GEN

ERA

LT

HO

RA

CIC

Minute Ventilation-to-Carbon Dioxide Output(V̇E/V̇CO2) Slope Is the Strongest Predictor ofRespiratory Complications and Death AfterPulmonary ResectionAlessandro Brunelli, MD, Romualdo Belardinelli, MD, Cecilia Pompili, MD,Francesco Xiumé, MD, Majed Refai, MD, Michele Salati, MD, andArmando Sabbatini, MD

Divisions of Thoracic Surgery and Cardiology, Ospedali Riuniti, Ancona, Italy

Background. This study assessed whether the minuteventilation-to-carbon dioxide output (V̇E/V̇CO2) slope, ameasure of ventilatory efficiency routinely measuredduring cardiopulmonary exercise testing (CPET), is anindependent predictor of respiratory complications aftermajor lung resections.

Methods. Prospective observational analysis was per-formed on 225 consecutive candidates after lobectomy(197 patients) or pneumonectomy (28 patients) from2008 to 2010. Inoperability criteria were peak oxygenconsumption (V̇O2) of less than 10 mL/kg/min in asso-ciation with predicted postoperative forced expiratoryvolume in 1 second of less than 30% and diffusioncapacity of the lung for carbon monoxide of less than30%. All patients performed a symptom-limited CPETon cycle ergometer. Respiratory complications (30 daysor in-hospital) were prospectively recorded: pneumo-nia, atelectasis requiring bronchoscopy, respiratoryfailure on mechanical ventilation exceeding 48 hours,adult respiratory distress syndrome, pulmonaryedema, and pulmonary embolism. Univariable andmultivariable regression analyses were used toidentify independent predictors of respiratory

complications.

Address correspondence to Dr Brunelli, Via S. Margherita 23, 60129 An-cona, Italy; e-mail: brunellialex@gmail.com.

© 2012 by The Society of Thoracic SurgeonsPublished by Elsevier Inc

Results. Cardiopulmonary morbidity and mortalityrates were 23% (51 patients) and 2.2% (5 patients). The 25patients with respiratory complications had a signifi-cantly higher V̇E/V̇CO2 slope than those without compli-cations (34.8 vs 30.9, p � 0.001). Peak V̇O2 was notassociated with respiratory complications. Logistic re-gression and bootstrap analyses showed that, after ad-justing for other baseline and perioperative variables, thestrongest predictor of respiratory complications was V̇E/V̇CO2 slope (regression coefficient, 0.09; bootstrap fre-quency, 89%; p � 0.004). Patients with a V̇E/V̇CO2 slopeexceeding 35 had a higher incidence of respiratory com-plications (22% vs 7.6%, p � 0.004) and mortality (7.2% vs.0.6%, p � 0.01).

Conclusions. V̇E/V̇CO2 slope is a better predictor ofrespiratory complications than peak V̇O2. This inexpen-sive and operator-independent variable should be con-sidered in the clinical practice to refine operability selec-tion criteria.

(Ann Thorac Surg 2012;93:1802–6)

© 2012 by The Society of Thoracic Surgeons

The cardiopulmonary exercise test (CPET) has beenrecommended as a pivotal step in the preoperative

evaluation of lung resection candidates [1]. Althoughpeak oxygen consumption (V̇o2 peak) is the most usedvariable for patient selection, several other direct orderived indicators can be used for risk stratification andfor determining and possibly correcting deficits in theoxygen transport system. Among ventilatory variables,the ratio of minute ventilation (V̇e) to carbon dioxideoutput (V̇co2), defined as ventilatory efficiency (V̇e/V̇co2

Accepted for publication March 5, 2012.

Presented at the Poster Session of the Forty-eighth Annual Meeting ofThe Society of Thoracic Surgeons, Fort Lauderdale, FL, Jan 28–Feb 1, 2012.

slope), has been used with increasing frequency toclassify functional limitations in patients with heartdisease and to stratify clinical outcome in heart failurepatients [2– 6] and, most recently, in thoracic surgicalpatients [7].

The objective of this investigation was to verify the roleof the V̇e/V̇co2 slope in predicting respiratory complica-tions (RCs) after major lung resections.

Patients and Methods

This was a prospective observational study of 225 con-secutive candidates for lobectomy (197 patients) or pneu-monectomy (28 patients) at our institution from 2008 to2010. The study was approved by the local Institutional

Review Board, and all patients gave their consent to

0003-4975/$36.00doi:10.1016/j.athoracsur.2012.03.022

1803Ann Thorac Surg BRUNELLI ET AL2012;93:1802–6 V̇e/V̇co2 SLOPE AND LUNG RESECTION

GEN

ERA

LT

HO

RA

CIC

participate in the institutional prospective database anduse of their data for research and clinical purposes.

Patients were operated on by board-certified thoracicsurgeons through a muscle-sparing, nerve-sparing lat-eral thoracotomy [8]. Perioperative management fol-lowed standardized pathways of care. Patients wereusually extubated in the operating room and transferredto a dedicated thoracic ward.

Functional inoperability criteria were: peak V̇O2 of lessthan 10 mL/kg/min in association with a predicted post-operative (ppo) forced expiratory volume in 1 second(FEV1) of less than 30% and ppo diffusion capacity of thelung for carbon monoxide (Dlco) of less than 30%according to published guidelines [1, 9].

Postoperative treatment focused on early as possiblemobilization, chest physiotherapy, and physical rehabil-itation administered by a specialized physiotherapist.Thoracotomy chest pain was controlled by a continuoussystemic infusion of nonopioid drugs or paravertebralanalgesia with local analgesics to maintain the numericpain score below 4 in a scale ranging from 0 (no pain) to10 (excruciating pain).

Pulmonary AssessmentAll patients performed a preoperative (within 1 weekbefore their operation) symptom-limited CPET on anelectronically braked cycle ergometer using a ramp-pattern increase in work rate to reach exhaustion be-tween 8 and 12 min. Expired gases and volumes wereanalyzed, breath-by-breath, with a metabolic cart. CPETwas stopped when one or more of the following criteriawere present: fatigue, dyspnea, excessive systemic bloodpressure increase (ie, � 230/130 mm Hg), a 2-mm or moreST depression in at least 2 adjacent leads, or angina.

The following RCs occurring within 30 days of theoperation or during hospitalization were defined a prioriand prospectively recorded: pneumonia (chest roentgen-ogram infiltrates/consolidation, leukocytosis, fever), atel-ectasis requiring bronchoscopy, respiratory failure need-ing mechanical ventilation for longer than 48 hours, adultrespiratory distress syndrome (defined according to theAmerican-European consensus conference [10]), pulmo-nary edema, or pulmonary embolism (confirmed by V/Qscan or computed tomography scan).

In addition to the 225 patients considered for theanalysis, another 20 patients underwent major pulmo-nary anatomic resections during the same period butwere excluded because they were unable to perform aCPET because of logistic or severe incapacitatingcomorbidities.

Statistical AnalysisThe Shapiro-Wilk normality test was used to assessnormal distribution of numeric variables. Univariableanalysis was initially used to screen variables associatedwith RCs. The univariable comparisons of outcomes wereperformed with the unpaired Student t test for numericvariables with normal distribution and by using theMann-Whitney U test for numeric variables without a

normal distribution. Categoric variables were compared

using the �2 test or the Fisher exact test, wheneverappropriate.

The following variables were tested for a possibleassociation with postoperative RCs: age, sex, body massindex, FEV1 %, Dlco%, ppoFEV1 %, ppoDlco%, moder-ate to severe chronic COPD status (defined as FEV1 �80% � FEV1/FVC ratio � 0.7), peak V̇o2 (mL/kg/min),V̇e/V̇co2 slope, induction chemotherapy, pack-years ofsmoking, and type of operation. Variables with p of lessthan 0.1 at univariable analysis were used as indepen-dent predictors in a stepwise logistic regression analysis,with presence of RCs as the dependent variable.

All data were at least 95% complete. Sporadic missingdata were imputed by averaging the nonmissing values(numeric variables) or taking the most frequent responsecategory (categoric variables). To avoid multicollinearity,only one variable in a set of variables with a correlationcoefficient greater than 0.5 was selected (by bootstrapprocedure) and used in the regression model. For thepurpose of this investigation, V̇e/V̇co2 slope was ana-lyzed with peak V̇o2. The two variables were not corre-lated (r � 0.1). A value of p � 0.05 was selected forretention of variables in the final model.

The multivariable procedure was validated by boot-strap analysis with 1,000 samples [11–13]. In the bootstrapprocedure, repeated samples of the same number ofobservations as the original database were selected withreplacement from the original set observations. For eachsample, stepwise logistic regression was performed en-tering the variables with p less than 0.1 at univariableanalysis. The stability of the final model was assessed byidentifying the variables that entered most frequently inthe repeated bootstrap models and comparing thosevariables with the variables in the final stepwise model. Ifthe final model variables occurred in 50% of the boot-strap models, the original final stepwise regressionmodel was judged to be stable. A further analysis wasperformed using the V̇e/V̇co2 cutoff of more than 35,which has been shown to be associated with worseprognosis in patients with heart disease [14]. All statisti-cal tests were two-tailed and a significance level of 0.05was accepted. The analysis was performed on Stata 10software (Stata Corp, College Station, TX).

Results

The characteristics of the study patients are reported inTable 1. The overall cardiopulmonary morbidity andmortality rates were 23% (51 cases) and 2.2% (5 cases).RCs were present in 25 patients (11%) and had a signif-icantly higher V̇e/V̇co2 slope compared with those with-out complications (34.8 vs 30.9, p � 0.001). The 5 patientswho died had a higher value of V̇e/V̇co2 slope thansurvivors (36.3 vs 31.2, p � 0.07). Unlike patients withRCs, patients with cardiac complications (mainly atrialfibrillation) did not have a significantly higher value ofV̇e/V̇co2 slope compared with those without cardiaccomplications (31.8 vs 31.2, p � 0.6). In addition toV̇e/V̇co2, the following variables were also associated

with RC at univariable analysis: FEV1 % (75.6% vs 82.3%,

1804 BRUNELLI ET AL Ann Thorac SurgV̇e/V̇co2 SLOPE AND LUNG RESECTION 2012;93:1802–6

GEN

ERA

LT

HO

RA

CIC

p � 0.07), ppoFEV1 % (58.6% vs 64.3%, p � 0.07), presenceof moderate to severe COPD (17% vs 8%, p � 0.06).PeakV̇o2 (mL/kg/min or percentage of predicted value)was not associated with RCs or mortality in this series(Table 2).

Stepwise logistic regression and bootstrap analysesshowed that the only predictor reliably associated withRCs remained the V̇e/V̇co2 slope (regression coefficient,0.09; standard error, 0.03; bootstrap frequency, 89%; p �0.004). Compared with patients with a lower V̇e/V̇co2

slope value, those with a V̇e/V̇co2 slope exceeding 35 hada threefold higher rate of RCs (22% vs 7.6%, p � 0.004)

Table 1. Characteristics of the Study Patients

Variablesa Outcome

Age, years 67.2 (9.8)Male sex, No. (%) 183 (81)Body mass index, kg/m2 26.7 (4.7)FEV1 % 81.6 (18.1)Dlco% 76.4 (19.8)Predicted postoperative

FEV1, % 63.7 (14.9)Dlco, % 60 (16.7)

COPD, No. (%)b 70 (31)Pack-years 47.5 (34)Peak V̇o2, mL/kg/min 16.1 (3.7)Peak V̇o2 % 63.1 (16.5)V̇e/V̇co2 slope 31.3 (6.1)

a Results are expressed as means (standard deviation) unless otherwiseindicated. b Assessed as moderate to severe.

COPD � chronic obstructive pulmonary disease; Dlco�diffusioncapacity of the lung for carbon monoxide; FEV1 � forced expiratoryvolume in 1 second; V̇e/V̇co2 � expired volume–to–carbon dioxideoutput; V̇o2 � volume of oxygen consumption.

Table 2. Results of the Univariable Comparison Between Pati

WithVariablesa (n �

Age, years 68.8 (Body mass index, kg/m2 25.1 (FEV1 % 75.6 (Dlco% 74.3 (Predicted postoperative

FEV1 % 58.6 (Dlco% 59.3 (

Pneumonectomy, No. (%)c 3 (Induction chemotherapy, No (%)d 4 (COPD,e No. (%) 12 (Peak V̇o2, mL/kg/min 15.6 (Peak V̇o2, % 63 (V̇e/V̇Co2 slope 34.8 (

a Continuous variables are expressed as means (standard deviation) andpneumonectomy patients with or without respiratory complications.underwent induction chemotherapy. e Percentage of patients with m

COPD � chronic obstructive pulmonary disease; Dlco �diffusion capacityin 1 second; RCs � respiratory complications; V̇e/V̇co2 � expired volu

and 12-fold higher rate of mortality (7.2% vs 0.6%, p �0.01).

Figure 1 plots the V̇e/V̇co2 slope values against the RCrate. There is a direct relationship between the V̇e/V̇co2

slope values and the incidence of RC with a linearincrease of morbidity, which is particularly high withvalues above 35.

The interaction between V̇e/V̇co2 and RCs occurred inpatients without (complicated 34.4 vs noncomplicated31.1, p � 0.02) and with COPD (complicated 35.2 vsnoncomplicated 30.3, p � 0.03; Fig 2).

The difference between values of V̇e/V̇co2 slope inpatients with and without RCs was more pronounced inthe patients with peak V̇o2 exceeding 15 mL/kg/min(complicated 35.2 vs noncomplicated 29.8, p � 0.004)compared with those with lower peak V̇o2 (complicated34.3 vs noncomplicated 32.3, p � 0.2). Nevertheless,compared with patients with lower V̇e/V̇co2 slope values,the incidence of RCs in patients with V̇e/V̇co2 exceeding35 was more than double in both groups (peak V̇o2 � 15:21% vs 9%, p � 0.15; peak V̇o2 � 15: 27% vs 7%, p � 0.02),and the relationship between V̇e/V̇co2 slope and respi-ratory complications was similar (Fig 3).

Finally, we plotted the values of V̇e/V̇co2 slope againstthe postoperative length of stay, showing a linear directrelationship between them. Higher V̇e/V̇co2 slope valuescorrespond to a longer hospital stay, reflecting a morecomplicated and difficult postoperative course (Fig 4).

Comment

The CPET is increasingly used for stratifying the risk oflung resection candidates. A peak V̇o2 below 10 mL/kg/min or 35% predicted value has been indicated as aprohibitive threshold for major anatomic resection [1].

With and Without Respiratory Complications

Without RCs(n � 200) p Value

67 (9.9) 0.4b

26.9 (4.6) 0.2b

82.3 (18.2) 0.0776.6 (20) 0.6

64.3 (15.1) 0.0760 (17.1) 0.825 (13) 0.932 (16) �.9958 (28) 0.06

16.1 (3.8) 0.5b

63.1 (13.5) 0.8b

30.9 (6.1) 0.001b

oric variables as indicated. b Mann-Whitney test. c Percentage ofercentage of patients with or without respiratory complications whote to severe COPD with or without respiratory complications.

ents

RCs25)

8.4)4.8)16.8)18)

12.8)13)12)16)48)3.4)14.5)5.5)

categd P

odera

of the lung for carbon monoxide; FEV1 � forced expiratory volumeme–to–carbon dioxide output; V̇o2� volume of oxygen consumption.

1805Ann Thorac Surg BRUNELLI ET AL2012;93:1802–6 V̇e/V̇co2 SLOPE AND LUNG RESECTION

GEN

ERA

LT

HO

RA

CIC

Although peakV̇o2 is certainly the most widely usedvariable, CPET provides several other direct or indirectindicators that change in response to incremental work-loads. This not only allows assessment of over-allcardiopulmonary reserves, but also, in case of a limi-tation of exercise capacity, the detection of its cause,such as pulmonary, cardiovascular, or musculoskeletallimitations [15].

Among the ventilatory expired gas variables, an abnor-mally high relationship between minute ventilation (V̇e)and carbon dioxide output (V̇co2), expressed as theV̇e/V̇co2 slope, has been recently associated with pooroutcome in patients with chronic heart failure [4]. A highV̇e/V̇co2 reflects a combination of factors that underlieventilatory inefficiency. Although many reports havedemonstrated the independent prognostic role of V̇e/V̇co2 in patients with heart disease, scant information isavailable about its utility in those with normal heartefficiency. In particular, only one report investigated therole of V̇e/V̇co2 slope in COPD candidates for lung

Fig 1. Relationship between the level of the minute ventilation-to-carbon dioxide output (V̇e/V̇co2) slope and the incidence of respira-tory complications.

Fig 2. Relationship between the level of the minute ventilation-to-carbon dioxide output (V̇e/V̇co2) slope and the incidence of respira-tory complications in patients with (dotted line) and without (solid

line) chronic obstructive pulmonary disease (COPD).

resection [7]. These authors found that a higher V̇e/V̇co2

slope was associated with an increased mortality rate butdid not find a significant association of this parameterwith cardiopulmonary morbidity.

The objective of this study was to verify the role ofV̇e/V̇co2 slope value in predicting RCs after major pul-monary resections. We found that the V̇e/V̇co2 sloperemained the only significant factor associated with RCsafter controlling the effect of other confounders in astepwise logistic regression analysis. In particular, byusing the cutoff value of more than 35, found by otherauthors to be predictive of adverse outcome in patientswith chronic heart failure [14, 16], we found that patientswith V̇e/V̇co2 slope exceeding 35 had an incidence of RCsand mortality that were threefold and 12-fold higher thanthose with a lower V̇e/V̇co2 slope, respectively.

The association of ventilatory efficiency with the risk ofRCs occurred both in patients with and without moderate

Fig 3. Relationship between the level of the minute ventilation-to-carbon dioxide output (V̇e/V̇co2) slope and the incidence of respira-tory complications in patients with peak oxygen consumption (V̇o2)greater (solid line) or lower (dashed line) than 15 mL/kg/min.

Fig 4. Relationship between the level of the minute ventilation-to-carbon dioxide output (V̇e/V̇co ) slope and postoperative length of

2

stay.

1806 BRUNELLI ET AL Ann Thorac SurgV̇e/V̇co2 SLOPE AND LUNG RESECTION 2012;93:1802–6

GEN

ERA

LT

HO

RA

CIC

to severe COPD. This expands the findings of Torchioand colleagues [7], who found this factor was associatedwith death only in patients with COPD. The V̇e/V̇co2

slope appears to be a useful marker of some underlyingventilatory alterations that become more evident duringexercise, irrespective of the baseline pulmonary function.

Another interesting finding of our study was that theassociation between a high V̇e/V̇co2 slope and the risk ofRCs was present not only for patients with reducedaerobic capacity but even, and perhaps in a more pro-nounced way, at higher levels of peakV̇o2 (� 15 mL/kg/min). This confirmed previous studies in patients withheart failure, suggesting a differential interpretation ofthese two ergometric variables [4]. V̇o2 level is influencedby the combined contribution of the heart, lungs, theoxygen transport system, and skeletal muscles to externalwork. In this regard, it is a more global variable thanV̇e/V̇co2 slope alone, which is specifically the expression ofventilatory efficiency. In this regard, the V̇e/V̇co2 slope mayrepresent a useful indicator to refine risk-stratification in-dependent of the peak V̇o2 level.

The main limitation of this study is its retrospectivenature, which may have introduced inherent selectionbiases. However, the V̇e/V̇co2 slope value was not usedin any way to select patients for operation, minimizingthe risk of an influence on the results. Furthermore, theRCs were defined a priori and recorded prospectively inan electronic database.

In conclusion, we found that V̇e/V̇co2 slope is a betterpredictor of RCs and death than peak V̇o2. This variableis routinely calculated during CPETs and may be used,along with peak V̇o2, to refine operability selectioncriteria.

References

1. Brunelli A, Charloux A, Bolliger CT, et al; European Respi-ratory Society and European Society of Thoracic Surgeonsjoint task force on fitness for radical therapy. ERS/ESTSclinical guidelines on fitness for radical therapy in lungcancer patients (surgery and chemo-radiotherapy). Eur Re-spir J 2009;34:17–41.

2. Arena R, Myers J, Williams MA, et al. Assessment of func-

tional capacity in clinical and research settings. A scientificstatement from the American heart association committee

on exercise, rehabilitation, and prevention of the council onclinical cardiology and the council on cardiovascularnursing. Circulation 2007;116:329–43.

3. Myers J. Applications of cardiopulmonary exercise testing inthe management of cardiovascular and pulmonary disease.Int J Sports Med 2005;26(Suppl 1):S49–55.

4. Corra’ U, Mezzani A, Bosimini E, Giannuzzi P. Cardiopul-monary exercise testing and prognosis in chronic heartfailure: a prognosticating algorithm for the individual pa-tient. Chest 2004;126:942–50.

5. Arena R, Myers J, Aslam SS, Varughese EB, Peberdy MA.Peak VO2 and VE/VCO2 slope in patients with heart failure:a prognostic comparison. Am Heart J 2004;147:354–60.

6. Tumminello G, Guazzi M, Lancellotti P, Pie=rard LA. Exer-cise ventilation inefficiency in heart failure: pathophysiolog-ical and clinical significance. Eur Heart J 2007;28:673–8.

7. Torchio R, Guglielmo M, Giardino R, et al. Exerciseventilatory inefficiency and mortality in patients withchronic obstructive pulmonary disease undergoing sur-gery for non small cell lung cancer. Eur J CardiothoracSurg 2010;38:14 –9.

8. Cerfolio RJ, Bryant AS, Maniscalco LM. A non dividedintercostal muscle flap further reduces pain of thoracotomy:a prospective randomized trial. Ann Thorac Surg 2008;85:1901–6.

9. Brunelli A, Salati M. Preoperative evaluation of lung cancer:predicting the impact of surgery on physiology and qualityof life. Curr Opin Pulm Med 2008;14:275–81.

10. Bernard GR, Artigas A, Brigham KL, Carlet J, Falke K,Hudson L, Lamy M, Legall JR, Morris A, Spragg R. TheAmerican-European Consensus Conference on ARDS.Definitions, mechanisms, relevant outcomes, and clinicaltrial coordination. Am J Respir Crit Care Med 1994;149:818 –24.

11. Blackstone EH. Breaking down barriers: helpful break-through statistical methods you need to understand better.J Thorac Cardiovasc Surg 2001;122:430–9.

12. Grunkemeier GL, Wu YX. Bootstrap resampling method:something for nothing? Ann Thorac Surg 2004;77:1142–4.

13. Brunelli A, Rocco G. Internal validation of risk models inlung resection surgery: bootstrap versus training and testsampling. J Thorac Cardiovasc Surg 2006;131:1243–7.

14. Corrà U, Mezzani A, Bosimini E, Scapellato F, Imparato A,Giannuzzi P. Ventilatory response to exercise improves riskstratification in patients with chronic heart failure and inter-mediate functional capacity. Am Heart J 2002;143:418–26.

15. Ferguson MK, Lehman AG, Bolliger CT, Brunelli A. The roleof diffusing capacity and exercise tests. Thorac Surg Clin2008;18:9–17.

16. Chua TP, Ponikowski P, Harrington D, et al. Clinical corre-

lates and prognostic significance of the ventilatory responseto exercise in CHF. J Am Coll Cardiol 1997;29:1585–90.