610 CAPNOGRAPHY Salen et al. • SONOGRAPHY AND CAPNOGRAPHY TO PREDICT OUTCOMES

CLINICAL INVESTIGATIONS

Can Cardiac Sonography and Capnography Be UsedIndependently and in Combination to Predict

Resuscitation Outcomes?

PHILIP SALEN, MD, ROBERT O’CONNOR, MD, PAUL SIERZENSKI, MD,BRENT PASSARELLO, MD, DIANA PANCU, MD, SCOTT MELANSON, MD,

STEVEN ARCONA, PHD, JAMES REED, PHD, MICHAEL HELLER, MD

Abstract. Objective: To measure the ability of car-diac sonography and capnography to predict survivalof cardiac arrest patients in the emergency depart-ment (ED). Methods: Nonconsecutive cardiac arrestpatients prospectively underwent either cardiac ul-trasonography alone or in conjunction with capnog-raphy during cardiopulmonary resuscitation at twocommunity hospital EDs with emergency medicineresidency programs. Cardiac ultrasonography wascarried out using the subxiphoid view during pausesfor central pulse evaluation and end-tidal carbon di-oxide (ETCO2) levels were monitored by a main-stream capnograph. A post-resuscitation data collec-tion form was completed by each of the participatingclinicians in order to assess their impressions of thefacility of performance and benefit of cardiac sonog-raphy during nontraumatic cardiac resuscitation. Re-

sults: One hundred two patients were enrolled overa 12-month period. All patients underwent cardiacsonographic evaluation, ranging from one to fivescans, during the cardiac resuscitation. Fifty-threepatients also had capnography measurements re-corded. The presence of sonographically identifiedcardiac activity at any point during the resuscitation

was associated with survival to hospital admission,11/41 or 27%, in contrast to those without cardiac ac-tivity, 2/61 or 3% (p < 0.001). Higher median ETCO2

levels, 35 torr, were associated with improvedchances of survival than the median ETCO2 levels fornonsurvivors, 13.7 torr (p < 0.01). The multivariatelogistic regression model, which evaluated the com-bination of cardiac ultrasonography and capnogra-phy, was able to correctly classify 92.4% of the sub-jects; however, of the two diagnostic tests, onlycapnography was a significant predictor of survival.The stepwise logistic regression model, summarizedby the area under the receiver operator curve of 0.9,furthermore demonstrated that capnography is anoutstanding predictor of survival. Conclusions: Boththe sonographic detection of cardiac activity andETCO2 levels higher than 16 torr were significantlyassociated with survival from ED resuscitation; how-ever, logistic regression analysis demonstrated thatprediction of survival using capnography was not en-hanced by the addition of cardiac sonography. Key

words: cardiac sonography; capnography; prediction;survival; outcomes; cardiac arrest. ACADEMICEMERGENCY MEDICINE 2001; 8:610–615

THE ABSENCE of a palpable central pulsedoes not always reflect cardiac standstill dur-

ing cardiac resuscitation. Inefficient cardiac con-tractions occur as a result of many factors, includ-ing pericardial tamponade, pulmonary embolism,and pneumothorax. Because some causes of pulse-lessness are more reversible than others, the early

From St. Luke’s Hospital Emergency Medicine Residency (PS,DP, SM, SA, JR, MH), Bethlehem, PA; and Christiana CareHealth System (RO, PS, BP), Newark, DE.Received September 28, 2000; revision received December 29,2000; accepted January 9, 2001. Presented at the ACEP annualmeeting, Philadelphia, PA, October 2000.Address for correspondence and reprints: Philip Salen, MD, St.Luke’s Hospital, 801 Ostrum Street, Bethlehem, PA 18015.e-mail: salenp@slhn.orgA related commentary appears on page 654.

identification of groups of pulseless patients withhigher likelihood of successful resuscitations mightbe useful. Research has suggested that cardiacsonography may be beneficial in identifying pa-tients with potentially treatable conditions andtherefore better prognosis when central pulses arenot palpable.1–6 Since end-tidal carbon dioxide(ETCO2) levels are dependent on the cardiac out-put generated by cardiopulmonary resuscitation incardiac arrest victims, capnography represents an-other potential predictor of survival or death aftercardiac arrest.7 Both cardiac sonography and cap-nography are well suited to bedside emergency de-partment (ED) practice in the care of critically illpatients because both tests are rapidly performed,noninvasive, and easily available. The objectives ofthis study were to determine the feasibility of per-

ACADEMIC EMERGENCY MEDICINE • June 2001, Volume 8, Number 6 611

Figure 1. Cardiac sonographic images.

forming cardiac sonography during resuscitationand to evaluate the utility of cardiac sonographyand capnography, both separately and together, aspredictors of successful resuscitation of pulselesspatients.

METHODS

Study Design. This was a prospective clinical ob-servation study of nonconsecutive cardiac arrestpatients undergoing cardiac ultrasonography aloneor in conjunction with capnography. The goal wasto measure the ability of these noninvasive tech-niques to predict survival after cardiac arrest. In-formed consent was waived because of the patients’critical condition and the noninvasive character ofcardiac sonography and capnography.

Study Setting and Population. This study wasconducted at two community hospital EDs withemergency medicine residency programs. PulselessED patients were enrolled in a prospective andnonconsecutive fashion. The minimal required ex-perience for performing ultrasonography on studyparticipants was a four-hour trauma sonographycourse during which focused cardiac sonography,using the subxiphoid view, was taught and prac-ticed on a series of different human and manne-quin simulator models. Participating emergencyphysicians, residents and attendings, performedand interpreted the cardiac sonography studiesduring the pulse check pause of the Advanced Car-diac Life Support (ACLS) algorithm.8

Study Protocol. Focused cardiac sonographywas performed primarily by the subxiphoid view;the apical view was used as an adjunct in obesesubjects (Fig. 1). St. Luke’s Hospital physiciansused the Pie Medical Scanner 200 (Pie Medical,Maastricht, The Netherlands) and Christiana Hos-pital physicians used the General Electric RT3200

Advantage II (General Electric, Milwaukee, WI);both used a 3.5-MHz curvilinear transducer. Thecapnography monitor, available only at ChristianaHospital, was a mainstream capnograph with in-frared analysis of ETCO2 that provided quantita-tive capnographic waveforms and measurements.Patient data, including cardiac rhythm, sono-graphic findings, impression of the clinical utilityof cardiac sonography, and capnography results,were recorded. The study protocol recommendedthat cardiac sonography should be performed inless than 10 seconds and that it should not inter-fere with ACLS-mandated interventions. Cardiacsonography was carried out during the pulse eval-uation and with any change in cardiac electricalrhythm; the number of scans was determined bythe progress of the resuscitation. Capnographywas performed upon patient arrival in the ED ifthe subject was an out-of-hospital-intubated car-diac arrest victim, or immediately after intubationin the ED. Capnography levels were noted simul-taneous to cardiac sonography exams; only peakETCO2 levels are reported.

Data Analysis. Chi-squared analysis was usedto compare survival rates in groups with and with-out sonographically detected cardiac activity. Sur-vival was defined as survival to admission to thehospital. The Mann-Whitney test was used to com-pare the differences in medians between capnog-raphy data for survivors and nonsurvivors. Thepoint-biserial correlation coefficient tested thestrength of association between the capnographicdata and survival. Alpha was set at 0.05, two-tailed. Multivariate logistic regression, which eval-uated cardiac ultrasonography and ETCO2 results,was used to construct a multivariate model pre-dicting survival to hospital admission based onETCO2 levels and the presence of sonographicallyidentified cardiac activity. Stepwise multivariatelogistic regression generated significant predictors

612 CAPNOGRAPHY Salen et al. • SONOGRAPHY AND CAPNOGRAPHY TO PREDICT OUTCOMES

Figure 2. Box-and-whisker chart of the end-tidal carbondioxide (CO2) median concentration level correlated withsurvival.

TABLE 1. Data Summary of Cardiac Sonography and Capnography Correlated with Survival*

1Cardiac Activity

PEA Asystole V Tach V Fib

2Cardiac Activity

PEA Systole V Tach V FibEnd-Tidal

CO2

DiedSurvived

238

30

12

31

231

321

20

20

13.735

*CO2 = carbon dioxide; PEA = pulseless electrical activity; V Tach = ventricular tachycardia; V Fib = ventricular fibrillation.

of survival from a set of candidate predictors tocreate a receiver operating characteristic (ROC)curve.

RESULTS

Table 1 documents the cardiac rhythms, the pres-ence of cardiac activity, the survival status, and thecapnography data of the 102 subjects enrolled over12 months. Subjects underwent a mean of 1.8 car-diac ultrasound scans during the resuscitation,with a range of one to five examinations. Patientswith sonographically identified cardiac activity atany time during the resuscitation were more likelyto survive to hospital admission, (11/41 or 27%)than those without cardiac activity (2/61 or 3%; x2

= 12.2; f = 0.35; p < 0.001). Among pulseless elec-trical activity (PEA) patients with no sonographicdetection of contractions, only one of 24 (4%) sur-vived to admission compared with eight of 31 (26%)of patients with contractions (x2 = 4.6; f = 0.3; p< 0.05). Sonographic detection of cardiac activityon the first sonographic evaluation in the resusci-tation was associated with survival (12/39 or 31%),as compared with those without cardiac activity(1/63 or 2%; x2 = 18.5; f = 0.6; p < 0.001). Of pa-tients with sonographically identified contractionspresent throughout resuscitation, (i.e., every timethe heart was sonographically identified, it wascontracting), 100% (6/6) survived to admission. Of

those who lacked contractions on at least one son-ographic evaluation, 7.0% (6/86) survived to hos-pital admission (x2 = 42.8; f = 0.7; p < 0.001). Twopatients (1 PEA, 1 asystole) without sonographi-cally identified contractions survived to admission.Cardiac sonography, used eight times to confirm orrefute transcutaneous or transvenous pacemakercapture, confirmed pacemaker capture of cardiacelectrical activity in one subject and refuted cap-ture in seven subjects. Of 11 subjects who werenoted to have either ventricular tachycardia orventricular fibrillation, seven had sonographicallyidentified cardiac activity. Pericardial effusionswere identified in four subjects.

The median ETCO2 level for survivors (39 torr)was significantly elevated compared with the me-dian ETCO2 level for nonsurvivors (13.7 torr; Z =3.26; p < 0.01). The point-biserial correlation be-tween ETCO2 levels and survival was 0.53. Figure2 is a box-and-whisker plot that demonstrates theassociation between ETCO2 and survival. Althoughelevations of up to 48 torr were noted in nonsur-vivors, no patient with an ETCO2 less than 16 torrsurvived. Of 22 subjects with ETCO2 levels lessthan 16 torr and the absence of sonographicallydetectable cardiac activity, none survived; sevensubjects with ETCO2 less than 16 torr had cardiacactivity detected and none survived. Of 19 subjectswith ETCO2 higher than 16 torr and sonographi-cally detectable cardiac activity, five survived; fivesubjects with ETCO2 higher than 16 torr had nosonographically detectable cardiac activity, onlyone survived.

Results of the multivariate logistic regressionmodel reveal that sonographically identified car-diac activity was not a significant predictor of sur-vival (odds ratio 1.09; 95% CI = 0.07 to 16.4). How-ever, ETCO2 was a significant predictor to hospitaladmission. For each increase of 1 torr in ETCO2,the odds of surviving increased by 16% (odds ratio1.16; 95% CI = 1.05 to 1.29). The model was ableto correctly classify 92.4% of the subjects as a pre-dicted probability cut value of 0.50 (sensitivity50%; specificity 98%; positive predictive value 75%;negative predictive value 94%). Summary statis-tics for the entire model suggest that it fit the datawell (Hosmer-Lemeshow goodness of fit = 4.8; p =0.8; McFadden r2 = 0.4). The results of the step-wise logistic regression model are summarized by

ACADEMIC EMERGENCY MEDICINE • June 2001, Volume 8, Number 6 613

Figure 3. Receiver operating characteristic (ROC)curve.

Figure 4. Clinicians’ impressions of the difficulty of per-forming cardiac sonography.

the ROC curve in Figure 3, with an area under thecurve of 0.9, which demonstrates that ETCO2 is anoutstanding predictor of survival.

Clinicians reported successfully identifying theheart sonographically in all 102 cases. Figure 4demonstrates the self-reported degree of the diffi-culty in identifying the heart sonographically. Ona one-to-five Likert scale (1 = easy to 5 = difficult),the performance of cardiac sonography was gradedas 1 or 2 in 73% of cases. On the post-resuscitationquestionnaire, clinician participants reported car-diac sonography to be helpful in their managementof 96% of cases and unhelpful in 4%.

DISCUSSION

An increasing body of evidence suggests that cap-nography can be used to predict survival or to de-termine when ACLS can be discontinued.7,9–11

There is also evidence suggesting that cardiac son-ography may be a useful adjunct in cardiopulmo-nary resuscitation and in terms of predicting sur-vival.1,2,12 This study was conducted to determinewhether cardiac sonography either alone or incombination with capnography in pulseless pa-tients undergoing resuscitation is predictive of sur-vival.

Overall, sonographically detected cardiac activ-ity was noted in 40% (41/102) of pulseless subjects(Table 1). The sonographic identification of cardiacactivity was associated with increased survival asopposed to those without sonographically identi-fied cardiac activity (p < 0.001). The presence ofcardiac activity on the first sonographic evaluationwas also associated with survival (p < 0.001). Fur-thermore, there was a moderately strong associa-tion between survival and the presence of sono-graphically detected contractions throughout theresuscitation as opposed to those who had one ormore episodes in which cardiac sonography did notdetect activity (p < 0.001). Previous research hasdemonstrated that as many as 87% of patients inPEA have myocardial wall and valve motion.3 Ourdata confirm that a large percentage, 56%, of sub-jects with PEA do have cardiac activity and thesePEA victims may be the most amenable to resus-citative therapy; the association between survivaland sonographically detected cardiac activity inPEA subjects as compared with no cardiac activitywas significant (p < 0.05). Our data support thehypothesis that the demonstration of cardiacstandstill is associated with an unsuccessful resus-citation, whereas the detection of cardiac output isassociated with an improved likelihood of sur-vival.1,13

Cardiac sonography may have other uses in theevaluation of cardiac arrest patients. There havebeen case reports regarding cardiac sonography in

the evaluation of pacemaker capture of cardiacelectrical activity, in the evaluation of pericardialeffusion in pulseless electrical tachycardia pa-tients, and in the detection of ventricular fibrilla-tion.2,4–6 In our study, cardiac sonography refutedpacemaker capture in seven cases and confirmedcapture in one case by either transcutaneous ortransvenous pacemaker. Historically, postmortempresence of pericardial effusion was found in 1% to14% of pulseless patients who underwent cardio-pulmonary resuscitation for nontraumatic cardiacarrest4,5; we identified a rate of 4% of pericardialeffusion in pulseless victims. Finally, of 11 victimsof pulseless ventricular tachycardia or ventricularfibrillation, seven had sonographically detectedcardiac activity (Table 1). This finding of sono-graphically detected cardiac activity in ventricularfibrillation patients adds credence to two case re-ports in which echocardiography identified cardiacactivity in subjects with ventricular fibrillation.6,12

614 CAPNOGRAPHY Salen et al. • SONOGRAPHY AND CAPNOGRAPHY TO PREDICT OUTCOMES

Higher ETCO2 values have been associatedwith return of spontaneous circulation.7,11,14 Ourcapnography results (Table 1) add to this evidencebase by demonstrating that higher ETCO2 valuesare associated with better prognosis. MedianETCO2 for survivors (35 torr) was significantlyhigher than for nonsurvivors (13.7 torr). The me-dian ETCO2 value of 35 torr in this study mirrorsmean ETCO2 levels in survivors in prior studies,which ranged from 31 torr to 33 torr.11,13 In com-parison with previous studies in which nonsurvi-vors had mean ETCO2 levels ranging from 4.4 torrto 20.6 torr, nonsurvivors in our study had a me-dian ETCO2 level of 13.7 torr.11,13 An ETCO2 valueof 16 torr or less successfully discriminated be-tween the survivors and nonsurvivors in our studybecause no patient survived with an ETCO2 lessthan 16 torr. Our logistic regression model furthershowed that for every increase of 1 torr in ETCO2,the odds of surviving increased by 16%.

Both the sonographic demonstration of cardiacactivity and ETCO2 higher than 16 torr were sig-nificantly associated with survival. Of the two cor-relation coefficients, f of 0.35 for cardiac sonogra-phy and point-biserial coefficient of 0.53 forcapnography, elevated ETCO2 levels were morestrongly associated with survival than the pres-ence of sonographically detected cardiac activity.The multivariate logistic regression model combin-ing cardiac sonography and capnography correctlyclassified 92.4% of subjects. However, only capnog-raphy was found to be a significant predictor; car-diac sonography did not enhance the predictive ac-curacy of the model. Furthermore, the results ofthe stepwise logistic regression model not only con-firmed the multivariate logistic regression model,but also demonstrated that capnography alone isan outstanding predictor of survival.

Clinicians did not report any instances in whichcardiac sonography appeared to interfere with re-suscitation, and the responses of the clinician par-ticipants demonstrated that cardiac sonographycan be carried out with relative ease during theperformance of cardiopulmonary resuscitation(Fig. 3). Furthermore, the impression that cardiacultrasound provided useful information about theresuscitation was reported by 96% of the clinicianparticipants.

LIMITATIONS AND FUTURE QUESTIONS

This study used a convenience sample because ofseveral methodological constraints. The ultra-sound scanner was occasionally in use for other in-dications. Sometimes there were no trained emer-gency physician sonographers available during theresuscitation. Additionally, use of cardiac sonog-raphy in resuscitation was not mandatory. It also

should be noted that this study did not attempt toquantify the vigor with which the sonographicallyidentified heart was beating. Future studies mightattempt to quantify the strength of sonographicallyidentified cardiac contractions. In this study, sur-vival was defined as survival to admission; follow-up studies may want to evaluate the long-term out-come of the survivors. Although we cannot excludethe possibility that, for example, the absence ofcardiac activity coupled with low ETCO2 shortenedthe duration of resuscitation efforts, we believethat all the resuscitations in this study were car-ried out according to ACLS guidelines. Finally, ourmodest sample size may account for the fact thatcardiac sonography was not a predictor of survivalin our logistic regression analysis, and furtherstudy with larger cohorts may be warranted.

CONCLUSIONS

Sonographic observation of cardiac activity andETCO2 levels higher than 16 torr were both sig-nificantly associated with survival from ED resus-citation; however, logistic regression analysis dem-onstrated that prediction of survival using ETCO2

higher than 16 torr was not enhanced by cardiacsonography. Clinician participants perceived thatcardiac sonography could be performed with rela-tive ease at the bedside without hampering cardi-opulmonary resuscitation efforts.

References

1. Varriale P, Maldonado JM. Cardiac sonography observa-tions during inhospital cardiopulmonary resuscitation. CritCare Med. 1997; 25:1717–20.2. Ettin D, Cook T. Using ultrasound to determine externalpacer capture. J Emerg Med. 1999; 16:1007–9.3. Bocka JJ, Overton DT, Hauser A. Electromechanical disso-ciation in human beings: a cardiac sonography evaluation. AnnEmerg Med. 1988; 17:450–2.4. Glasser SP, Harrison EE, Amey BD, Straub EJ. Cardiacsonography incidence of pericardial effusion in patients resus-citated by emergency medical technicians. J Am Coll EmergPhysicians. 1979; 8:6–11.5. Corbett SW, O’Callaghan T. Detection of traumatic compli-cations of cardiopulmonary resuscitation by ultrasound. AnnEmerg Med. 1997; 29:317–22.6. Amaya SC, Langsam A. Ultrasound detection of ventricularfibrillation disguised as asystole. Ann Emerg Med. 1999; 33:344–6.7. Asplin BR, White RD. Prognostic value of end-tidal carbondioxide pressures during out-of-hospital cardiac arrest. AnnEmerg Med. 1995; 25:756–61.8. Cummins RO (ed). American Heart Association AdvancedCardiac Life Support Textbook. Dallas: AHA, 1997.9. Falk JL, Rackow EC, Weil MH. End-tidal carbon dioxideconcentration during cardiopulmonary resuscitation. N Engl JMed. 1988; 318:607–11.10. Idris AH, Staples ED, O’Brien DJ, et al. End-tidal carbondioxide during extremely low cardiac output. Ann Emerg Med.1994; 23:568–72.11. Levine RL, Wayne MA, Miller CC. End-tidal carbon diox-ide and outcome of out-of-hospital cardiac arrest. N Engl JMed. 1997; 337:301–6.12. De Piccoli B, Rigo F, Raviele A, et al. Transesophageal car-

ACADEMIC EMERGENCY MEDICINE • June 2001, Volume 8, Number 6 615

diac sonography evaluation of the morphologic and hemody-namic cardiac changes during ventricular fibrillation. J Am SocCard Echo. 1996; 9:71–8.13. Paradis NA, Martin GB, Goetting MG, Rivers EP, FeingoldM, Nowak RM. Aortic pressure during human cardiac arrest:

identification of pseudo-electromechanical dissociation. Chest.1992; 101:123–8.14. Sanders AB, Kern KB, Otto CW, et al. End-tidal carbondioxide monitoring during cardiopulmonary resuscitation: aprognostic indicator for survival. JAMA. 1989; 262:1346–51.

v

REFLECTIONS

For Jonathan (at 5 Years)

We are but older children, dear,Who fret to find our bedtime near.

—Lewis Carroll1

Who made the sky? Playing basketballOn a cloudless day, you ask, between shots.I, who should know better, say I don’t know.You persist.Grandma and grandpa later tell you, ‘‘God made the sky.’’You look at me and shrug.Metaphysics is a children’s primer:Who made the first day?What’s the last number?Who was the first person?Eternity, infinity, and creation pass in your mindcasually, while you push a toy truck, orat bedtime, or on the way to school.My answers are weak rockets I send up—theysputter, then fall to the ground.The flames of philosophy, theology, and sciencecrackle in young minds, energized by wonder.Only grown-ups think they have the answer.But they’ve forgotten the questions.Fundamental questions asked in life’s foundation—The most basic by the most basic of minds.Years etch trivialities into concerns: mutual funds and mortgages,politics and gossip, schedules bordered by deadlines—All push back the early questions.Until we turn the curve to the end, and the heart slows, the brainfalters, the joints freeze. Only then the questions of childrenrush in a flood: The why of who we are, and how we came to be.Some lean on religion, or philosophy, or the latest trend.Some learn new Eastern meditations, adapt special diets,discover herbs and roots with medicinal import.Some find love again, others leave, some come back:All together in a heap at the end, asking the same questionsof children, mindless of destiny, shooting baskets on cloudlessdays, hours before bedtime.

BONNIE SALOMON, MDLake Forest HospitalLake Forest, Illinois

Reference

1. Carroll L. Through the Looking-Glass. London: Anness Publishing,1999.