fax: +34 93 227 54 04
Liver Failure and Liver Disease
Gas exchange mechanism of orthodeoxia in hepatopulmonary syndrome
Article first published online: 30 AUG 2004
Copyright © 2004 American Association for the Study of Liver Diseases
Volume 40, Issue 3, pages 660–666, September 2004
How to Cite
Gómez, F. P., Martínez-Pallí, G., Barberà, J. A., Roca, J., Navasa, M. and Rodríguez-Roisin, R. (2004), Gas exchange mechanism of orthodeoxia in hepatopulmonary syndrome. Hepatology, 40: 660–666. doi: 10.1002/hep.20358
- Issue published online: 30 AUG 2004
- Article first published online: 30 AUG 2004
- Manuscript Accepted: 21 MAY 2004
- Manuscript Received: 25 MAR 2004
- Red Respira-ISCIII-RTIC-03/11
- Fondo de Investigación Sanitaria. Grant Number: FIS-95/0752
- Generalitat de Catalunya. Grant Number: 2001SGR00286
- Laboratorios Dr. Esteve, SA
The mechanism of orthodeoxia (OD), or decreased partial pressure of arterial oxygen (PaO2) from supine to upright, a characteristic feature of hepatopulmonary syndrome (HPS), has never been comprehensively elucidated. We therefore investigated the intrapulmonary (shunt and ventilation-perfusion [V̇A/Q̇] mismatching) and extrapulmonary factors governing PaO2 in 20 patients with mild to severe HPS (14 males, 6 females; 50 ± 3 years old SE) at upright and supine, in random order. We set out a cutoff value for OD, namely a PaO2 decrease ≥5% or ≥4 mm Hg (area under the receiver operating characteristic curve, 0.96 each). Compared to supine, 5 patients showed OD (PaO2 change, −11% ± 2%, −7 ± 1 mm Hg, P < .05) with further V̇A/Q̇ worsening (shunt + low V̇A/Q̇ mode increased from 19% ± 7% to 21% ± 7% of cardiac output [Q̇T], P < .05), as opposed to 15 patients who did not (+2% ± 2%, +1± 1 mm Hg) with V̇A/Q̇ improvement (from 20% ± 4% to 16% ± 4% of Q̇T, P < .01). Cardiac output was significantly lower in OD patients in both positions. Changes in extrapulmonary factors at upright, such as increased minute ventilation and decreased Q̇T, were of similar magnitude in both subsets of patients. In conclusion, our data suggest that gas exchange response to OD in HPS points to a more altered pulmonary vascular tone inducing heterogeneous blood flow redistribution to lung zones with prominent intrapulmonary vascular dilatations. (HEPATOLOGY 2004;40:660–666.)
Hepatopulmonary syndrome (HPS), defined as a clinical triad including hepatic disease, arterial oxygenation defect, and intrapulmonary vascular dilatations,1 associate other characteristic features, such as low transfer factor for carbon monoxide (DLCO) and orthodeoxia (OD). OD (i.e., decreased arterial PO2 [PaO2] from supine to upright) is one of the hallmarks of HPS, with a wide prevalence ranging from 80%2 to 20%.3 Notwithstanding, the underlying mechanism of OD remains insufficiently elucidated. Further, OD cutoff values have never been validated. There is strong evidence that increased intrapulmonary shunt and ventilation-perfusion (V̇A/Q̇) mismatching are the predominant intrapulmonary determinants of arterial oxygenation in HPS.1, 3 It has been suggested that increased pulmonary blood flow to dependent, lower lung zones, when HPS patients adopt the upright position, results in increased intrapulmonary shunt, hence inducing OD.4 However, gas exchange pathophysiology is a complex process in which PaO2 is the end-primary outcome of the interplay between intrapulmonary, namely shunt, V̇A/Q̇ mismatch, diffusion impairment, and extrapulmonary (i.e., cardiac output, minute ventilation, and oxygen uptake) factors.5, 6 Accordingly, we investigated the positional responses of all determinants of PaO2 in patients with HPS to assess the mechanism of OD and propose an operational definition. Arterial deoxygenation, one of the best predictors of postoperative mortality after liver transplantation in HPS,7 is one major prognostic factor in the natural history of cirrhosis, particularly in HPS-induced hypoxemia.8 To the extent that OD in HPS could represent a more frail positional adaptive gas exchange process, it might become a relevant prognostic factor as a surrogate outcome of disease severity.
Patients and Methods
Twenty patients (14 males, 6 females; 50 ± 3 years old) with stable HPS were prospectively recruited over a period of 3 years. Fifteen of them were evaluated and diagnosed with HPS in the context of a work-up for elective orthotopic liver transplantation, while the remaining 5 patients were referred for investigation of dyspnea associated with chronic liver disease. Cirrhosis was confirmed in all but 1 patient (who had idiopathic portal hypertension) by clinical, biochemical, and echographic findings, and hepatic histology, whenever a liver biopsy specimen had been obtained. The degree of liver dysfunction was graded according to the Child-Turcotte-Pugh classification. HPS was defined when, in addition to liver disease and/or portal hypertension, patients showed (1) increased alveolar to arterial oxygen partial pressure gradient (AaPO2) ≥15 mm Hg (normal range, 4–8 mm Hg)9 while breathing room air and in the sitting position, irrespective of the presence of arterial hypoxemia; and (2) evidence of intrapulmonary vascular dilatations using transthoracic contrast-enhanced echocardiography.3 The study was approved by the Ethics Committee of the Hospital Clínic, and all patients gave written informed consent. Data on lung function and V̇A/Q̇ distributions at upright in 14 of the patients have been previously reported.3
Patients were studied at both upright (seated in an armchair) and supine positions in random order. Measurements were performed in all but 2 patients under steady-state conditions after they stayed 30 minutes in each position while breathing room air. The latter 2 patients, treated with long-term oxygen therapy, were studied while breathing 30% and 40% oxygen, and their AaPO2 was not included in the analysis. Maintenance of steady-state conditions was confirmed by stability (±5%) of both ventilatory and hemodynamic variables and also by the close agreement between duplicate measurement of mixed expired and arterial O2 and CO2 (within ±5%).
Pulmonary Function Tests.
V̇A/Q̇ distributions were estimated by the multiple inert gas elimination technique in the customary manner.9, 12 In 8 patients, in whom a pulmonary artery catheter was inserted (Swan-Ganz catheter, Baxter Healthcare, Irvine, CA), the multiple inert gas elimination technique was calculated using mixed venous inert gases, and cardiac output (QT) was determined by thermodilution. In the other 12 patients, Q̇T was measured by the dye dilution technique (DC-410, Waters Instruments, Rochester, NM), whereas both venous inert gases and partial pressure of mixed venous oxygen (Pv̄O2) were computed by mass balance. The agreement between these 2 different approaches has been consistently validated in our laboratory.13 The dispersion of pulmonary distribution blood flow on a logarithmic scale (Log SDQ) and dispersion of alveolar ventilation distributor on a logarithmic scale (Log SDV) on a logarithmic scale (upper normal limit, 0.60 and 0.65, respectively)9 and the difference among measured retentions and excretions of the inert gases corrected for the elimination of acetone (DISP R-E*) (normal values <3.0)12 were used as V̇A/Q̇ indices.
Intrapulmonary shunt and low V̇A/Q̇ mode were defined, respectively, as the fraction of blood flow perfusing lung units with V̇A/Q̇ ratios <0.005 and <0.1 (excluding shunt).
Pulmonary artery pressures were continuously monitored using a Swan-Ganz catheter on a multichannel recorder (HP-7754 B, Hewlett-Packard, Waltham, MA) in 8 patients. Two measurements were performed at each position, and the mean value was reported as the final result. Pulmonary and systemic vascular resistances were calculated using standard formulae.
All of the analyses were performed with version 10.0 of the SPSS statistical package (SPSS, Chicago, IL). Descriptive data are expressed as mean ± SEM and range, as appropriate. Comparisons were made with Wilcoxon rank sum test or Mann-Whitney test. Pearson or Spearman correlation was used when necessary to assess relationships between variables. A P value < .05 was considered to be significant in all instances.
Nine patients had hepatitis C cirrhosis, 6 alcoholic cirrhosis, 2 hepatitis C and alcoholic cirrhosis, 2 cryptogenic cirrhosis, and 1 extrinsic liver disease due to idiopathic portal hypertension with mild liver dysfunction. Three patients were in class A, 10 in class B, and 7 in class C of the Child-Turcotte-Pugh classification. Fourteen patients referred mild to severe dyspnea (mean duration, 1.9 ± 0.6 years; range, 0.6–4.5 years). Spider naevi were present in all patients and finger clubbing in 14 (70%). Four patients were smokers (22 ± 4 pack-years; forced expiratory volume in 1 second [FEV1], 85% ± 14% predicted; FEV1/forced vital capacity ratio, 71% ± 5%; total lung capacity, 97% ± 8% predicted), 9 were exsmokers (13 ± 3 pack-years; FEV1, 81% ± 5% predicted; FEV1/forced vital capacity ratio, 74% ± 3%; total lung capacity, 86% ± 4% predicted) and the remaining 7, nonsmokers with normal lung function (FEV1, 92% ± 4% predicted; FEV1/forced vital capacity ratio, 77% ± 4%; total lung capacity, 98% ± 6% predicted). All patients had an abnormal low transfer factor at upright (52% ± 4% predicted; range, 21%–79% predicted). Four patients had mild unilateral or bilateral pleural effusions.
Gas Exchange at Upright
Overall, patients were moderately hypoxemic (PaO2 <80 mm Hg) and hypocapnic (PaCO2 <35 mm Hg) (Table 1). Fourteen patients (70%) were moderately to severely hypoxemic (PaO2, 58 ± 3 mm Hg; 39–79 mm Hg; AaPO2, 53 ± 5 mm Hg; 22–72 mm Hg), and 6 were normoxemic (PaO2, 93 ± 2 mm Hg; 87–99 mm Hg) with an increased AaPO2 (22 ± 2 mm Hg; 16-28 mm Hg) only. Two of the most influential extrapulmonary factors governing gas exchange, namely minute ventilation and Q̇T, were mildly to moderately increased (abnormal) in most of the patients, whereas oxygen uptake was on average within normal limits (Table 1). Distributions of V̇A/Q̇ ratios were mildly to severely abnormal in all patients, as shown by increases in Log SDQ and in DISP R-E*. In addition, mild to moderate (<20% of Q̇T) (7% ± 2% of Q̇T, n = 16), or severe (≥20% of Q̇T) (40% ± 4% of Q̇T; n = 4) increases in intrapulmonary shunt and mild areas with low V̇A/Q ratio were observed (Table 1). The mean distribution of pulmonary blood flow (mean Q) was close to normal (∽1.0) or mildly decreased. By contrast, Log SDV was normal or slightly increased without areas of high V̇A/Q̇ ratio (>10.0, excluding dead space). The mean distribution of alveolar ventilation was normal or mildly to moderately increased. Dead space was normal or slightly reduced. Predicted (multiple inert gas elimination technique) PaO2 (73 ± 4 mm Hg; 43–111 mm Hg) was slightly but significantly higher than measured (actual) PaO2 (69 ± 4 mm Hg; 39–99 mm Hg) (P < .05), suggesting the coexistence of diffusion limitation to oxygen.3 An inverse correlation was shown between actual PaO2 and both intrapulmonary shunt (r = −0.74; P < .001) and DISP R-E* (r = −0.73; P < .001).
|PaO2, mm Hg||69 ± 4||(39–99)|
|PaCO2, mm Hg||30 ± 1||(21–38)|
|AaPO2, mm Hg||43 ± 5||(16–72)|
|Pv̄O2, mm Hg†||37 ± 1||(26–46)|
|V̇E, L · min−1||9.5 ± 0.6||(5.9–13.1)|
|Q̇T, L · min−1||7.9 ± 0.6||(3.3–14.9)|
|Cl, L · min−1 · m2||4.4 ± 0.3||(2.0–6.9)|
|V̇O2, mL · min−1||245 ± 12||(162–346)|
|Shunt, % QT‡||13.5 ± 3.3||(0–48.1)|
|Low V̇A/Q̇ (<0.1), % Q̇T§||3.8 ± 0.8||(0–11.8)|
|Shunt + Low V̇A/Q̇, % Q̇T||17.3 ± 3.4||(0.4–52.6)|
|Mean Q||0.84 ± 0.06||(0.30–1.44)|
|Log SDQ∥||1.05 ± 0.08||(0.55–1.77)|
|Mean V||1.48 ± 0.12||(0.60–2.52)|
|Log SDV¶||0.64 ± 0.03||(0.42–0.98)|
|DISP R-E* **||14.6 ± 2.0||(5.5–34.5)|
|Dead space, % of V̇A††||27 ± 2||(11–44)|
|PaO2-difference, mm Hg‡‡||4.7 ± 2.0||(−15–16)|
To determine the relative contribution of the changes of intrapulmonary shunt, Log SDQ, QT, minute ventilation, and Pv̇O2 in influencing position-induced changes in PaO2, a multiple forward stepwise regression analysis was performed. As a result, it was established that only the change in intrapulmonary shunt was independently related to the change in PaO2 (r2= 0.28; P < .05). Furthermore, we used a receiver operating characteristic analysis to determine the cutoff threshold for the PaO2 decrease that better predicted an upright-induced increased intrapulmonary shunt. Accordingly, OD was defined as a fall in PaO2 ≥5% at upright, or ≥4 mm Hg (area under the receiver operating characteristic curve, 0.96; sensitivity, 80%; specificity, 93% each). Using this threshold, 5 patients (3 males, 2 females) exhibited OD (% PaO2 change at upright, −11% ± 2%, or −7 ± 1 mm Hg), whereas the other 15 patients (11 males, 4 females) did not (non-OD) (+2% ± 2%, or +1 ± 1 mm Hg) (Table 2 and Fig. 1). Two OD patients referred platypnea. Although all 5 OD patients were hypoxemic at both positions, OD was not related to supine nor to upright PaO2. Nine non-OD patients (60%) showed similar levels of hypoxemia to OD patients at supine (Fig. 1, top panel).
|OD (n = 5)||Non-OD (n = 15)||P†|
|PaO2, mm Hg||67 ± 5||59 ± 6‡||70 ± 5||72 ± 5||<.001|
|PaCO2, mm Hg||31 ± 2||29 ± 2‡||31 ± 1||30 ± 1‡||NS|
|AaPO2, mm Hg||41 ± 8||50 ± 9‡||41 ± 6||40 ± 5||<.005|
|Pv̄O2, mm Hg§||37 ± 2||31 ± 2‡||41 ± 1||39 ± 1‡||<.05|
|V̇E, L · min−1||7.7 ± 0.7||8.5 ± 1.1‡||8.8 ± 0.6||9.9 ± 0.6∥||NS|
|Q̇T, L · min−1||6.2 ± 1.0¶||5.6 ± 0.9‡¶||10.0 ± 0.9||8.6 ± 0.7∥||NS|
|Cl, L · min−1 · m2||3.5 ± 0.6¶||3.2 ± 0.5‡¶||5.5 ± 0.4||4.8 ± 0.3∥||NS|
|V̇O2, mL · min−1||206 ± 16||206 ± 16||239 ± 9||258 ± 13||NS|
|Shunt, % Q̇T**||15 ± 7||17 ± 7||14 ± 4||12 ± 4∥||<.05|
|Low V̇A/Q̇, % Q̇T††||4.2 ± 1.2||4.6 ± 1.7||5.1 ± 1.7||3.6 ± 1.0||NS|
|Shunt + Low V̇A/Q̇, % Q̇T||19 ± 7||21 ± 7‡||20 ± 4||16 ± 4∥||<.001|
|Mean Q||0.79 ± 0.10||0.98 ± 0.16||0.63 ± 0.05||0.80 ± 0.07∥||NS|
|Log SDQ‡‡||1.06 ± 0.13||1.12 ± 0.11||1.12 ± 0.10||1.02 ± 0.10‡||<.05|
|Mean V||1.50 ± 0.08||1.84 ± 0.09‡¶||1.17 ± 0.11||1.36 ± 0.15∥||NS|
|Log SDV§§||0.65 ± 0.09||0.65 ± 0.09||0.67 ± 0.02||0.64 ± 0.02||NS|
|DISP R-E*∥∥||15.7 ± 4.0||17.5 ± 4.2||15.4 ± 2.4||13.6 ± 2.3∥||<.005|
|Dead space, % of V̇A¶¶||31 ± 2||33 ± 3||25 ± 3||26 ± 2||NS|
|PaO2-difference, mm Hg***||2.6 ± 4.0||0.1 ± 2.2||4.8 ± 2.4||6.2 ± 2.5||NS|
|DLCO, % predicted†††||58 ± 10||53 ± 8||58 ± 6||51 ± 5||NS|
Differences Between OD and Non-OD Patients
While the etiology and severity of liver disease of OD and non-OD patients were similar, OD patients tended to be older (58 ± 2 years; 52–62 years; non-OD patients, 47 ± 3 years; 24–62 years, P = .056). Table 2 sets out position-induced gas exchange changes within OD and non-OD populations and also differences in position-induced changes between each subset of patients. Except for a lower Q̇T (and cardiac index) at supine and upright and a higher mean distribution of alveolar ventilation at upright in OD patients, no other difference was shown between OD and non-OD patients at each position.
Compared to supine, both OD and non-OD patients showed at upright a small but significant decrease in PaCO2 and Pv̄O2; by contrast, AaPO2 increased in OD patients only. Similarly, there were increases in minute ventilation and decreases in Q̇T without changes in oxygen uptake in both subsets of patients. Changes in V̇A/Q̇ distributions at upright within OD and non-OD patients were different. Whereas in non-OD patients all V̇A/Q̇ disturbances ameliorated, altogether reflecting a more homogeneous V̇A/Q̇ pattern, in OD patients V̇A/Q̇ disturbances deteriorated, as expressed by the combined increase in both intrapulmonary shunt and areas with low V̇A/Q̇ ratio (shunt + low V̇A/Q̇ mode). Figure 2 further illustrates differences in position-induced changes of the major functional variables, shown in Table 2, between each subset of patients. Thus, OD patients had inferior PaO2 and Pv̄O2 and further V̇A/Q̇ worsening, while minute ventilation and cardiac output changes were close to those shown in non-OD patients. No differences were shown in the responses of both intrapulmonary and extrapulmonary factors to positional changes when patients were assessed according to the presence or absence of hypoxemia.
Overall, there was a hyperdynamic circulatory state at supine—i.e., high Q̇T and low systemic vascular resistance, with normal or mild reductions in both mean pulmonary artery pressure and pulmonary vascular resistance (<120 dynes/sec/cm−5) (Table 3). At upright, Q̇T decreased significantly and this was paralleled by a decrease in mean pulmonary artery pressure and an increase in systemic vascular resistance (Table 3 and Fig. 3). Due to the small number of patients, no differences in position-induced hemodynamic changes were observed between OD (n = 2) and non-OD (n = 6) patients.
|PvO2, mm Hg||41.2 ± 1.0||39.0 ± 2.0||< 0.05|
|QT, L · min−1||10.8 ± 1.5||8.9 ± 1.2||< 0.05|
|mPAP, mm Hg||13.2 ± 2.1||8.9 ± 1.9||< 0.05|
|PAOP, mm Hg||7.6 ± 1.9||4.6 ± 1.9||< 0.05|
|TPG, mm Hg||5.6 ± 0.7||4.4 ± 0.5||NS|
|PVR, dyne/sec/cm−5||50 ± 9||42 ± 7||NS|
|SVR, dyne/sec/cm−5||620 ± 91||740 ± 89||< 0.05|
Two are the novel findings of our study. Firstly, we were able to unravel thoroughly the pathophysiological mechanism that underlies OD in HPS patients, essentially characterized by further V̇A/Q̇ imbalance as assessed by increases in intrapulmonary shunt and in areas with low V̇A/Q̇ ratio. Secondly, we set out a cutoff value for OD, defined by a PaO2 decrease ≥5% or≥4 mm Hg from supine. These data extend and complement previous findings and provide a more comprehensive assessment of gas exchange during OD in HPS patients.
The notion that OD in HPS is related to an increased intrapulmonary shunt points to a more altered pulmonary vascular tone inducing more heterogeneous gravitational pulmonary blood flow redistribution to dependent lung zones, possibly with more pronounced vascular dilatations, as suggested by thoracic computed tomography scans.14 This mechanism was reinforced once we demonstrated that the positional effects on intrapulmonary and extrapulmonary determinants of oxygenation interplayed differently in OD and non-OD patients (Table 2). In non-OD patients, V̇A/Q̇ imbalance improved at upright in the face of increased minute ventilation and decreased cardiac output, both of which tend to reduce intrapulmonary shunt,15 hence increasing PaO2, but that effect is offset by decreased mixed venous PO2, the net result being a negligible change in PaO2. On the contrary, compared to non-OD patients, those with OD deteriorated V̇A/Q̇ inequality despite similar changes in minute ventilation and cardiac output, resulting in more intense arterial and mixed venous hypoxemia (Fig. 2). It is therefore conceivable that the redistribution of pulmonary blood flow in OD patients is more heterogeneous throughout the pulmonary vasculature and more dependent gravitationally than in non-OD patients, hence favoring further V̇A/Q̇ imbalance. Presumably, OD reflects a weaker pulmonary vascular tone, more rigid and fixed, due to more functional pulmonary vasculature abnormalities, which is less liable to proportionately accommodate gravitational blood flow changes to ventilation in dependent alveolar units. This is consistent with the finding that the pulmonary circulation of cirrhotic patients behaves paradoxically, combining a lower, or even absent, hypoxic vascular response but also some degree of hypoxic vasoconstriction release during 100% oxygen.16 One striking finding of our study was that cardiac output in OD patients at both positions was significantly lower than in non-OD patients, thereby suggesting that cardiac function may be more affected in OD patients.
We acknowledge, however, that there are some limitations in our study. The number of HPS patients was relatively small. The magnitude of OD changes was moderate, and we ignore whether these changes can be more relevant in more severe hypoxemic patients. Assessment of the hypoxic vascular response in HPS patients with abnormal gas exchange, however, to confirm the hypothesis of a more altered pulmonary vascular tone, can be life-threatening, given the severity of gas exchange disturbances.
All in all, these results are akin to early anecdotal findings of OD in patients with chronic liver disease.17 A similar pattern of upright-induced increased intrapulmonary shunt was shown in 3 OD patients with severe HPS. In another study,18 2 out of 9 patients with cirrhosis and severe hypoxemia, presumably having HPS, showed a marked increase in intrapulmonary shunt (assessed with the oxygen method) at upright. Originally, OD in HPS was defined as any reduction in PaO2 at upright.17 Subsequently, an arbitrary cutoff value was set as >10% decrease in PaO2, sometimes just referred as >10 mm Hg, irrespective of the inspired oxygen fraction,2, 17 a threshold never validated. The major differences observed during OD in HPS studies2–4, 17, 19 may be multifactorial, including differences in sampling while standing or sitting, lack of agreement in cutoff values for OD, and/or an ill-defined notion of OD while breathing 100% oxygen. In the era of evidence-based medicine, the operational definition of OD stays as one of the unmet needs in HPS.1 Akin to our findings, a cutoff value for PaO2 decline >4 mm Hg, to define OD in patients with pulmonary vascular malformations,20 and an upper normal limit of ≤3 mm Hg, for upright-induced PaO2 fall in healthy individuals,21 have been previously suggested.
In summary, we demonstrated that further increased intrapulmonary shunting emerges as the key gas exchange mechanism of OD in HPS, presumably in the context of a more altered pulmonary vascular tone. Furthermore, we recommend a PaO2 decline ≥5%, or ≥4 mm Hg, at upright to diagnose the presence of OD in HPS. Whether OD reflects a systemic circulatory widespread derangement or simply a disrupted pulmonary vascular network needs to be further investigated.
- 5Pulmonary and peripheral gas exchange and their interactions. In: RocaJ, Rodriguez-RoisinR, WagnerPD, eds. Pulmonary and peripheral gas exchange in health and disease. New York, NY: Marcel Dekker, Inc.; 2000: 1–25., , .