiopulmonary diseases). Without more detailed clinical data, the risk of misclassification remains. Hence, there is uncertainty whether the included study population is reflective of true OHS based on the traditional definition or includes patients in whom obesity overlaps with other hypoventilation contributors.

The study cannot determine whether NIV prescriptions were appropriate or potentially overprescribed. Similarly, the meaning of CPAP failure is difficult to interpret without information on the rationale for switching from CPAP to NIV in clusters 3 and 4. Because the source of patients came from NIV reimbursement, the study excludes patients with non–NIV-treated OHS (OHS treated solely with CPAP or undiagnosed or untreated OHS). Potentially, the study may overrepresent the more severe or system-engaged cases and cannot provide any insight into the natural course of patients with untreated or CPAP-managed OHS.

Nonetheless, the study demonstrates that valuable real-world insights can be gathered from a large database. The trajectory-cluster approach should be replicated in other countries with different healthcare systems to assess variations in care delivery and confirm the generalizability of key findings from the French National Healthcare Data System dataset. In the era of big data research and centralized data banks, future studies could look to integrate national databases with clinical registries, sleep and/or pulmonary laboratory data, electronic medical records, and device-linked datasets, thereby allowing the validation of different OHS phenotypes as well as patients with mixed pathologic conditions, who are often neglected in clinical trials. We need better strategies to identify and connect with patients with OHS who are trending toward cluster 2: how to best establish treatment of these patients to improve their long-term prognosis and prevent hospitalization. Additional research is desired to provide a better understanding of clinician- and/or system-related barriers to using CPAP in patients with OHS with sOSA, but also to clarify what constitutes CPAP failure. Finally, the research and management focus will need to be multimodal, with increased attention to addressing these patients’ underlying obesity and cardiometabolic disease.