| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Guest Access | Sign In via User Name/Password
|
|
|
|||||||
Guest Access | Sign In via User Name/Password |
|||||||||
Pulmonary and Critical Care Division New England Medical Center Tufts University School of Medicine Boston, MA
To the Editor:
The recent study of Drs. Melendez and Carlon1
demonstrated that an increased cardiopulmonary risk index (CPRI) score
was predictive of cardiopulmonary complications associated with
pneumonectomy (positive predictive value [PPV] 1.0, negative
predictive value [NPV] 0.64). In contrast to our Veterans
Administration studies,2
,3
these investigators found an
increased CPRI not predictive for other thoracic procedures. Important
differences in patient populations, technique of scoring, methods for
determining complications and preoperative and postoperative treatment
(of CPRI risk factors) may explain part of the discrepancy. Although
interpretation should be cautious given their preliminary nature, there
are now four additional prospective studies (including my
own4
) totaling 350 patients, that have examined the
issue.4
,5
,6
,7
Two blinded trials in non-Veterans
Administration populations4
,5
found a CPRI 
4
predictive but with less accuracy than our initial experience with PPV
(0.62 to 0.64) and NPV (0.77 to 0.83). This may result from reduced
predictive accuracy in female patients (PPV 0.42, NPV
0.79)4
or among those undergoing non-lung-cancer surgery.
Importantly, approximately 50% of the patients reported by Drs.
Melendez and Carlon were women.
It is surprising that the well-established Goldman index was not
predictive of cardiac complications in the Melendez and Carlon study.
To be consistent with Goldman, we added three points to each cardiac
risk index score (CRI) to account for thoracic surgery.2
Therefore, the presence of any other Goldman risk factor results in a
CRI 2. As anticipated in this cohort at high risk for
cardiovascular disease based on age and tobacco use, we found a CRI
2 in 45% of patients.2
,3
,4
In contrast, Drs. Melendez
and Carlon report that only approximately 15% of their patients had a
CRI 2. Similarly, we noted a left ventricular ejection fraction
(LVEF) 
40% in 10% of patients, with > 50% ultimately suffering
a perioperative complication.2
,3
Strikingly, in the July
1998 CHEST paper,1
of 180 patients, none had an
LVEF 40%. This suggests that these authors were studying either a
remarkably fit population or that CRI scores were spuriously low. The
latter can result in falsely decreased PPVs and NPVs for the CPRI.
I agree with these authors that one should avoid observation bias in
assessing risk factors. That is precisely why we used explicitly stated
criteria and time frames for pulmonary risk index factors (obesity body
mass index 27, tobacco within 8 weeks of surgery, productive
cough/wheeze within 5 days of surgery). In contrast, although Drs.
Melendez and Carlon offer exact criteria for increased
PaCO2 and decreased FEV1/FVC, explicit
definitions for the other parameters are inexplicably not provided.
Similarly, different definitions for pneumonia and prolonged mechanical
ventilatory support are used (allowing for inclusion of early
reintubations possibly related to inadequate recovery from anesthesia
rather than true respiratory failure). Most importantly, in contrast to
the stated method in their study, other recently completed trials have
used a blinded format to prevent observer bias, with the investigator
scoring complications blinded to the assigned CPRI
score.4
,5
Lastly, all studies lack sufficient control for the efficacy of specific preoperative and postoperative care directed at CPRI risk factors. The CPRI was originally conceived to help quantify underlying cardiopulmonary disease to better understand the correlation between exercise oxygen consumption and postoperative complications.2 In isolation, though predictive of complications, it was not sufficiently accurate in predicting mortality to preclude surgery. To the contrary, we recommended the CPRI be used to explicitly identify risk factors amenable to correction, leading to more intensive preoperative preparation and postoperative care for those at increased risk.2 When such therapies are effective, we anticipate that predictive value will decrease. Perhaps the recently reported reductions in PPV reflect attainment of that goal.
Correspondence to: Scott K. Epstein, MD, Pulmonary and Critical Care Division, Box 369, New England Medical Center, 750 Washington St., Boston, MA 02111
References
Vittoria Arslan-Carlon, MD Anesthesiology and CCM Memorial Sloan-Kettering Cancer Center New York, NY
To the Editor:
Dr. Epstein is correct when he mentions the need to add three points to patients undergoing thoracic surgery in the calculation of the cardiac risk index (CRI) score. The CRI range mentioned in his paper should have been 3 to 49 not 47 (Fig 1 and its legend).1-1 Furthermore, Dr. Goldman included the three points in his index in order to weight the effect of cavitary surgery on the outcome of all operations,1-2 and clearly that was not the case in our population. All patients underwent the same cavity operation. As a result of this discrepancy and the belief that the additional three points, applicable to all patients, would have only shifted the curve to the right without affecting the overall distribution, we chose not to use them. In light of the criticism, we have added the three points to our population and reviewed the statistics. The results were unchanged to the previous, and we confirmed our statement that cardiopulmonary risk index (CPRI) is not a good index to predict complications after thoracic surgery. We were able to compare receiver operative characteristic curve constructed with and without "the 3 points" for both the entire patient population (p = 0.156) and those undergoing pneumonectomy (p = 03.11), and the results are not statistically significant. Regarding the left ventricular ejection fraction, it is possible that our population is "remarkably fit" or the underling disease limits intervention to those whose expectancy of life is really improved by the thoracic surgery.
We failed to explicitly describe some of the preoperative and
postoperative criteria. This was an oversight. All criteria were either
identical or negligibly dissimilar to Epstein et al.1-1
The
preoperative criteria were obesity (body mass index 27), recent
smoking (tobacco use within 8 weeks of surgery), productive
cough, and wheezing or rhonchi (on admission). The
postoperative complications criteria were pneumonia (temperature
38.5°C, antibiotic treatment as a corroboration of the
diagnosis), and postoperative ventilation (lasting > 3 h not 48
h). At the beginning of the study, we felt the change in postoperative
ventilation criteria to be significant at our institution; we practice
immediate postoperative extubation in the operating room. However, none
of our patients fulfilled either criteria for postoperative
ventilation. All recorded postoperative respiratory failures were
reintubations performed after discharge from the recovery room and
could not be considered anesthesia-related.
To demonstrate the predictive power of any variable, it is imperative that the methods include preoperative estimation of risk and subsequent comparison to actual outcome. It is not sufficient to demonstrate a prospective relationship between a variable and a certain outcome. The accuracy of estimation of the risk experienced by a particular patient during a particular operation should be the goal of this exercise, and not the demonstration of a relationship between a variable and an outcome. We have failed to see this kind of analysis in studies that find CPRI predictive of outcome in thoracic surgery. In addition, the population size of each study is not only smaller than ours, but insufficient to establish the certainty of predictability.1-3 ,1-4 ,1-5 We do agree with Dr. Epstein on the importance of blinding investigators to the CPRI scoring. Although we did not perform a blinded study, the individual collecting the preoperative data was not the same as the one tabulating the outcome, in essence, blinding the investigators. In addition, these kinds of studies are difficult, as Dr. Epstein points out, because inevitably, physicians will direct extra care to those individuals who may be perceived as having a higher cardiopulmonary risk thereby reducing positive predictive value.
Our aim was to identify a simple usable criteria to predict outcome after thoracic surgery. We failed to recognize the importance of the relationship between exercise oxygen consumption and CPRI. CPRI may be a good tool to identify those patients who need more preoperative preparation and/or postoperative care, but it is prone to institutional bias related to investigator observation. An index using only numeric measurements that can be reproduced at different institutions, and is not prone to observer bias, would better fit this unique purpose.
Correspondence to: Jose A. Melendez, MD, Department of Anesthesiology and CCM, Memorial Sloan-Kettering Cancer Center, 1275 York Ave, New York NY 10021
References
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
