Here, we showed that, when present, true RL shunting is related to intracardiac shunt in only a minority of patients. Hence, intracardiac RL shunting (through reopening of a patent foramen ovale or atrial septal defect) was present in only 12% of the patients, despite a shunt fraction higher than physiological values (5%) and elevated AaPO2 in all patients. This frequency is comparable to that found in the general population [18, 19], and in previous studies in PAH (18%)  or in precapillary PH (26%) . Contrast echocardiography was performed by experienced operators involved in our previous studies on RL shunting [15, 21]. Although transesophageal echocardiography was not performed, it is unlikely that important intracardiac shunt may have been missed by transthoracic contrast echocardiography, which has excellent sensitivity when the second harmonic mode is used [22, 23]. Thus, we consider that RL shunting and elevated AaPO2 were not explained by intracardiac shunt in most patients.
Our findings are relevant for the clinical management of patients with precapillary PH, whatever the etiologic group. Because hypoxemia is a potent pulmonary vasoconstrictor, and can contribute to the progression of PH, it is recommended to maintain oxygen saturation at > 90% at all times . However, hypoxemia related to RL shunting is poorly improved by supplemental oxygen therapy. Therefore, maintaining the adequacy of oxygenation, as recommended by clinical practice guidelines , may be difficult in patients with PH and true RL shunting. In addition, our results challenge the clinical utility of contrast echocardiography in patients with hypoxemia related to precapillary PH.
Our study included patients with most categories of the etiologic spectrum of precapillary PH, with no marked differences, thus showing that our findings are not restricted to any etiologic subgroup. Since the frequency and severity of RL shunting did not significantly differ between etiologic groups, and because no known cause of shunting other than PH was present, we consider that RL shunting was more likely related to precapillary PH than to the associated disease when present. It cannot be excluded that ventilation/perfusion mismatch participated to hypoxemia. Since Qs/Qt was greater than 5%, and AaPO2 did not correlate with cardiac output, it is unlikely that increased AaPO2 was due to low cardiac output. Interestingly, RL shunting (with increased AaPO2 and a median of Qs/Qt of 19%) was previously reported in 8 patients with severe PH associated with chronic obstructive pulmonary disease (with mPAP higher than 40 mmHg, "disproportionate" to the lung disease), with no evidence of intracardiac shunting at contrast echocardiography . PH was moderate or severe in 88% of our cases. Whether treatment of PH affects RL shunting and hypoxemia remains to be determined, however some improvement of AaPO2 was observed with treatment of PH in few patients.
The pathophysiology of RL shunting and increased AaPO2 in our patients remains largely unknown. RL shunting was higher than physiological shunting, which represents less than 5% of cardiac output . Transthoracic contrast echocardiography reportedly has excellent sensitivity for the detection of intrapulmonary shunt . Experimental studies in normal humans and dogs have shown increased RL shunting at exertion demonstrated by elevated AaPO2, positive transthoracic contrast echocardiography, and isotope-labeled microspheres, in proportion to the increase of cardiac output [28–30], especially under hypoxic conditions , although with unclear consequences on PaO2. Studies in infants  and adults  have demonstrated intrapulmonary arteriovenous shunts, with up to 200 μm in diameter. Large-diameter (> 25 μm) intrapulmonary arteriovenous pathways may be recruited with physiological exercise , thereby limiting the rise in PAP despite cardiac output increase . In patients with PAH, dilated and distorted capillary circulation were assumed to reflect collateral flow around obliterated pulmonary arterial segments . Intrapulmonary shunting in PH may be regulated by pulmonary vascular pressure and flow , and may take place at the capillary level, the diameter of which may be higher than normal (7–11 μm) but small enough to prevent the transit of microbubbles (60–90 μm). RL shunting may further increase at exercise in patients with PH . Thus, RL shunting in patients with PH might represent shunting through intrapulmonary arteriovenous pathways recruited with increase in microvascular pressure, similar to mechanisms seen during physiological exercise [30, 31] or hepatopulmonary syndrome . Alternatively, it might be due to increase in complex anatomic anastomosis of bronchial and/or pleural circulation with the pulmonary circulation, as suggested in PH  and especially chronic thromboembolic PH .
Our study had limitations, including its retrospective design, and heterogeneity of causes of PH with potential selection bias. Although patients with various causes of precapillary PH were included, the distribution of causes of PH was similar in patients with or without RL shunting within the overall population of patients with PH in our center, and we consider that similar results would have been obtained had the patient population been restricted to PAH. We could not determine whether occurrence of RL shunting was related to more severe hemodynamic parameters, although RL shunting was observed mostly in patients with moderate to severe PH. The effect of exercise on RL shunting was not evaluated. Contrast echocardiography was not performed in normoxic patients.