| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
J Thorac Cardiovasc Surg 2003;126:761-768
© 2003 The American Association for Thoracic Surgery
General thoracic surgery |
a Centre de recherche, Hôpital Laval, Institut universitaire de cardiologie et de pneumologie de l'Université Laval, Ste-Foy, Québec, Canada
Received for publication December 4, 2001; revisions received March 28, 2002; revisions received October 7, 2002; accepted for publication October 18, 2002.
* Address for reprints: Yves Lacasse, Hôpital Laval, Centre de Pneumologie, 2725 Chemin Ste-Foy, Ste-Foy, Québec G1V 4G5, Canada
yves.lacasse{at}med.ulaval.ca
 |
Abstract |
|---|
 Top Abstract MethodsResultsDiscussionAppendixReferences |
|---|
METHODS: We conducted a mailed survey among Canadian physicians including pulmonologists, thoracic surgeons, radiologists, and internists. Five hypothetical clinical scenarios designed to illustrate a wide spectrum of clinical situations (including nodules associated to very low, indeterminate, and high probabilities of malignancy in patients medically fit or unfit to undergo surgical resection) were submitted to each physician who had to choose among options of diagnostic procedures and to estimate the pretest probability of malignancy.
RESULTS: The survey response rate was 47%. Overall, the physicians tended to overestimate the probability of malignancy. Twenty-eight percent of the respondents would have ordered a transthoracic needle biopsy in a medically fit patient presenting with a very low probability (< 2%) of malignancy, whereas 53% would have done so in a medically fit patient presenting with a very high probability (75%) of malignancy. In a regression analysis, the significant predictors of the use of transthoracic needle biopsy were (1) the pretest probability of cancer, (2) the patient's operability, (3) specialty (pulmonologists and thoracic surgeons being less likely to order a transthoracic needle biopsy than radiologists), and (4) the respondent's gender.
CONCLUSIONS: There are widespread variations in clinical opinions among Canadian physicians regarding the use of transthoracic needle biopsy in the investigation of solitary nodules. Hypothetical transthoracic needle biopsy was often preferred when the result was unlikely to affect patient management.
Transthoracic needle biopsy (TNB) is a diagnostic technique that provides histologic or cytologic material from lung lesions.5 Guidelines regarding the indications of TNB remain nondirective, and the American Thoracic Society recommends that "experience and sound clinical judgment...determine its use in individual cases."6 Reappraisal of the use of TNB in the diagnosis of localized pulmonary lesions requires a better understanding of its accuracy and clarification regarding its indications. A recent meta-analysis7 of the diagnostic accuracy of TNB indicated that it provides definitive results only when findings of "malignant" or of a specific diagnosis of a benign condition provide definitive results. Findings of "suspicious" markedly increases probability of malignancy, and "benign" markedly decreases it, but may not be considered definitive. Nevertheless, biopsies are often performed routinely, regardless of the probability of malignancy, when it is unclear that the biopsy will modify the clinical management of the patient.8,9
The objective of this study was to identify what determines physicians' choices of diagnostic procedures in patients presenting with a solitary pulmonary nodule. More specifically, we sought to (1) describe Canadian physicians' current opinions relative to the choice of diagnostic procedures in patients with a solitary pulmonary nodule; (2) identify the attributes (including physicians' training, clinical practice, and appointment) that may influence their decision; and (3) examine whether the physicians' estimate of the probability of malignancy modulates the choice of diagnostic procedure in patients with a solitary pulmonary nodule. To do so, we conducted a national survey among Canadian physicians most likely to be involved in the management of patients with a solitary pulmonary nodule. We hypothesized that the estimated pretest probability of malignancy would determine the decision to perform a TNB and that physicians' attributes (especially their specialty) would predict their use of TNB.
|
Methods |
|---|
|
TopAbstractMethodsResultsDiscussionAppendixReferences |
|---|
Survey
The first section of the survey was designed to collect general information on respondents including gender, year of graduation, specialty, clinical experience with the investigation of solitary pulmonary nodules, and work environment (university affiliated or not). Each respondent was then presented with 5 hypothetical clinical scenarios that were designed to illustrate a wide spectrum of situations currently encountered in clinical practice (see appendix), including benign, indeterminate, or malignant nodules in patients medically fit or unfit to undergo surgical resection. Each case was designed to correspond to a pretest probability of malignancy that was determined from a set of clinical and radiologic characteristics and from a clinical prediction model recently published by Swensen and coworkers.1 For each scenario, physicians had to choose among options of diagnostic procedures and to estimate the probability of malignancy from a 100-mm (0%-100%) analog scale.
The survey was conducted between March and May 2000 according to the total design method described by Dillman.10 Each physician in the sample was mailed a package containing a cover letter signed by the investigators, a booklet of questions in his/her language (French or English), and a stamped, addressed reply envelope. Each survey was numbered and coded for identification and to protect confidentiality. One week later, this first mailing was followed by a reminder postcard. Six weeks later, those who had not responded to the initial survey were sent a second package that was identical to the first one except for the cover letter, which mentioned that their survey had not yet been received.
Statistics
Sample size
In calculating the sample size needed for this study, we hypothesized that patients in whom the biopsy was truly indicated (ie, in our opinion, the patient with intermediate pretest probability of malignancy who was medically fit for surgery) would be at least twice as likely to undergo a biopsy as those in whom the indication of biopsy was not clear. We anticipated that 25% of those patients in whom the biopsy was not clearly indicated would actually undergo a biopsy. Given the large number of planned statistical comparisons, the level of statistical significance in the regression analyses was set at the .01 level. Study power was set at 85%. From these specifications, we calculated that 500 responses were needed. This sample size was augmented to take into account the anticipated response rate that was likely to vary according to the subspecialties. The highest response rate was expected from pulmonologists and the lowest from internists. Also, we anticipated a low response rate from the thoracic surgeons since most of them are classified in the Canadian Medical Directory as cardiothoracic surgeons. Given the small number of cardiothoracic surgeons, 100% of them were surveyed to obtain a reasonable number of responses. Accordingly, a total of 1470 physicians were surveyed: 458 radiologists, 455 internists, 320 pulmonologists, and 237 surgeons.
Response rate
The response rate was calculated as the ratio of analyzable responses over the number of eligible physicians that were reached.11
 |
A "noneligible" designation was applied to physicians who returned their survey indicating that they are never called to evaluate patients with a solitary pulmonary nodule in their practice. A "nonreachable" designation was applied to physicians whose first package was returned to us. Newfoundland, Nova Scotia, Prince Edward Island, and New Brunswick were analyzed as the Atlantic provinces and Alberta, Saskatchewan, and Manitoba as the Prairies. The other provinces (Ontario, Québec, and British Columbia) were analyzed separately.
Analysis
Descriptive statistics (proportions, means, and standard deviations when appropriate) were used to characterize the study population. A frequency distribution of the different diagnostic options was computed for each scenario and by specialty. Between-specialty comparisons were conducted using 
2 tests. The estimated pretest probabilities of malignancy were described using the mean (± standard deviation) and compared to the pretest probability derived from Swensen's clinical prediction model using t tests. Between-specialty comparisons were conducted using analyses of variance.
We conducted a logistic regression analysis to test the hypothesis that some physicians' attributes (including their training, practice, and appointment), their estimate of pretest probability of malignancy, as well as the patient's medical fitness to undergo surgical resection may influence their decision to have a biopsy performed on a patient. Since each physician we surveyed had to examine 5 different case scenarios, the responses provided by an individual physician could not be regarded as independent observations. To correct for this situation, we adopted the method introduced by Liang and Zeger12 by constructing a generalized linear model with unstructured covariance. In this model, the physicians' estimates of the pretest probability of malignancy were categorized according to thresholds defining very low (<5%), low (5%-30%), intermediate (31%-50%), high (51%-70%), and very high (>70%) probabilities of cancer.13
|
Results |
|---|
|
TopAbstractMethodsResultsDiscussionAppendixReferences |
|---|
All the regions of the country were represented, with response rates ranging from 44% in the Prairies to 50% in British Columbia. The response rate from the 4 disciplines was heterogeneous: respirology, 62%; radiologists, 44%; thoracic surgeons, 36%; and internists, 20%. The respondents were unevenly divided by specialty: respirology, 187; radiologists, 167; thoracic surgeons, 81; and internists, 81. Sixteen percent of the respondents were women. On average, the respondents had been in clinical practice for 21 years (SD: 10); 58% worked in a university-affiliated teaching hospital. Most of them had extensive experience in managing patients with a solitary nodule, as 304 (59%) claimed that they investigate more than 12 patients a year in such circumstances. Only 141 (27%) of the respondents (mostly radiologists, n = 120) performed TNB themselves.
Estimates of pretest probability of malignancy
The pretest probability of malignancy determined from Swensen's clinical prediction model are presented in Table 1.
These probabilities are compared with those elicited by the respondents. Overall, the physicians tended to overestimate the probability of malignancy. There were small (although statistically significant) differences in the estimates across specialties.
|
|
|
|
|
Discussion |
|---|
|
TopAbstractMethodsResultsDiscussionAppendixReferences |
|---|
In our survey-based study, we found that physicians tended to overestimate the pretest probability of cancer. This situation may partly explain the perceived indication of TNB in case 3 (ie, the patient with the nodule associated to a very low probability of malignancy), as this nodule was often considered by the respondents as of intermediate probability. Similarly, this situation may partly explain the referral for radiation therapy even though the diagnosis of malignancy was not proven in case 5 (ie, the patient with the nodule associated to an intermediate probability of malignancy), as this nodule was often considered by the respondents as of very high probability of malignancy. On the other hand, TNB was overused in case 1 (ie, the medically fit patient with a nodule associated to a very high probability of malignancy), when the result was unlikely to affect the management of the patient. This interpretation of the results assumes that the pretest probability estimated from Swensen's model is correct. An alternative interpretation would be that clinicians are more accurate in their current pretest probability estimates than Swensen's model. Our computation of the pretest probabilities of malignancy from other prediction models3,4 demonstrated very good agreement in the probabilities, although the models were developed in different settings and in different points in time (data not shown).
From the American Thoracic Society guidelines, the patient most likely to benefit from TNB was case 2, represented by the medically fit patient with a moderate probability (38%) of malignancy; 71% of the respondents would have referred this patient to TNB. Of the other 29%, two thirds would have directly referred the patient to surgical resection.
Whether a patient is medically fit for surgery was a predictor of the use of TNB, but not to the extent we expected. In case 4, we posed the case of a 68-year-old woman with a nodule associated to a high probability of malignancy (83%). She was medically unfit for surgery because of severe COPD and left ventricular dysfunction. Also, she had already informed her treating physician that she would refuse chest radiotherapy as a treatment should this lesion be proved to be malignant. This patient would have been referred for TNB by 25% of the respondents. This situation is considered by the American Thoracic Society a contraindication to TNB.6 More controversy surrounded case 5, represented by the medically unfit patient with an intermediate probability nodule. In this case and according to the American Thoracic Society guidelines, TNB could be useful to diagnose cancer in a patient who, because of personal or clinical reasons, is not a candidate for resection of the lesion unless it is definitely malignant.6 Clearly, this indication is not widely accepted among physicians, as this individual would have been referred for TNB by 56% of the respondents. An alternative explanation for this high rate of biopsy in patients who are apparently "unfit for surgery" is that a general consensus may not exist as to how to define "fitness for surgery," especially in the era of limited resection by thoracoscopy.
Do our findings really reflect the use of TNB in Canada? The response rate we obtained (47%) is typical of the reply of physicians to mailed surveys.11 In their review of 321 distinct mail surveys examining attitudes and behaviors in health care settings published in American journals in 1991, Asch and coworkers11 reported that physicians had the lowest mean response rate among all groups examined (54% ± 17%). Nonresponse to any survey may be a problem because of the possibility that respondents differ in a meaningful way from nonrespondents, thus biasing the results. However, surveys with low response rates may provide a representative sample of the study population. We attempted to investigate the potential of nonrespondent bias by making a priori predictions on the response rates across specialties and by examining the representation in the survey of physicians from all the regions on the country. Roughly, the response rates met our a priori predictions. Also, all parts of Canada were proportionally represented in the survey. Whether the results of our survey reflect the actual use of TNB by the sample population remains unanswered since the respondents were presented hypothetical clinical scenarios.
We acknowledge that the inclusion of flexible bronchoscopy as a routine diagnostic procedure in all of our 5 scenarios may be controversial.14,15 All 5 bronchoscopies were nondiagnostic. We made this decision on purpose. This survey was not designed to examine whether clinicians would select flexible bronchoscopy or TNB in the management of patients with solitary pulmonary nodules after the initial clinical and radiologic evaluation. By incorporating a negative bronchoscopy in each scenario, we wanted to isolate the option of TNB as the next step in management. In doing so, the bronchoscopy slightly modified the probability of malignancy, even if it was nondiagnostic. The authors of most studies of the diagnostic accuracy of bronchoscopy reported their results in terms of "diagnostic yield" in malignant nodules.16 The most recent study by Baaklini17 included patients with both malignant and benign solitary nodules. From their data, we computed that the likelihood ratio (which indicates by how much a given test result raises or lowers the pretest probability of the target disorder18) associated to a nondiagnostic bronchoscopy in patients with solitary nodules smaller than 3 cm is 0.9. This likelihood ratio merely reduces the probability of malignancy for each scenario to 73%, 36%, < 2%, 82%, and 34%, respectively.
Several reports have questioned the current use of TNB.8,9 In a retrospective study of 113 patients who underwent 117 biopsies between 1991 and 1996, the diagnostic accuracy of the procedure was high (sensitivity: 85%; specificity: 100%).9 Of the biopsies, 54.7% were associated with a iatrogenic pneumothorax, about half of them necessitating hospitalization for either observation or chest tube drainage. Nevertheless, for most patients, the result of the procedure did little to modify the course of surgical management. The authors recommended that TNB should be limited to patients who are medically unfit and could not tolerate surgery, to patients who refuse surgery, and to patients with advanced disease. In a different setting, the same conclusions were reached by independent investigators.8 In this study, Larscheid and coworkers retrospectively reviewed the charts of 130 consecutive patients who had undergone TNB. The sensitivity of TNB for the detection of malignancy was 74% and its specificity was 100%. The authors concluded that the information gained from TNB rarely resulted in a change in their clinical management and that TNB is most useful for diagnosing pulmonary nodules that are likely to be malignant in patients who are not candidates for other surgical procedures.
The recent introduction of positron emission tomography (PET) scanning further complicates the integration of TNB in the diagnostic algorithm of localized pulmonary lesions. Several studies have reported high sensitivity and specificity of this new technology in the detection of malignant solitary pulmonary nodules.19 A recent head-to-head comparison between TNB and PET scanning20 suggested that PET imaging of the lung is as accurate as TNB in the evaluation of lung lesions suspected to be malignant. For methodologic reasons, this study did not meet, however, the inclusion criteria of our meta-analysis.7 PET scanning was rarely elicited by our respondents as a diagnostic option in the management of the patients we presented. This situation may reflect the fact that, as of January 1, 2001, only 7 PET centers were in function in Canada.
Finally, the role of TNB in the investigation of solitary pulmonary nodules may increase if induction chemotherapy in early-stage, non–small cell lung cancer proves effective in improving survival.21 In the meantime, the indications of TNB must be based on the estimate of pretest probability of malignancy and the likelihood that the test result will affect the management of the patient. The recent interest in screening of lung cancer by low-dose computed tomography launched by the Early Lung Cancer Action Project22 will likely result in the finding of a large number of pulmonary nodules for which TNB is seldom truly indicated.
|
Appendix |
|---|
|
TopAbstractMethodsResultsDiscussionAppendixReferences |
|---|
patient 2
A 57-year-old farmer, ex-smoker of 20 pack-years, is referred to you for the investigation of a pulmonary nodule discovered on a routine chest radiograph. He has no known medical problems and he is asymptomatic. The chest radiograph shows a 2-cm nodule in the apical segment of the right upper lobe. There is no prior chest radiograph available. The chest CT scan confirms the presence of this well-defined, noncalcified lesion. The rest of the pulmonary parenchyma is normal and there is no mediastinal adenopathy. The bronchoscopy is normal and the cytologic examination of bronchial secretions does not reveal any malignant cells. His pulmonary function tests are normal.
patient 3
A 35-year-old man is referred to you for the evaluation of a pulmonary nodule found on a routine chest radiograph. His medical history is unremarkable and he has never smoked. There is no history of tuberculous contact and he is asymptomatic. His chest radiograph shows a 1-cm nodule in the right lower lobe. The chest CT scan confirms the presence of a solitary, smooth, noncalcified, 1.2-cm nodule in the right lower lobe. The rest of the pulmonary parenchyma is normal. No prior chest radiograph is available. A bronchoscopy has already been done by the referring physician and was normal.
patient 4
You are referred a 68-year-old woman for the assessment of a solitary pulmonary nodule. She has documented chronic obstructive pulmonary disease and severe coronary artery disease. She stopped smoking 8 years ago. A chest radiograph done in the follow-up of her chronic lung disease shows a 3-cm lesion in the left lower lobe. This lesion was absent on a radiograph done 2 years earlier. The chest CT scan confirms the presence of this unique 3.1-cm peripheral lesion with indistinct margins. There is no mediastinal lymphadenopathy. The bronchoscopy is normal and the cytologic examination of bronchial secretions does not reveal any malignant cells. Her best FEV1 [forced expiratory volume in 1 secone] is 30% of predicted and her left ventricular ejection fraction is 25%. The patient has already informed you that she would refuse chest radiotherapy as a treatment should this lesion be proved to be malignant.
>Patient 5
You see a 72-year-old woman for an abnormal chest radiograph. She has diabetes complicated by nephropathy and peripheral vascular disease. She is also known to have severe ischemic heart disease. She has been smoking since her youth. Her FEV1 is 800 mL (40% of predicted). Her chest radiograph shows a dense 2-cm nodular opacity in the right lower lobe and some cicatricial markings in the upper lobes. The chest CT scan confirms the presence of this lesion and signs of a previous granulomatosis. There is no mediastinal adenopathy. The bronchoscopy is normal and the cytologic examination of bronchial secretions does not show any malignant cells. There is no prior chest radiograph available for comparison.
Following each scenario, the physicians were asked the same 3 questions:
|
Acknowledgments |
|---|
|
Footnotes |
|---|
|
References |
|---|
|
TopAbstractMethodsResultsDiscussionAppendixReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  J. Strausz Transthoracic needle biopsy Interventional Pulmonology, June 18, 2010; 109 - 118. [Abstract] [Fulltext] [PDF]
|
|
|
|
 |
|
 T Targowski, K Jahnz-Rozyk, T Szkoda, S From, N Qandil, and T Plusa Telomerase activity in transthoracic fine needle biopsy aspirates as a marker of peripheral lung cancer Thorax, April 1, 2008; 63(4): 342 - 344. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| ANN THORAC SURG | ASIAN CARDIOVASC THORAC ANN | EUR J CARDIOTHORAC SURG |
| J THORAC CARDIOVASC SURG | ICVTS | ALL CTSNet JOURNALS |
