Thoracic outlet syndrome (TOS) is common among those engaged in professional activities involving repetitive arm movements as well as among athletes (Chandra et al., 2014; Shutze et al., 2017; Hilberg et al., 2021). TOS is reported to occur in almost half of vascular diseases in sports (Arko et al., 2001). The diagnosis of thoracic outlet syndrome (TOS) remains difficult (Demondion et al., 2006; Ammi et al., 2017; Ahmad and Murthy, 2018; Burt, 2018; Illig, 2018). Surgical treatment of TOS is a significant source of malpractice claims against vascular surgeons (Li and Dissanaike, 2021). Consequently, documenting ischemia in cases of compression (impaired inflow) of the subclavian artery is of major interest in patients with TOS, even if symptoms are generally considered to be of neural origin.

According to the standards proposed for thoracic outlet syndrome by the Society for Vascular Surgery, TOS should be considered as being of arterial origin only if patients have “clinical symptoms due to documented symptomatic ischemia or objective subclavian artery damage caused by compression” (Illig et al., 2016). We assume that transcutaneous oxygen pressure (TcpO2) could fill a gap and provide objective arguments for the presence of ischemia in patients with suspected TOS.

Among positional manoeuvres aiming to induce a compression of the neurovascular bundle, the Roos test (also referred to as the elevated arm stress test: EAST) is the most widely used provocative manoeuvre. The test consists of abduction and external rotation, also known as the surrender/candlestick position (Nord et al., 2008; Fried and Nazarian, 2013; Adam et al., 2018). We recently showed that during the Roos test in TOS, the forearm ischemia resulting from impaired positional perfusion to the upper limb can be estimated by TcpO2 bilaterally and simultaneously (Henni et al., 2019; Abraham et al., 2020). TcpO2 has been used for years at the lower limb level in patients with arterial disease (Abraham et al., 2018a; Constans et al., 2019). In patients with suspected critical lower limb ischemia, when TcpO2 remains within normal limits at rest, elevation of the limb allows for better discrimination between normal and abnormal TcpO2 results. Indeed, when perfusion is impaired but TcpO2 at rest remains within normal values, the post-stenotic perfusion pressure is insufficient to compensate for the hydrostatic pressure decrease that is induced by limb elevation. As a result, TcpO2 decreases more in impaired perfused limbs than in normally perfused limbs where only a moderate decrease is observed as a result of decreased hydrostatic pressure (Paraskevas et al., 2006; Ruangsetakit et al., 2010). Maximal amplitude of the decrease in TcpO2 can be expressed as the “minimal decrease from rest of oxygen pressure” (DROPm) index (Henni et al., 2019; Abraham et al., 2020). We also previously showed that positional compression of the subclavian artery does not necessarily result in arterial ischemia (Hersant et al., 2022).

We then hypothesised that similarly to what is observed in the lower limb, elevation of the upper limb would facilitate the detection of impaired but non-abolished perfusion. If so, the position of TcpO2 probes on the upper limb could be important in the evaluation of ischemia because there is a wide difference in the amplitude of elevation at the level of the forearm (F-), as compared to the upper arm (U-) in the candlestick/surrender position. We hypothesised that the U-DROPm value would be higher than the F-DROPm value during manoeuvres potentially inducing arterial compression. We also hypothesised that F-DROPm would better predict the presence of positional arterial compression of the subclavian artery and/or positional upper limb symptoms than U-DROPm, but with different cut-off values for F-DROPm compared to U-DROPm. To test these hypotheses, we performed a cross-sectional study of patients referred for suspected TOS by simultaneously recording U-TcpO2 and F-TcpO2 during a standardised provocative test. Because no ideal gold standard exists for diagnosing TOS, we compared TcpO2 results to symptoms only, ultrasound only, angiography only and combinations of these three elements.

Among the 514 referred patients, only 175 subjects (350 arms) were eligible, had accessible or available imaging results, and had simultaneous arm and forearm TcpO2 recordings as shown in Figure 2. One hundred and three of these patients engaged in repetitive movements in the course of their professional activities or in weight training. Only 62 subjects were involved in regular sports activities including running (n = 32), cycling (n = 7), weightlifting (n = 4), and badminton (n = 3), and most had to stop doing sports because of their upper limb pain.

Among the studied subjects, 58 were male (33.1%). The subjects were 40.7 ± 12.1 years old, with a weight of 72.8 ± 15.9 kg and height of 167 ± 9 cm. Most subjects were right-handed (n = 153, 87.4%) and had displayed symptoms for a minimum of 6 months (n = 159, 90.9%). 94 patients took painkillers for their upper limb pain (n = 94, 53.7%). The disability of arm and shoulder (DASH) score was available (at least 27 answers) for 128 patients and was 41 ± 21. According to their history, patients complained of right (n = 57), left (n = 48) or bilateral (n = 70) positional symptoms. During the Ca-Pra manoeuvre, patients complained of right (n = 41), left (n = 46) or bilateral (n = 61) symptoms, but 37 patients remained asymptomatic.

All but two patients had ultrasound investigations and only 37 patients had an angiogram as a pre-surgical investigation. Results for arterial pressure, ultrasound, angiography and transcutaneous oximetry at rest or during the provocative manoeuvre are shown in Table 1.

A typical example of a recording and the corresponding angiography in a patient showing unilateral positional occlusion of the subclavian artery is reported in Figure 3.

The Kolmogorov-Smirnov test confirmed the normal distribution of resting U-TcpO2 (z = 1.354; p = 0.051) and F-TcpO2 (Z = 0.865; p = 0.442) values. No significant difference was found in relation to absolute value for resting U-Tcpo2 (70.1 ± 11.2 mmHg) and F-TcpO2 (71.1 ± 10.8 mmHg (p = 0.111). On the contrary, as shown in Figure 4, there was a non-gaussian distribution of U-DROPm (Z = 4.065; p < 0.001) and F-DROPm (z = 2.973; p < 0.001) values, with a median value of −5 [−11/2] mmHg and −12 [−19/−7] mmHg among the 350 upper arm and 350 forearm values respectively (p < 0.001).

On a limb-by-limb basis, as reported in Tables 2, 3, the absolute TcpO2 value at rest was not predictive of the presence of Symp+, US+, Ang+, Imag + or CS+, whereas all areas (except F-DROPm for Ang+) were significant for DROPm results with areas ranging from 0.596 to 0.689.

As shown in Figure 4, the areas under ROC were not significantly different for upper arm and forearm DROPm values, but the DROPm values that provided the shortest distance to the 100% sensitivity/100% specificity angle (best compromise of sensitivity and specificity to be used for diagnosis) were not the same for U-DROPm and F-DROPm.

These threshold differences were found whatever the end point used, with thresholds obtained being −4.5 vs. −14.5 mmHg, −6.5 vs. −14.5 mmHg, −7.5 vs. −14.0 mmHg, −6.5 vs. −14.5 mmHg and −7.5 vs. −14.5 mmHg, for the detection of Symp+, US+, Ang+, Imag+ and CS + results at the upper arm vs. forearm levels respectively. Note again that the forearm DROPm value for Ang+ is provided but that the area of the ROC curve is not significant as shown in Table 2.

As per our hypothesis, there was a lower DROPm (larger decrease of DROP) at the forearm level than at the upper arm level as a result of the difference in altitude changes of the probes during provocative manoeuvres performed in the standing position: median values found for DROPm (see Table 2) being systematically lower by 6–10 mmHg at the forearm than at the upper arm level. Contrary to our hypothesis, diagnostic performances estimated from the AUROC were similar at the forearm and upper arm levels, but with a lower DROPm threshold value at the forearm level than at the upper arm level.

Many authors advocate that neurogenic TOS comprises over 90% of all TOS cases, while arterial TOS is rare (Sanders et al., 2007; Illig et al., 2016). Nevertheless, in many cases the classification of TOS remains difficult with intricate signs of arterial and neurogenic compression (Likes et al., 2014; Ferrante and Ferrante, 2017). Inversely, there is a high rate of positive ultrasound or angiographic results (evidence of arterial compression during provocative manoeuvres) in asymptomatic patients (Plewa and Delinger, 1998; Nord et al., 2008). This underlines the interest of improving arterial investigations in patients with suspected TOS (Likes et al., 2014; Molina and D'Cunha, 2008). Ultrasound allows for simple, low-cost and accurate recordings and remains relatively simple. It is widely used despite some debate regarding its accuracy (Sobey et al., 1993; Nord et al., 2008; Stapleton et al., 2009). Nevertheless, Doppler (as would pulse plethysmography) provides evidence for flow impairment but not for the presence of ischemia itself. Indeed, even arterial positional non-occlusive compression can provide sufficient perfusion to result in no significant oxygen delivery deficit and cause no arterial ischemia (Hersant et al., 2022). There are potential advantages of using TcpO2 to diagnose an arterial compression and argue for arterial TOS. It enables direct monitoring of oxygen availability as one of the determinants of pain and of the decline in force during ischemia (Hogan et al., 1994; Hogan et al., 1999). TcpO2 can be used during dynamic tests to provide quantitative results of regional blood flow impairment (Abraham et al., 2018a). The specific interest in the present study was the ability to objectively monitor both upper arms and forearms throughout the period of, and in the recovery period from, the provocative tests. The DROP calculation was of specific interest here to account for eventual systemic pO2 changes (as it could result from increased arterial pO2 accompanying hyperventilation due to upper limb ischemic pain). The absolute pO2 measurements that we took fall within the range of previous studies that report a mean normal TcpO2 of 52–65 mmHg at the upper limb level (Manabe et al., 2004; Babilas et al., 2008; Blake et al., 2018). From a technical point of view, it could appear surprising that analysing TcpO2 at the upper limb in patients with suspected TOS had never been performed. This probably relies on the fact that for years TcpO2 results were expressed as regional perfusion index (RPI). The transcutaneous oxygen gradient makes absolute values only fairly reliable and results in RPI being of poor reproducibility. Indeed, this gradient is unpredictable, variable from one patient to another or one probe position to another in close proximity, but constant over time for a defined probe position. The specific interest of DROP is to account only for pO2 changes over time, and as a result, to be completely insensitive to the transcutaneous oxygen gradient. Other microvascular techniques such as laser doppler or near-infrared spectroscopy are potential candidates for the future, but to the best of our knowledge have never been tested in TOS.

As expected, arm elevation induces a decrease in TcpO2 that is roughly proportional to the degree of upper limb elevation. Blake et al. (2018) observed in patients in the lying position that the mean decrease in TcpO2 during upper limb elevation of 45° above its resting level was only 1 mmHg at the upper arm level, approximately 5 mmHg at the forearm level and in excess of 15 mmHg at the hand level, while chest TcpO2 remains almost unchanged. Therefore, with the average upper limb length in adults being 70 cm, the estimated change in TcpO2 was approximately 3 mmHg for a 10 cm elevation. The amplitude of elevation from the resting position in our study was estimated at 40–45 cm for the forearm probe but only approximately 10–15 cm for the upper arm probe. Subsequently, the ∼8 mmHg difference between U-DROPm and F-DROPm observed in the prediction of Imag + results is consistent with previously cited results (Blake et al., 2018). This also suggests that any F-DROPm higher than −10 mmHg during EAST tests is expected to result only from the normal fall in TcpO2 due to elevation of the probe above heart level. This is also consistent with the optimal cut-off point that we found in predicting the presence of subclavian arterial compression on arteriography (Abraham et al., 2020). From a clinical point of view, it is unlikely that an arterial compression that would result in no ischemia could be the cause of upper limb symptoms of ischemic origin. Consequently, we believe that, since it measures the presence or absence of ischemia, TcpO2 provides unique evidence for a causal relationship between the arterial compression and the symptoms of upper limb positional pain or weakness.

We noted that the DROPm values observed in patients with positive imaging were not as low as the ones observed in our previous experiments. Nevertheless, the diagnostic performances obtained in the present study and assessed by the AUROC are very close (at both the forearm and upper arm level) to those reported in our previous studies: 0.69 ± 0.06 for arteriography (Abraham et al., 2020) and 0.69 ± 0.04 for ultrasound (Henni et al., 2019). These two observations are very likely to result from the fact that the duration of the candlestick/surrender position of a Ca-Pra manoeuvre was limited to 30 s, which possibly reduced the severity of upper limb ischemia in cases of arterial compression. Due to the relatively slow responsiveness of TcpO2 to abrupt tissue pO2 changes, it is possible that an early reperfusion occurred, preventing the TcpO2 DROP values from decreasing further. It is clear that normalising the duration of the provocative manoeuvres allows for a better inter-individual comparison, but it might have reduced the diagnostic performance of the test by limiting the DROPm value resulting from impaired perfusion. Whether, at the individual level, the provocative manoeuvre should be as long as possible to enable better discrimination between the DROPm that would result from gravity changes that would remain close to zero and abnormal DROPm results, has yet to be determined.

From a physiological and practical point of view, the present work suggests that, when monitoring upper limb ischemia during provocative manoeuvres that include an upper limb elevation, one should account for lower DROPm values at the forearm than at the upper arm due to differences in the altitude. We believe that a measurable ischemia (estimated through DROPm calculation) may strengthen the responsibility of arterial compression in upper limb symptoms in patients with suspected TOS.

From a diagnostic point of view, the diagnosis of TOS is a holistic approach based on a combination of patients’ histories, clinical diagnostic testing and mainly ultrasound imaging. The cost-benefit of TcpO2 relative to the other diagnostic approaches used for patients with suspected TOS remains to be determined. Nevertheless, Tcpo2 might be critical in providing recordable evidence for the presence of ischemia because of arterial impaired inflow (as can be determined by radiological or ultrasound imaging). Whether portable multichannel TcpO2 devices might be of interest to confirm the presence of ischemia during usual professional or sports activities is a fascinating direction for future studies.

To the best of our knowledge, this is the first report comparing upper arm and forearm TcpO2 indices in patients with suspected TOS to show upper arm or forearm DROPm show similar diagnostic performance for detecting upper limb arterial compression and/or symptoms during provocative manoeuvres. Nevertheless, for detecting positional arterial compression or symptoms, and symptoms associated with arterial compression during arm abduction, different thresholds for U-DROPm or F-DROPm must be used.