We conducted a cross-sectional study at the Department of Rheumatology at Aarhus University Hospital in Denmark from 27 July 2021 to 21 December 2021 involving adult patients diagnosed with SSc according to the ACR/EULAR 2013 criteria [15].

Patients were excluded in case of recent or ongoing pneumonia, pregnancy, a diagnose of connective tissue overlap syndrome [16] or in case of severe physical or mental comorbidity, which prevented the performance of the 6MWT. Patients were allowed to use supplemental oxygen or walking aid during the 6MWT if needed.

The 6MWT was performed at room temperature by the same investigator (ALE) according to the American Thoracic Society guidelines [4]. Patients had acclimatised and rested for minimum 20 min before test start. SpO2, 6MWD and Borg dyspnoea score were collected [4, 17]. Raynaud’s attacks during the 6MWT were noted in case the patient fingers turned white and/or blue. SpO2 was continuously measured during the 6MWT by pulse oximeters (Vyntus® WALK, Nonin Model 3150, Viare Medical, Germany) using sensors at the finger, earlobe and forehead [18]. The pulse oximeters measured SpO2 as integers. The accuracy of the SpO2 measurement (interval of 70–100%) with low perfusion was ± 2%. At the first visit, an arterial line was placed in all patients by a trained anesthesiologist (HH or PJ) on the opposite arm of the finger oximetry sensor. Arterial blood was drawn immediately before (pre-exercise) and after (post-exercise) the 6MWT and analysed with a blood gas analyser (ABL800, Radiometer Medical, Brønshøj, Denmark).

A subgroup of patients (n = 40) repeated the 6MWT one week later without an arterial line. At visit 2 the earlobe and finger probe were placed on the same earlobe and finger as in visit 1 (Supplementary Fig. 1).

The oximeters were electronic and paired with tablets via Bluetooth, which generated a graph of the continuous measurement of SpO2 during the 6MWT (Supplementary Fig. 2). The quality of data was assessed by ALE using the following pre-specified criteria for exclusion of data (Supplementary Table 1).

(1) Technical error in collection or transfer of data from pulse oximeter and tablet (i.e., no readings from pulse oximeter or no data transfer from pulse oximeter to tablet at time of arterial blood gas test).

(2) Technical error in performance or analyse of arterial blood gas test.

In case of doubt, a consensus was reached in cooperation with KS.

The following clinical and paraclinical parameters were collected from the electronical patient record (MidtEPJ, Systematics, Aarhus, Denmark): SSc-disease characteristics, (ii) medication, (iii) modified Rodnan Skin score (mRSS), (iv) routine blood samples, (v) electrocardiogram (ECG), (vi) the latest pulmonary function test (PFT, median time since latest PFT was 10 months [interquartile interval(IQI): 2–18]), (vii) the latest high-resolution computed tomography (HRCT, median time since latest HRCT was 72 months [IQI: 20–113]), (viii) the latest transthoracic echocardiography, (ix) PAH detected by right-heart catheterisation, and (x) comorbidities. SSc-ILD was defined according to the HRCT criteria for ILD patterns [19].

All patients had Nailfold Videocapillaroscopy (NVC) images recorded of the 2nd to 5th finger. The capillary density was assessed with the 90-degree method from minimum 3 images per patient. The general capillary density was defined as the mean capillary density of the available pictures from the same hand.

Patients answered two self-reporting questionaries: Raynaud’s attacks the last month, including the Raynaud’s Condition Score (RCS) [20] and burden of ischaemic ulcers, and the Scleroderma Health Assessment Questionnaire (SHAQ) [21].

Categorial data are reported as counts and percentages, and continuous data as mean values and standard deviation (± SD) when normally distributed or otherwise as median values and interquartile interval (IQI). Data distribution was investigated Q–Q plots and histograms.

The agreement of the SpO2 of the three anatomical sites was examined using Bland–Altman plots to display the difference between SpO2 and SaO2 (bias) at pre-exercise and at post-exercise [22], and the re-test reliability was examined using Bland–Altman plots to display the agreements between the minimum SpO2 during the 6MWT, the 6MWD and Borg dyspnea score at visit 1 and visit 2. Furthermore, the overall accuracy of the peripheral oxygen measurement was calculated by the accuracy root mean square (Arms) (√ [(bias)2 + (precision)2) [23]. In accordance with the Food and Drug Administration recommendation, we used a cut-off of Arms < 3% as the acceptable accuracy of the SpO2 measurements [24].

The intraclass correlation coefficient (ICC, 95% confidence interval [95% CI]) for repeated measurement of minimum SpO2 was calculated based on an absolute agreement, two-way mixed effect model [25]. The ICC values were interpreted using the following definitions: ICC < 0.5: poor reliability, ICC: 0.5–0.75: moderate reliability, ICC 0.75–0.9: good reliability, and ICC > 0.90: excellent reliability. Furthermore, we calculated the frequency of measurement error of SpO2 at post-exercise for the finger, forehead, and earlobe sensors. Measurement errors was defined as values being ± 4% different from SaO2 values. In explorative analysis, we examined the impact of demographic parameters, comorbidities, and SSc specific parameters on the risk of measurement errors of the SpO2. Statistical significance was tested using Student’s t test, the nonparametric Mann–Whitney U test and Fisher’s exact test. All analyses were carried out in Stata17, where p < 0.05 was considered statistically significant.

The research project was approved by the Central Denmark Region Committees on Health Research Ethics (1-10-72-203-20) at the 30 October 2020 and listed in the Central Denmark Region register of internal research projects (1-16-02-270-20) at the 15 June 2020. ClinicalTrials.gov identifier: NCT04650659.