The study design used for this proof-of-concept trial was an open-label randomised one-way crossover design to rule out regression to the mean (S1 Fig). This intervention is not suitable to compare to a genuine placebo effect. A concealed computer-based system randomised participants to an early intervention group or late intervention group in a 1:1 ratio. Stratification was performed for duration of disease and disease activity by ASDAS-CRP.

The early intervention group started with the intervention at baseline and continued with the intervention until week 8. Study visits in the intervention period were done at baseline, week 4 and week 8. Sixteen weeks after the intervention period, thus at week 24, a follow-up visit was done. The total study duration for the early intervention group was 24 weeks, in which 4 study visits were done.

The late intervention group started with the control period that lasted for 8 weeks. Study visits were performed at baseline, week 4 and week 8. After 8 weeks, the late intervention group started with the intervention and continued with the intervention for 8 weeks. The week 8 visit of the control period also served as the baseline visit for the intervention period. Study visits in the intervention period for the late control group were done after 4 and 8 weeks of intervention (thus at week 12 and week 16). Sixteen weeks after the intervention period, thus at week 32, a follow-up visit was done. The total study duration for the late intervention group was 32 weeks, in which 6 study visits were done.

Data from the first 8 weeks of study participation of the late intervention group were used as control data. Data from both intervention periods (of the early and late intervention group) were combined in the analysis. In both groups, the 8-week intervention period was followed by a 16-week safety follow-up period. Patients were enrolled between May and December 2016. Participating centres were the Academic Medical Centre (AMC) and Bernhoven Hospital. All training sessions took place in the AMC.

Patients aged between 18 and 55 years were eligible if they had a clinical diagnosis of axSpA according to the treating physician, fulfilled the ASAS classification criteria[6] and had active disease defined as ASDAS>2.1[7] and a high-sensitive CRP (hs-CRP) ≥5mg/L. Patients were allowed to use concomitant non-steroidal anti-inflammatory drugs (NSAIDs), corticosteroids (prednisone equivalent up to 10mg/day) and disease modifying anti-rheumatic drugs (DMARDs) (both synthetic and biologic) provided they have been initiated at least 8 weeks before screening (exception: two weeks for NSAIDs) and that the dose had been stable for at least 6 weeks prior to screening. Doses also had to remain stable throughout the study (from screening to end of intervention at week 8). Exclusion criteria were significant comorbidities that, in the opinion of the investigators, could interfere with the study or lead to deleterious effects for the patient, a recent history or persistence of an infection requiring hospitalization or antibiotic treatment within 4 weeks of baseline and pregnancy.

All participants gave written informed consent. Patients received a travel allowance that was dependent on the distance of the patients’ residence to the training site, but no other reimbursement. This study was approved by the Medical Ethics Committee of the Academic Medical Centre in Amsterdam under reference number 2015_328.

The intervention consisted of an 8-weeks add-on training program comprising three elements: 1) breathing exercises (further detailed below), 2) gradual cold exposure (immersions in ice cold water), and 3) meditation (third eye meditation).

The breathing techniques consisted of two exercises. First, patients were asked to hyperventilate for an average of 30 breaths. Subsequently, the patient exhaled and held their breath in an unforced manor for ∼2–3 min until they felt a stimulus to inhale (“retention phase”). The duration of breath retention was entirely at the discretion of the patient himself. For safety reasons, it was instructed to not hold the breath longer than 3.5 minutes. Breath retention was followed by a deep inhalation breath, which was held for 10 s. Subsequently a new cycle of hyper/hypoventilation began. After the last cycle, patients were instructed to do a strenuous exercise such as push-ups. The induced state of intermittent respiratory alkalosis and hypoxia typically “empowers” the patient to outperform their standard capability in any physical exercise. The second breathing exercise consisted of deep inhalations and exhalations in which every inhalation and exhalation was followed by breath holding for 10 s, during which the patient tightened all his body muscles. An additional element this part of the training program consisted of strength exercises (e.g., push-ups and yoga balance techniques).

During the intervention period, patients voluntarily exposed themselves to cold in two ways. During weekly training sessions, the patients immersed whole-body in ice-cold water (0–1°C) for several minutes, incrementally up to a maximum of 5 minutes (for safety reasons). At home, daily cold showers were taken incrementally up to 5 minutes (10–14°C).

The so-called “third eye meditation,” is a form of meditation including visualizations aimed at total relaxation. It consisted of an unguided meditation (in silence) with the eyes closed in any posture as desired by the patient for a period of 15–20 minutes. It was generally used at the end of each training session.

Consistent with the previous study,[2] patients were trained at the academic rehabilitation centre by a Dutch individual Wim Hof and four trainers who previously received an instructor course by Wim Hof to become a trainer. Medical personnel were present during all training sessions. During the first 4 weeks, participants had group trainings twice weekly, the second 4 weeks once weekly. During the intervention period, and after extensive written and oral instructions, participants practiced the exercises daily at home and registered their progression in a diary. In this diary the participants reported per day a) their progression in breath retention at home and b) their mental and physical state. The adherence to the training program was discussed and assessed weekly by the trainers and during the study visits by the medical personnel.

Primary safety assessments included vital signs, physical examination, electrocardiogram, haematology and chemistry at baseline, 4, 8, and 24 weeks and upon indication, and recording of adverse events (AEs) and serious adverse events (SAEs). The number and severity of adverse events was used to assess safety, other safety assessments, such as measuring blood pressure, heart rate, body temperature and physical exam, were done to detect potential adverse events that were not reported by the participants. Secondary endpoints were the change in serum hs-CRP (mg/L) levels between baseline and week 8 of the intervention/control period as a quick response measure, and erythrocyte sedimentation rate (ESR, mm/hr) and serum calprotectin levels[8] (measured by ELISA). Exploratory end-points were disease activity measured by ASDAS-CRP[6] and BASDAI[9], quality of life measured by the short-form 36 (SF-36)[10] and the EuroQol-5D (EQ-5D)[11] and depressive symptoms measured by the Hospital Anxiety Depression Score (HADS)[12]. All questionnaires were self-administered and used and scored according to the test manuals.

Because of the PoC design with safety as primary outcome, sample size could not be calculated and was estimated based on previous work[2]. All patients are included in the analysis of the primary outcome. For the primary outcome, safety, Mann-Whitney U tests were performed. For the secondary and exploratory outcome, a per protocol analysis was performed in which only patients that completed the intervention period were included. For the analyses of the intervention period, the intervention periods of the early and late intervention group were combined. The control group consisted of the first 8 weeks of the late intervention group. Secondary and exploratory outcomes were primarily assessed within groups over the 8-week intervention/control period using Wilcoxon signed rank tests comparing baseline and week 8. Because of the proof-of-concept nature of this trial, we did not adjust for multiple testing. A p value of <0.05 was considered statistically significant.

This intervention has received extensive media attention in the past years, resulting in patients with various conditions practicing this intervention. However there was little known about safety and putative immunomodulatory effects. The importance of a research project that would increase insight into these matters was highly appreciated by patients as mentioned in patient panels. To properly set the training program for patients with axSpA, 3 patient experts were consulted. As their strong preference was that all patients would receive the intervention, the study was designed as a one-way crossover open-label randomized (within-group) controlled trial. The results will be disseminated to study participants by personal communication and press release.

No serious adverse events occurred during the intervention period. During the intervention period, 10 participants in the intervention group reported a total of 12 AEs (2 participants reported 2 AEs), and 7 participants of the control group reported a total of 9 AEs (2 participants reported 2 AEs)(p = 0.268). The most common AE was common cold (4 vs. 2, respectively, Table 2). Three AEs during the intervention period had a moderate severity (1 in the intervention group and 2 in the control group). During the follow-up period (n = 19), there was one SAE of a participant who experienced a hypertensive crisis 8 weeks after end of the intervention and was hospitalised for one night. Upon further examination by a ophthalmologist there were signs of longstanding disease, therefore this SAE was judged ‘unrelated’ to the study procedures and did not lead to discontinuation of the study.

All secondary endpoints are listed in Table 3. In the intervention group, ESR significantly declined over time: from 16[9–26.5] mm/hr at baseline to 9[5–23] mm/hr at week 8, p = 0.040, while it did not in the control group: from 14[8.3–27.3] mm/hr to 16[5–37] mm/hr, p = 0.406). Hs-CRP did not significantly change within both groups during the 8-week intervention period (intervention group: from median [IQR] 10.2[6.5–17.1] mg/L to 6[3.9–15.6] mg/L, p = 0.103, control group: from 8.7[5.2–17.2] mg/L to 13.2[7.9–20.1] mg/L, p = 0.286). Calprotectin tended to decline in the intervention group (from 2295[1648–4923] pg/mL to 2165[953–3734] pg/mL, p = 0.064), whereas it remained stable in the control group (2115[1300–2571] pg/mL to 2279[1770–4989 pg/mL], p = 0.307).

All exploratory endpoints are listed in Table 3. There was a significant improvement in the median ASDAS-CRP during the 8-week period in the intervention group (p = 0.044), but not in the control group (p = 0.213). Median BASDAI declined significantly in the intervention group (p = 0.012) but not in the control group (p = 0.755). The physical component score (PCS) of the SF-36 increased significantly during the intervention period (p = 0.004) but not in the control group (p = 0.859).The MCS of the SF-36 increased significantly in the intervention group (p = 0.004) but not in the control group (p = 0.859).The EQ-5D did not significantly change in the intervention group (p = 0.102) nor in the control group (p = 0.933).A similar effect was observed for the EQ-5D VAS (p = 0.090 and p = 0.674 in the intervention and control groups, respectively. There was no significant effect on the HADS anxiety and depression scales within groups.

The main burden of the intervention assessed by patients themselves was the commuting time to the academic center for group training and follow-up visits. The time spent on group and personal trainings were assessed as purposeful and the burden of cold exposure was transient. The participation during the training sessions was judged as high by the trainers. The adherence to the exercises at home was discussed in the group sessions and reviewed in the diaries, and was also judged as high.