par Pauwen, Nathalie Yaël
Président du jury Deboeck, Gaël
Promoteur Faoro, Vitalie
Co-Promoteur Bruyneel, Marie
Publication Non publié, 2026-01-15
Thèse de doctorat
Résumé : Introduction -- Hyperventilation syndrome (HVS) is ventilation exceeding metabolic requirements, often accompanied by reduced arterial and end-tidal CO₂ (PETCO₂) and transient respiratory alkalosis. It presents with a constellation of respiratory, cardiovascular, neurological, gastrointestinal, and affective symptoms, with a marked impact on quality of life. Historically, the same clinical picture was labelled ‘irritable heart’ or ‘circulatory neurasthenia’, underscoring the syndrome’s position at the interface of physiology and psychiatry. Idiopathic HVS refers to cases without an identifiable organic disorder after structured exclusion; and may present as either acute episodes or chronic forms. Although early publications on HVS already mentioned some physiological clues (such as abnormalities in CO₂ uptake and/or altered chemosensitivity), psychogenic explanations of the HVS prevailed for decades, crystallizing around a symptom-hypocapnia paradigm that shaped research and clinical reasoning for many years. A major challenge still persists: the aetiology of HVS remains unknown, no gold standard exists, prevalence estimates are unreliable, and its polymorphic symptom profile overlaps with other conditions, contributing to uncertainty for both clinicians and patients. Symptom-based tools illustrate these limitations: the Nijmegen Questionnaire (NQ), although widely used, was validated without a physiological reference and its discriminative properties for HVS were later questioned by its own authors. Despite this, the NQ remains the most common diagnostic tool, largely due to the absence of alternatives. In parallel, the original Hyperventilation Provocation Test (HVTest) has undergone major reinterpretations and modifications. Initially, the HVTest was strictly physiological, based on end-tidal CO₂ dynamics, without requiring symptoms as an endpoint. Over the following decades, the HVTest was combined with symptom reproduction and questionnaire scores, introducing circular reasoning (symptoms defining the HVS and then ‘confirming’ themselves) and reducing specificity. A well designed placebo-controlled design further highlighted the fragility of this approach. These developments highlight overlooked physiological mechanisms, such as the contribution of central/peripheral chemoreceptors, CO₂ stores, plant gain and ventilatory-control instability (loop gain). However, recent efforts have returned to physiology, emphasizing objective PETCO₂ recovery. These new insights could support a transition from symptom-based screening towards an integrated physiological framework that may more clearly define HVS, reduce misclassification, and shorten the diagnostic pathway. Objectives -- This work aims to move beyond symptom-based approaches to HVS by investigating objective physiological mechanisms: characterising ventilatory features of HVS, using Cardiopulmonary Exercise Testing (CPET), HVTest and Hypercapnic-Hyperoxic Challenge Test (HHCT); redefining diagnostic criteria based on physiological markers; evaluating the validity and usefulness of existing diagnostic tools and finally, proposing an integrated physiological model of HVS. This study therefore examined three hypotheses: (1) ventilatory alterations during CPET reflect distinct patterns in HVS; (2) CO₂ kinetics during HVTest provide a more robust diagnostic marker than symptoms; (3) HHCT reveal changes in chemosensitivity, CO₂ stores, plant gain characteristic of HVS. Methods -- Building on insights gained from supervising 28 master’s theses on HVS over a 7-year period, 3 retrospective and 2 prospective studies were conducted to test the hypotheses: [1]. Ventilatory patterns and cut-offs at CPET: Retrospective analysis of CPET was conducted in 20 female HVS+ patients (NQ ≥ 23/64) and 20 matched HVS- controls. All participants underwent lung function testing and a maximal CPET on a cycle ergometer. Ventilatory, gas exchange, and cardiovascular responses were compared, allowing the identification of distinct patterns. In a second retrospective study using the HVTest as gold standard, 14 HVS+ and 14 matched HVS- were analysed to establish diagnostic cut-off values, using group comparisons, ROC curves, sensitivity, specificity, and AUC. [2]. Validity of the NQ in primary care: In a primary care population, the NQ was retrospectively evaluated in 112 HVS+ patients (HVTest-positive, used as gold standard) and 212 HVS- controls (88 healthy, 124 with somatic disorders: COPD, cardiac disease, asthma, gastric, thyroid, or psychological/neurologic conditions). Structural/content validity, as well as predictive properties (AUC, sensitivity, specificity, PPV, NPV) were assessed. Principal component analysis was conducted to explore underlying dimensions of the NQ. Scores were stratified and compared by disorder, to assess potential misclassifications. [3]. Analysis of the PETCO₂ kinetics of the HVTest: This study began with a pilot study (2019), followed by a larger prospective investigation (2022) comprising a training cohort (37 HVS+ patients and 37 matched HVS- controls) and a validation cohort of 152 patients. The standardised HVTest consisted of 3-minute adaptation, 3-minute hyperventilation, and 5-minute recovery phases. PETCO₂ kinetics were modelled with a curvilinear TAU function, and diagnostic performance was evaluated using group comparisons, ROC curves, sensitivity, specificity, and AUC. [4]. Determinants of ventilatory control using HHCT: This prospective study included HVS+ adult patients defined by 4 concordant criteria: clinical suspicion, positive NQ, normal lung function and most importantly, a positive HVTest. HVS- controls showed none of these features. HVS+ and HVS- were carefully matched (for sex, age, height, and weight). Ventilatory and drive-to-breathe control were assessed with the HHCT using Read’s non–steady-state rebreathing method. Responses were standardised with Read’s approach. From these, the following parameters were estimated: α-intercept, controller gain (β-slope = ΔVE/ΔPETCO2), Ventilatory and Drive Recruitment Thresholds (VRT, DRT), Apnoeic and Chemoreceptor thresholds (AT, CT), Peripheral Sensitivity Range (PSR), CO₂ stores, and Plant gain (PG = CO₂ stores/ΔVE, below eupnoea) was estimated indirectly under isometabolic conditions. Statistical analyses included group comparisons and correlations. Multivariate linear regression was used to identify determinants of PG, while logistic regression assessed the predictive value of ventilatory-control parameters for HVS status, with bootstrap resampling applied to ensure robustness and external validity. Results & Discussion -- (a) Ventilatory patterns and cut-offs at CPET: In the first retrospective CPET analysis, HVS+ patients identified with NQ exhibited lower VO₂max and Wmax compared to controls, despite reporting similar levels of dyspnoea and fatigue. Ventilatory inefficiency persisted throughout effort, with elevated EqCO₂. Three distinct ventilatory profiles emerged: ‘Normalised’ (≈15%, with preserved performance during exercise), ‘Exacerbated’ (≈35%, persistent hypocapnia and ventilatory inefficiency with severely reduced capacity), and ‘Questionable’ (≈50%, ventilatory dissociation during exercise, despite normal resting parameters), highlighting heterogeneity in exercise responses. These findings suggest that the NQ may be prone to false positives, and that HVS could represent a continuum rather than a strictly binary condition. In the second retrospective study using the HVTest as gold standard, diagnostic cut-off values at first ventilatory threshold (VT1) were identified: EqCO₂ ≥ 33.2 (AUC=0.78, Se/Sp=0.70/1.00) and PECO₂ ≥ 26.7 mmHg (AUC=0.79, Se/Sp=0.77/0.82). Although CPET is not suitable as a universal gold standard (since ≈70% of patients fail to reach VT1), it provides valuable objective markers and helps differentiate distinct ventilatory phenotypes of HVS. (b) Validity of the NQ in primary care: Factor analysis confirmed a three-component structure of the NQ, consistent with the original model. In primary care, the NQ demonstrated fair discriminative power (AUC=0.89; Se/Sp=0.86/0.79), but its PPV was limited (0.32) at 10% prevalence of HVS in general population. Misclassification was frequent in patients with comorbidities, with false-positive rates up to 75% for certain conditions and ≈14% of false-negatives overall. Overall, while the NQ remains useful for a symptom-screening tool, it cannot serve as a stand-alone diagnostic instrument. (c) Analysis of the PETCO₂ kinetics of the HVTest: During HVTest, some HVS+ patients showed significantly lower PETCO₂ values already during the adaptation phase, with a steeper negative slope compared to controls. At the end of the voluntary hyperventilation phase, a PETCO₂ fall < 17.6 mmHg discriminated HVS+ from HVS- (AUC=0.81; Se/Sp=0.76/0.73), although predictive accuracy decreased in the validation cohort (AUC=0.67; Se/Sp=0.66/0.61). Recovery kinetics provided stronger diagnostic value at the 3rd minute of the recovery phase, AUC reached 0.87 (Se/Sp=0.81/0.81), whilst the 5th minute criterion (ΔPETCO₂ < 12.8 mmHg) was the most robust in both cohorts (training: AUC=0.91; Se/Sp=0.92/0.84 – validation: AUC=0.85; Se/Sp=0.80/0.84). These findings support PETCO₂ recovery kinetics as a robust physiological marker, reflecting impaired CO₂ reuptake after voluntary hyperventilation in HVS+ patients. (d) Determinants of ventilatory control using HHCT: At resting eucapnia, HVS+ patients exhibited higher VE, VT and VTi/Ti with greater ventilatory variability (CV of VE) and more dyspnoea, while PETCO₂ was similar to HVS- controls. During HHCT, HVS+ showed preserved controller gain (β-slope), but reduced thresholds (VRT, DRT), lower AT and a 50% larger PSR. Mean CO₂ stores tended to be in opposite directions in HVS+ and HVS-, whereas PG was reduced in HVS+. At peak hypercapnia, groups reached similar PETCO₂, VE, VTi/Ti and dyspnoea. In multivariate linear regression, a lower PG was predicted by lower AT, higher PSR, and a [PSR× CO₂-stores] interaction, highlighting the buffering capacity of CO₂ stores in reducing the negative impact of carotid body stimulation on PG. In logistic regression, higher CV of VE and a lower PG independently predicted HVS status. Together, findings indicate impaired CO₂ buffering (↓PG) and a widened peripheral chemoreceptors operating range at rest (↑PSR) with variability-driven instability near eucapnia in HVS+. These findings support a physiological framework in HVS, in which altered peripheral chemosensitivity, an expanded PSR, reduced PG and diminished CO₂ stores buffering may interact to destabilise resting VE, thereby contributing to ventilatory instability in HVS+. Rather than a symptom-defined entity, it might be best characterised by objective physiological markers. Conclusion -- From this research, four converging results delineate the pathophysiological and diagnostic features of hyperventilation syndrome. First, the HVTest proved robust, with a ΔPETCO₂ < 12.8 mmHg at 5 minutes providing a reliable diagnostic marker. Second, CPET showed that an EqCO₂ ≥ 33.2 at the VT1 may help identify HVS, though its use is limited in patients who do not reach VT1. Third, the Nijmegen Questionnaire remains useful for screening symptoms, but leads to substantial misclassifications if applied as a stand-alone diagnostic tool. Finally, HHCT analyses converge on a physiological profile in HVS: reduced PG, with CO₂ stores emerging as a crucial buffering factor, tempering the destabilising interaction between PG and heightened peripheral drive. This fragile balance is expressed as increased ventilatory variability and a dissociation between eucapnic ventilation and CO₂ levels at rest. Together, these features may indicate that HVS involves a disorder of ventilatory control stability, rooted in weakened CO₂ buffering and increased reliance on peripheral chemoreceptor drive.

Documents en relation

DI-fusion