Adding cave air therapy to treatment for asthma, COPD, and long COVID in a controlled trial

This study was designed as a multicentre, parallel-group, randomized controlled superiority trial with a 1:1 allocation ratio.

Nine speleotherapy facilities in Germany, Austria, and South Tyrol (Italy) were included to address the primary research question. Five of these sites (marked with *) also participated in the CO2 subproject: Grube Bindweide, Rhineland-Palatinate * Silberbergwerk Bodenmais, Bavaria * Tiefer Stollen Aalen, Baden-Württemberg Silberbergwerk Neubulach, Baden-Württemberg * Eisensteinstollen Bad Grund, Lower Saxony Kluterthöhle Ennepetal, North Rhine-Westphalia * Feengrotten Saalfeld, Thuringia Klimastollen Prettau, South Tyrol, Italy * Friedrichstollen Bad Bleiberg, Austria

All facilities met the required climatic criteria and followed a standardized treatment protocol based on prior empirical experience.

Patients lay on comfortable reclining chairs for 1.5–2 h, wrapped in sleeping bags. At the beginning of each session, general physical and breathing exercises were performed for 5–10 min (7).

Data collection occurred at three time points: pre-treatment (T1), immediately post-treatment (T2), and 3 months after treatment completion (T3), or at corresponding time points for the control group.

The random allocation sequence was generated centrally using a computer-based random number generator.

No changes to the trial protocol, outcomes, or prespecified analyses were made after trial commencement. No interim analyses were planned or performed, and no formal stopping guidelines were defined.

Patients and the public were not involved in the design, conduct, reporting, or dissemination plans of this trial.

Participants aged 18 to 80 years with a physician-confirmed diagnosis of bronchial asthma, chronic obstructive pulmonary disease (COPD), or Long COVID were eligible for inclusion in this study. Recruitment was conducted through outpatient settings, and interested individuals with one of these diagnoses were able to register for participation.

The primary basis for study inclusion was the written diagnosis provided by the participant’s general practitioner or treating physician. No additional diagnostic verification by study physicians or specialists was performed as part of the study protocol.

Spirometry was performed primarily for follow-up monitoring and was not used as an inclusion criterion. Measurements were obtained under routine outpatient conditions and were conducted without standardized bronchodilator administration. Medication regimens and timing of intake were not controlled. Consequently, some participants classified as having COPD showed lung function values above commonly applied thresholds (e.g., FEV₁/FVC > 0.7). This may be attributable to current treatment effects, diagnostic variability in routine outpatient care, or intra-individual variability in lung function parameters (Additional File 5). Diagnostic group assignment therefore relied on documented physician diagnoses rather than spirometric cut-off values.

Exclusion criteria included the following: COPD (Global Initiative for Chronic Obstructive Lung Disease [GOLD]) stage 4, uncontrolled asthma (Global Initiative for Asthma [GINA] grade 3), frequent exacerbations (≥ 2 exacerbations requiring change in medication), decompensated heart failure, claustrophobia, pregnancy, acute upper respiratory infections (< 4 weeks prior to inclusion), active smoking within the past 3 months, and participation in speleotherapy within the previous year.

Following comprehensive written patient information and assessment of eligibility via a questionnaire, participants provided written informed consent for participation and data processing. Treatment was conducted on an outpatient basis; participants of the interventional group travelled daily from their place of residence (or temporary accommodation) to the therapy site, covering an average distance of 15 km. The control group received no speleotherapy during the study period but was allowed to participate after the study concluded.

The sample size was calculated based on the change between two measurement points for the primary endpoints and the two arms, assuming a medium to large effect (Cohen’s d ≈ 0.8), with α = 0.05 (two-sided) and a power of 90%. A total of n = 42 patients per arm (intervention/control) for each subgroup (Asthma, COPD, Long-COVID) was determined to account for non-parametric analysis methods and expected dropouts (15% and 10%, respectively).

Within each centre for each disease randomization was stratified by sex and treatment without additional blocking. Allocation concealment was ensured by central randomisation; investigators responsible for participant enrolment did not have access to the random allocation sequence and were informed of group assignment only after enrolment had been completed. It was conducted up to 2 weeks before the start of the study. In later phases, all eligible patients were included to meet the target sample size, which may have led to imbalances in sex and diagnosis distributions. Due to the nature of the intervention, participants and care providers were not blinded to group allocation. Outcome assessors and data analysts were not blinded.

ACT: 5 items. Assesses how well asthma is controlled¹².AQLQ: 32 items on a 7-point scale. Measures the extent to which asthma affects emotional and physical well-being¹³.CAT: 8 items on a 6-point scale. Evaluates the severity and impact of COPD on the patient’s daily life¹⁴.SGRQ: 50 items. Assesses disease-related impairments in patients with chronic respiratory diseases¹⁵.NQ: 16 items; cut-off score > 20. Assesses the degree of dysfunctional breathing¹⁶.FAS: 10 items. Used to assess the presence and severity of chronic fatigue¹⁷.Patient Questionnaire on Long COVID Syndrome (developed by the Median Clinic Group): 14 symptom-specific items with four response categories to rate severity. The Long-COVID Questionnaire was used with written permission from MEDIAN Klinik Flechtingen (Dr. P. O. Schüller, 2023). As this instrument is not formally published, it is cited as an unpublished clinical assessment tool¹⁸.

Additional file 1 provides an overview table of all measurement parameters, indicating whether they were defined as primary or secondary endpoints. It also summarizes the instruments used, the corresponding measurement scales, how the assessments were performed, and by whom the measurements were conducted.

Harms were defined as any adverse events or unintended symptoms occurring during the intervention period. Adverse events were assessed non-systematically by patient self-report during therapy sessions and at follow-up visits.

All methods were performed in accordance with the relevant guidelines and regulations and complied with applicable ethical and legal standards. The study was approved (or classified as not requiring formal review) by the ethics committees of the State Chambers of Physicians in Rhineland-Palatinate (initial vote), Baden-Württemberg, Bavaria, Thuringia, Westphalia-Lippe, and Lower Saxony, as well as by the responsible ethics committees in Italy (South Tyrol) and Austria (Carinthia). Registration: retrospectively registered on July 24, 2025). The study was conducted in accordance with the General Data Protection Regulation (GDPR, EU Regulation 2016/679).

All participants received written information regarding the study’s objectives, procedures, and data processing. Data collection was conducted using pseudonymized identifiers. The randomization list was stored separately with password protection.

Metric and ordinal data were presented descriptively in tables, including mean, standard deviation, and 95% confidence intervals for the mean. As alternative measures of central tendency and dispersion, the median (50th percentile) and interquartile range (IQR) were reported. Missing data were assessed across all primary outcomes and time points. As the overall rate of missingness was low (below 10%), no imputation method was applied. All analyses were therefore based on complete case analysis. Participants were analysed in the groups to which they were originally randomised. All continuous outcomes were summarised as medians with interquartile ranges and analysed as changes from baseline. Outcomes were assessed at baseline (T1), immediately after the intervention (T2), and at 3-month follow-up (T3).

Normal distribution was assessed using Q-Q plots and the Shapiro–Wilk test. The results typically contradicted the assumption of normality; therefore, data were analysed using the nonparametric approach for longitudinal data by Brunner and Langer¹⁹ to evaluate the main effects of treatment and time, as well as their interaction. Significant effects were further examined using post-hoc comparisons: comparisons between two groups and the change between two measurement points were conducted using the non-parametric Mann–Whitney U test and Wilcoxon signed-rank tests for within-group changes over time. The results of the repeated measures analysis according to Brunner and Langer, as well as a scheme of the post-hoc analysis for group and time effects, are presented in the additional file 9.

To control for type I error inflation, p-values were adjusted using the Bonferroni-Holm procedure. Given evidence suggesting a reduced therapeutic response in older patients with asthma²⁰, a post hoc subgroup analysis was conducted in asthma patients aged under 70 years using the same statistical methods as in the primary analysis. This analysis was exploratory and not prespecified.

The results were graphically visualized using boxplots. Data analysis was performed using R for Windows, version 4.4.1.

Of the 233 individuals initially enrolled, 208 patients with a diagnosis of bronchial asthma, COPD, or Long COVID were included in the analysis. Participants were recruited between 25 May 2024 and 15 September 2024, and follow-up for all outcomes was completed on 10 January 2025. A total of 25 participants were excluded for the following reasons: illness prior to the start of therapy (n = 6), illness during therapy (n = 5), fewer than 12 therapy sessions completed (n = 7), ineligible diagnosis (n = 6), and other reasons (n = 1).

Exclusions due to ineligible diagnoses were necessary based on subsequent diagnostic information. These included fibrosis (n = 1), allergic rhinitis (n = 1), multiple allergies (n = 1), post-infectious bronchitis (n = 1), a positive COVID-19 test without symptoms suggestive of Long COVID (n = 1), and a medically unconfirmed suspected COVID-19 case without Long COVID symptoms (n = 1).The average age of the 208 included patients was 62 (females: 59.2; males: 65.6). Additional file 2 shows the distribution of participants by diagnosis, sex, and group allocation (intervention/control).

The intervention was delivered by trained medical staff at the participating speleotherapy centres. Patient adherence was high, as reflected by a low rate of missing data of approximately 10%. Concomitant care, including regular pharmacological treatment, was continued unchanged in both the intervention and control groups throughout the study period.

In the total cohort, patients of the speleopherapy group (n = 98) and control group (n = 110) did not differ in baseline medication, FeNO, lung functional parameters, respiratory muscle strength (MIP, MEP) or questionnaire based scores at T1 before start of speleotherapy (Additional file 3 and 4). After 3 weeks of therapy (T2-T1) lung function parameters FVC and FEV₁improved significantly in both groups, whereas PEF only improved significantly in the interventional group. Moreover, respiratory muscle strength improved only in the interventional group and was significantly higher at T2 compared to the control group. Dysfunctional breathing according to the Nijmegen Questionnaire (NQ) decreased significantly in the interventional group at T2 and T3 and also in comparison to the control group. FeNO was neither changed in control, nor in the interventional group (Additional file 3).

In the total cohort of patients with bronchial asthma (n = 107; intervention: 54; control: 53), there were no differences between the intervention and control groups at baseline with regard to medication, FeNO, lung function, or respiratory muscle strength (MIP, MEP) (Additional file 4 and 5).

In the total cohort (n = 59; intervention: 27; control: 32), patients in the intervention and control groups did not differ in baseline medication, vital signs, FeNO, FVC, PEF or respiratory muscle strength (MIP, MEP). FEV₁at baseline was significantly lower in the intervention group compared to the control group (Additional file 4 and 6).

In the total cohort of patients with Long COVID (n = 42; intervention: 17; control: 25), there were no differences in baseline medication, vital signs, FeNO, lung functional parameters, respiratory muscle strength (MIP, MEP) or questionnaire-based scores at T1 (Additional file 4 and 7).

During the therapy period, CO₂ concentrations of the caves were measured in a sampling basis. At Bodenmais, only a slight increase was observed (462 ppm) compared to outside air (400 ppm). Other values recorded were: Prettau (970 ppm), Neubulach (1705 ppm), Ennepetal (2500 ppm), Bindweide (3665 ppm).

Two participants in the intervention group reported upper respiratory tract infections during the intervention period, which were considered possibly related to the therapy. No serious adverse events were observed.

In patients with asthma, the predefined primary endpoint FeNO remained unchanged, suggesting no measurable effect on type 2 airway inflammation.

In contrast, several secondary outcomes improved following speleotherapy. Notably, patient-reported outcomes reached established thresholds for clinically meaningful benefit. The median improvement in asthma control (ACT) met the minimal clinically important difference (MCID) of 3 points (²¹ [ACT MCID], and asthma-related quality of life (AQLQ) improved by a median exceeding or meeting the MCID of 0.5 points²²) [AQLQ MCID]. The application of MCID cut-offs to interpret clinically meaningful group-level differences in asthma trials is supported by expert consensus recommendations²³) [MCID consensus].

Physiological outcomes, including lung function parameters and respiratory muscle strength (MIP, MEP), also showed statistically significant changes. However, the magnitude of these changes was modest (e.g., an increase of approximately 4% in FEV₁ and 6–7 cmH₂O in MIP) and may fall within normal biological variation. In addition, clinically meaningful thresholds for these measures are less clearly defined in adult asthma, particularly for lung function parameters reported as percent predicted values. Therefore, these results should be interpreted cautiously and considered supportive of the clinically meaningful benefits observed in ACT and AQLQ. Improvements in Nijmegen Questionnaire scores further suggest reduced dysfunctional breathing symptoms, although no validated MCID is available for this instrument.

Overall, speleotherapy may exert its effects through two complementary mechanisms: improved airway function in asthma and enhanced breathing technique, reflected by better symptom control and quality of life.

Several review articles have addressed the effectiveness of non-pharmacological interventions in bronchial asthma. These approaches include environmental modifications, self-management strategies, physiotherapeutic applications, and multimodal interventions, which may also involve climate-based treatments such as stays at high altitude or by the sea^24,25. The present findings suggest that speleotherapy could represent an additional supportive option, either as an outpatient intervention or as part of a broader therapeutic setting combining multiple treatment modalities.

The impact of cold and humid environments on chronic respiratory conditions remains a subject of debate¹⁸. In therapeutic practice, underground climates with temperatures between 5°C and 12°C and relative humidity > 85% has been described in uncontrolled studies as being effective for speleotherapy in patients with asthma^26,27. As this air is warmed to 37°C within the airways, the relative humidity drops to < 30%, facilitating fluid absorption from the bronchial mucosa and promoting airway clearance. Additionally, the air is free from pollen, pathogens, and spores. High humidity binds dust particles, reducing mucosal irritation²⁸.

Another potential mechanism—particularly in asthma—may involve vagal nerve stimulation. The cool environment, exceptional stillness²⁹, and unique underground setting probably have vagotonic effects, contributing to reduced respiratory rates and dysfunctional breathing. This aligns with findings from Long COVID patients on the median questionnaire. We therefore hypothesize that, beyond a direct bronchial effect (e.g., reduction in mucosal oedema), indirect effects via vagal activation may play a role.

In Long COVID patients, no significant improvements in lung function were observed; however, signs of improved breathing were noted. The MIP, an indicator of diaphragmatic strength, improved significantly in the intervention group (p = 0.048), although this effect did not remain significant after Bonferroni-Holm correction. Improvements were also observed in dyspnoea, “problems with stair climbing/muscle exertion”, anxiety, and sleep disturbances. The reduction in NQ scores supports the hypothesis of improved dysfunctional breathing patterns, although these findings should be interpreted with caution given the small sample size of the Long COVID subgroup. This improvement may also be related to the observed decrease in fatigue severity, as reflected by lower FAS scores. Although FAS scores decreased, the median change of − 3 points remains slightly below the proposed minimal clinically important difference of approximately 4 points for the Fatigue Assessment Scale, indicating a trend toward clinically meaningful improvement but warranting cautious interpretation³⁰.

COPD patients reported improvements in CAT scores, despite no measurable improvements in lung function. The observed improvement in CAT scores (e.g., − 2 to − 3 points) reaches or exceeds the established minimal clinically important difference of about 2 points, suggesting that these changes are not only statistically but also clinically meaningful to patients³¹. This is consistent with clinical observations suggesting that COPD patients perceive subjective benefits from speleotherapy even without objective functional findings.

Exposure to the relatively cold environment and moderately elevated CO₂ levels did not result in any serious adverse effects. As reported in the Results section, only minor respiratory infections occurred in a small number of participants. This is consistent with recent evidence from a speleotherapy study in Long COVID patients, which likewise observed no serious adverse events, further supporting the safety of this therapeutic approach³².

Although the possibility of response bias in the questionnaire responses due to placebo expectations in the speleotherapy group cannot be fully ruled out, this appears unlikely to have substantially influenced the results, as no significant temporal changes were observed in the SGRQ and in most items of the MEDIAN Long-COVID Questionnaire.

Regarding FeNO values, no significant changes were observed in the overall group or subgroups. This may be due to the high proportion of participants on regular inhaled corticosteroid therapy (overall group: 64%; asthma: 79%).

An elevated CO₂ concentration in some therapeutic caves may have contributed to bronchial muscle relaxation. In caves with higher CO₂ concentrations, increased levels of CO₂ in the blood and exhaled air were observed, and these elevations partially persisted even three months after treatment (increased SpCO₂), possibly due to altered CO₂ sensitivity. Interestingly, improved dysfunctional breathing (NQ) in the interventional compared to the control group, was associated with higher pCO2 and lower breathing rate at T3 indicating decreased ventilation potentially correcting disease-associated hyperventilation. However, the pattern at T2 (lower pet-CO2 at maintained ventilation) suggest an increase in dead space ventilation.

A further limiting factor for detecting significant effects was the relatively high average age of study participants: total group (62), asthma (61), COPD (69), and Long COVID (55). Among participants under 70, the FEV₁ improved significantly in the intervention group compared to the control group.