Delay by Internists in Obtaining Diagnostic Biopsies in Patients with Suspected Cancer Sherif S. Farag, MB; Michael D. Green, MB; George Morstyn, MB, PhD; William P. Sheridan, MB; and Richard M. Fox, MB, PhD
• Objective: To investigate the degree and type of delays in performing diagnostic biopsies in medical patients with suspected malignancy. • Design: Retrospective survey of clinical histories of patients referred between January 1985 and March 1989. • Setting: Inner city teaching hospital internal medicine (nononcologic) services. • Patients: Patients with gastrointestinal and lung cancers, adenocarcinoma of unknown primary site, and lymphomas were referred as inpatients by internists. Two hundred fifty-five patients were eligible, and 177 were evaluable. • Main Outcome Measures: The number, type, and results of tests done before and after biopsy were analyzed. • Results: In 67% of patients the biopsied lesion was detected by the second day of evaluation; however, there was an 8- to 10-day delay before a biopsy was done. This delay was consistent across the four malignancy groups studied. Although logistic and other unavoidable delays occurred in 40% of the cases, in 60% delays could only be attributed to continued, frequently low yield, noninvasive tests. An average of 3.3 tests were made per patient, with only 24% leading to a definitive biopsy. • Conclusion: Because of the performance of many other tests, a substantial delay exists in proceeding to biopsy during the diagnosis of cancer by internists. Annals of Internal Medicine. 1992;116:473-478. From The Royal Melbourne Hospital, Victoria, Australia. For current author addresses, see end of text.
tigations and a delay in diagnosis. Although such a delay will rarely affect treatment outcome, it is likely to increase the length of hospital stay, cause more distress to patients, and increase the costs to health care systems. The importance of securing an early histologic diagnosis in the investigative process of patients presenting with occult primary malignancies has been previously emphasized (1-4). Ultmann and Phillips (1), writing about occult primary malignancy, recommend reversing the traditional approach: . . . The optimal sequence of steps when evaluating the patients with [occult primary malignancies] concentrates on the re-evaluation of the biopsy material. Note that this differs from the traditional order of enquiry (history, physical examination, laboratory studies) used to delineate other clinical problems. This biopsy should be reviewed first . . . any clues that the biopsy provides will facilitate a directed history or physical examination with emphasis on specific areas of interest. This approach will also target the investigations and is applicable to other forms of malignancy. Further, Ultmann and Phillips (5) are even more direct: "Regardless of the difficulty of the procedures involved histological proof of malignancy must be obtained before the risks involved in further staging." To document our impression, we reviewed the clinical histories of 255 patients presenting with four different groups of malignancies referred to a medical oncology service by internists. We attempted to quantitate the extent and cause of delay in obtaining a diagnostic biopsy. Methods
1 he appropriate management of patients presenting with malignancy depends almost entirely on obtaining an histologic diagnosis because this usually dictates the most appropriate form of treatment. Obtaining a biopsy is, therefore, a prime task in the diagnostic process. It was our impression, and that of other medical oncologists, that internists do a large number of noninvasive investigations in an attempt to either diagnose or stage the patient before resorting to biopsy. This appeared to contrast with surgical practice where biopsies are done more quickly. Such a practice appears to originate in the traditional philosophy in internal medicine to proceed from the least to more invasive investigations. In the diagnosis of cancer, where tissue diagnosis is vital, this approach often results in many unnecessary inves-
Four groups of malignancies, including gastrointestinal carcinomas, non-small cell lung carcinoma, adenocarcinoma of unknown primary site and lymphomas (including Hodgkin disease), were considered in this study. Between January 1985 and March 1989, 255 patients with those malignancies were referred by internists from a single general teaching hospital. The patients were worked up as inpatients on general and specialist (nononcologic) internal medicine services. The attending physicians and housestaff, as a team, decided on the details of the investigative process and ultimate referral. An histologic or cytologic diagnosis of malignancy was a requirement for inclusion. The clinical histories were reviewed retrospectively to obtain the number, type, and results of investigations done before and after biopsy, as well as the number and type of biopsies. The times from initial evaluation to suspicion of cancer, to detection of the subsequently biopsied lesion, and to the actual biopsy were also recorded. Specific delays in doing a biopsy were noted. The data was coded and analysed on a VAX computer (Digital Equipment Corporation, Maynard, Massachusetts) using the SPSSX package (SPSS Inc., Chicago, Illinois). ©1992 American College of Physicians
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19746/ by a University of California San Diego User on 06/07/2017
473
Figure 1. Delay in diagnostic biopsy. Time taken to suspect cancer as a diagnosis, detect an ultimately biopsied lesion, and perform the actual definitive diagnostic biopsy. Cumulative frequency plot with time on a logarithmic scale.
Clinical Presentations and Investigations Presenting symptoms and signs were recorded and those unrelated to the underlying malignancy were excluded from analysis. Investigations done before and after biopsy were recorded separately. Investigations done before biopsy were those performed in an attempt to reach or clarify the diagnosis of malignancy or its apparent stage. Investigations done for other reasons, such as management of concurrent illnesses or intercurrent complications, were not included in the analysis. Investigations that were repeated only to verify the initial investigation were included in the analysis. The yield of the investigations done before biopsy was examined according to whether the results led to definitive biopsy, whether they were abnormal but relevant, or whether they were "unhelpful" (including normal results, technical failures, or results irrelevant to the problem). Because the investigations after the biopsy were essentially staging investigations, their results were not analyzed. Biopsies The methods by which the definitively biopsied lesion was detected were analyzed. Because some patients had more than one biopsy, the mode of detection of the definitively biopsied lesion that gave the final diagnosis was the one recorded. Analysis of the methods of biopsy and their yield was also done to determine if a requirement for multiple biopsies and low-yield methods of biopsy contributed to delay in diagnosis. Specific delays other than those due to a prolonged investigative pathway were identified. These fell into three categories: "Patient unfit for biopsy procedure," "logistic problems," and other. Logistic problems included delays in obtaining investigation time or operating room time. A logistic problem was considered to be present when 3 or more days elapsed from the time of the last investigation to the biopsy and where it was clear in the clinical notes that there was an intention to do a biopsy. Results Of the 255 eligible patients, 177 (98 men and 79 women) were evaluable and included in the analysis. The remaining patients were excluded primarily because the diagnosis was made outside the hospital or because the clinical notes were inadequate to allow assessment. The mean age of the study group was 61.4 years (range, 474
15 to 86 years), with patients with lymphoma being significantly younger (mean age, 55 years) than the other groups of patients. One hundred twenty-five patients were referred by general internists and 52, by specialist internists; 168 of these patients were investigated wholly as inpatients. Clinical Presentations Almost all patients were suspected of having malignant disease at the time of the first evaluation (Figure 1). Eighty percent presented with physical signs that led to discovery of the biopsiable lesion. (These signs were either an abnormality that could be directly biopsied or an abnormality that initiated an investigation detecting a lesion amenable to biopsy, for example, signs of pulmonary consolidation leading to performance of a chest radiograph and detection of a biopsiable lung mass.) Eighty-five to ninety percent of patients with lung carcinoma, adenocarcinoma of unknown primary site, or lymphoma had physical signs on presentation that led to detection of a biopsiable lesion. In contrast, 17 of 36 patients with gastrointestinal tract cancer had no such signs. Fifty-nine percent of patients had more than one sign. Table 1 shows the distribution of lesions detected at the initial physical examination that were amenable to direct histologic or cytologic examination without any intermediate investigations. Detection of the Definitively Biopsied Lesion The definitively biopsied lesion was usually first detected either clinically or radiologically (Table 2). Overall clinical detection of the definitively biopsied lesion occurred in 67 (38%) patients and was more common in patients with lymphomas (54%) and adenocarcinoma of unknown primary site (50%) than in patients with lung (30%) or gastrointestinal tract (11%) cancer. Conversely, radiologic detection was more common in patients with gastrointestinal tract (64%) and lung (59%)
15 March 1992 • Annals of Internal Medicine • Volume 116 • Number 6
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19746/ by a University of California San Diego User on 06/07/2017
malignancies. The least common initial mode of detection was endoscopy (10%). Altogether, 52% of definitively biopsied lesions were detected at the initial evaluation and 67% were detected by the second day of evaluation (see Figure 1). Subgroup analysis showed that a greater delay occurred in detecting definitively biopsied lesions in the gastrointestinal tract malignancies compared with the other groups. Twenty-five percent of gastrointestinal lesions were detected at the first evaluation, 47% by day 2, and 67% by day 5; this delay was significantly greater than in the other patient groups (P = 0.005). The delay in the detection of gastrointestinal lesions compared with the other tumor groups occurred despite the fact that no difference between the groups was found with respect to when cancer was first suspected. Presumably, this greater delay reflected the scarcity of physical signs in the gastrointestinal tract group. Time to Definitive Biopsy The median time to definitive biopsy for the whole population was 8 days (see Figure 1). There were no significant differences between median times for patients with lung cancer, adenocarcinoma of unknown primary site, and lymphoma (9.5 days, 8 days, and 9 days, respectively). However, the median time to biopsy of 13 days was significantly longer for the patients with gastrointestinal tract disease (P = 0.03). Delays in Obtaining a Biopsy When cumulative frequency plots of times to suspicion of cancer, detection of the definitively biopsied lesion, and definitive biopsy for the whole study group are compared, it is apparent that a substantial delay existed between when the lesion was detected and when it was biopsied (see Figure 1). Although more than half of the lesions were detected at first evaluation, there was about an 8-day delay to biopsy. In 106 (60%) patients, delay appeared to be caused by a prolonged investigative pathway after the definitively biopsied lesion was detected. Delays other than a "prolonged investigative pathway" were present in 71 (40%) patients. The latter included delays due to logistic factors (38 patients), lack of fitness of patients to undergo the biopsy procedure (6 patients), equivocal biopsy results requiring a second biopsy (13 patients), biopsiable lesions missed at initial evaluation (10 patients), reluc-
tance of patients to have a biopsy (3 patients), and temporary loss of follow-up (1 patient). The Prebiopsy "Prolonged Investigative Pathway" A total of 1079 investigations were done, resulting in a mean of 6.1 investigations per patient. Patients with adenocarcinoma of unknown primary site had an average of 7.0 investigations each. This number was significantly higher than the numbers in the other groups (gastrointestinal tract, 5.9; lung, 5.5; lymphoma, 5.8; P = 0.001). Complete blood counts were done on 176 patients, routine serum biochemical screening (including liver function tests) was done on 173 patients, and chest radiographs were done on 148 patients. Because these three studies were done on almost all patients and could be reasonably considered as minimum investigations for a patient with possible malignancy, they were excluded from further analysis. After correcting for the exclusion of the "minimum" investigations, the mean number of investigations was 3.3 investigations (Table 3); patients with adenocarcinoma of unknown primary site still had a substantially higher number. The number of radiologic and endoscopic procedures in this group (251 and 95, respectively) was striking. Only 156 of the total of 1079 (15%) investigations led to a definitive biopsy. If investigations that yielded abnormal and relevant results but did not directly lead to biopsy are included, then 530 (49%) investigations may be deemed "useful." When the minimum investigations are excluded from analysis, 141 (24%) of the investigations led to definitive biopsy and a further 191 (33%) investigations were useful. There still remained a large percentage of investigations (43%) that were not useful and contributed little to defining the problem. The yield of individual investigations is shown in Table 4. Investigations with the highest yield in detecting biopsied lesions were radiologic (ultrasound, computed tomography, magnetic resonance scans, and contrast studies of gastrointestinal tract and spine) and endoscopic (other than lower gastrointestinal tract endoscopy), for which the overall yields were 35% and 40%, respectively. Biopsies A total of 252 biopsies and 124 cytologic examinations of body fluids were done, with an overall average
Table 1. Distribution of Abnormalities, Amenable to Biopsy, Found at the Initial Physical Examination among Patients with Cancer Physical Abnormality* Gastrointestinal Tract (n = 36) Palpable mass Effusion Superficial lymphadenopathy Total with at least one sign
6 3 1 10
Number of Patients in Each Category Adenocarcinoma Lung of Unknown Primary Site (n - 44) (n = 51) 2 15 6 23
Lymphoma (n = 46)
7 13 6 26
5 5 18 28
* Some patients presented with physical abnormalities from more than one category. 15 March 1992 • Annals of Internal Medicine
• Volume 116 • Number 6
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19746/ by a University of California San Diego User on 06/07/2017
475
of 2.6 tissue examinations per patient. The overall yield was 50% (Table 5), and if cytologic and bone marrow examinations are excluded, the yield increased to 76%. Open and endoscopic biopsies and fine-needle aspiration gave the highest yields. Only 16 patients (9%) required three or more types of biopsies to reach diagnosis, and 13 patients required repeated biopsy of the same lesion because of initially equivocal results.
this accounted for the longer time to biopsy in this group. Once the lesion was detected, however, the same delay in proceeding to biopsy occurred. Although logistic and similar delays in proceeding to biopsy may sometimes be outside the control of the physician, analysis of the causes of delay in our patients revealed that such a situation occurred only in about 40% of patients. In 60% of the patients, delays to biopsy could only be attributed to continued noninvasive investigations. We took care to include only investigations done to diagnose the malignancy, excluding those done to monitor or manage complications or intercurrent problems. Many of these investigations had low yield (see Table 4). In fact, the overall yield in detecting the lesion was 24% (when excluding the minimum investigations of complete blood counts, biochemical screening, and chest radiographs). Overall, 57% of investigations were useful in defining the clinical problem. Protein studies, tumor markers, mammography, skeletal radiographs, intravenous pyelography and angiography, nuclear medicine scans, lower gastrointestinal tract endoscopy, and cytologic examinations had particularly low yields. The lack of gain from lower gastrointestinal tract endoscopy was particularly striking, with none contributing to diagnosis (see Table 4); this has been previously reported (7). Many of these investigations may have been averted if a biopsy had been done earlier. Subsequent investigations may then have been directed by the results of the biopsy, presumably with a higher yield.
Postbiopsy Investigations A total of 89 studies were performed after biopsy. These procedures were essentially staging investigations. The mean number of studies per patient was 0.5; this was similar to the mean number of investigations per patient in the gastrointestinal tract, lung, and lymphoma subgroups (0.5, 0.64, and 0.85, respectively). Patients with adenocarcinoma of unknown primary site had significantly fewer investigations after biopsy (mean, 0.08 per patient; P < 0.001). Discussion In this study of 177 evaluablc patients with various types of malignancies, a definite delay existed in proceeding to biopsy after a potentially biopsiable lesion was detected. These findings confirm our clinical impression that an excess number of noninvasive investigations account in large part for this delay. Although excessive investigations have been previously documented in patients with adenocarcinoma of unknown primary site (2, 6, 7), this possibility has not been formally evaluated in the diagnostic process in advanced cancer per se. The group we studied was selected from referrals by internists and specifically excluded surgical referrals. Apart from an excess of gastrointestinal tumors in men (M:F = 2:1), the sex distribution of our patients and their presenting clinical features were not unusual. Most of the patients (73% to 80%) were suspected of having cancer at first evaluation, and there was no significant difference among the four groups studied. Further, 52% of the lesions that eventually provided the definitive diagnosis were detected at the initial evaluation, with 67% detected by the second day. Despite this high detection rate, overall there was approximately an 8- to 10-day delay before a biopsy was done. This delay was consistent across the four malignancy groups studied. Lesions were more difficult to detect in patients with gastrointestinal tumors (25% at initial evaluation), and
The nature of many of these studies suggested that perhaps the housestaff or internists were attempting to reach a diagnosis by noninvasive techniques before proceeding to biopsy. This conforms to the traditional medical approach of proceeding from the least to more invasive investigations. In managing malignant disease, however, this approach results in overinvestigation, delay in diagnosis, and an overall increase in costs without avoiding the ultimate need for invasive procedures. It also may reflect a lack of appreciation that ultimately a histologic or cytologic diagnosis is usually necessary. Low-yield biopsy investigations also contributed to a lesser extent to a delay in obtaining a definitive result. Overall, 2.6 tissue investigations per patient were done. This excess was largely due to the low yield of cytologic examinations and bone marrow biopsies (Table 5). Open biopsies had the best chance (87%) of giving a diagnosis and, when possible, these should not be delayed by awaiting results of cytologic examinations.
Table 2. Mode of Detection of Definitively Biopsied Lesions Mode of Detection
Gastrointestinal Tract (n = 36)
Number of Patients in Each Category Adenocarcinoma of Lung Unknown Primary Site (n = 44)
476
n (%)
fl
4 23 9 0
Total (n = 111)
(ft = 51)
(
Clinical Radiologic Endoscopic Other
Lymphoma (n = 46)
13 26 5 0
IS 23 0 3
25 12 4 5
67 (38) 84 (47) 18 (10) 8(5)
15 March 1992 • Annals of Internal Medicine • Volume 116 • Number 6
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19746/ by a University of California San Diego User on 06/07/2017
Table 3. Investigations Excluding Blood Count, Biochemical Screen, and Chest Radiograph Investigation Radiography* Endoscopy Tumor markers Isotope scans Protein studies Otherf Total (mean)$
Gastrointestinal Tract 67 33 11 6 2 5 124 (3.4)
Lung
Adenocarcinoma of Unknown Primary Site
Lymphoma
Total
99 21 31 23 11 24 209(4.1)
57 17 2 7 15 36 134 (2.9)
251 95 52 42 31 111
28 24 8 6 3 46 115 (2.6)
* Radiography numbers exclude chest radiographs. t Includes cytologic examination of effusions, cerebrospinal fluid, cervical smears, and urine and bone marrow examinations. $ Numbers in parentheses are the mean number of investigations per patient in each group.
Fine-needle aspiration was also associated with a high yield (77%), and this procedure is easy to do and has minimal complications. Its high diagnostic accuracy when done by experienced pathologists has been previously reported (9). However, it gives no information regarding tissue architecture and so may not always avert the need for an open biopsy, especially in lymphoma. Bone marrow biopsy is more appropriately a staging investigation and thus deferring it until after a diagnosis is obtained may result in a lesser number being performed. Inconclusive results requiring a second biopsy occurred in only 13 patients and thus were not an important factor for delay in diagnosis. Another interesting finding related to the apparent overinvestigation of patients with adenocarcinoma of unknown primary site. Such patients represent about 2% to 9% of patients with cancer (1, 6, 10), and their average survival is 3 to 7 months (6, 11). Many reports have emphasized the low yield and cost ineffectiveness of aggressively investigating these patients in search of a primary tumor (2, 6, 7, 8). Despite this, our study suggested that patients with adenocarcinoma were still extensively investigated. These patients had a significantly higher number of prebiopsy investigations despite the fact that lesions were not more difficult to detect in this group. Further, from the nature of these
investigations and the observation that these patients had fewer postbiopsy investigations, it appears that these patients were extensively studied in search of a primary site before a biopsy was done. In one study, a comprehensive search for a primary tumor was estimated to cost an extra $2 to $8 million for the additional benefit of 6 months' survival for 1000 patients compared with a more limited search based on directed investigations after review of histologic findings (7). It is noteworthy that our population was skewed by the exclusion of nonhospitalized patients and patients from nonteaching hospitals. Records of such patients are often difficult to access, and it is unlikely that their inclusion would have significantly changed our results. We conclude that a substantial delay in proceeding to biopsy of a detected lesion exists in the internists' management of cancer. A large contribution to this delay comes from the performance of a series of noninvasive investigations and low-yield tissue examinations. Although many of these may have been initiated by house officers, it is the responsibility of the attending physicians to direct the housestaff to use the most direct and cost-effective methods of obtaining an accurate histologic diagnosis. We suggest that early biopsy with accurate histologic or cytologic diagnosis of malignancy would result in more directed noninvasive investigations
Table 4. Yield of Investigations Investigation
Number
Normal or Irrelevant
Abnormal and Relevant
Directly Leading to Biopsy (%)
251 11 26 39
66 9 11 6
98 1 4 14
87 (35) 1(9) 11 (42) 19 (49)
142 5
65 0
26 0
51 (36) 5 (100)
28 95 45 18 32 52 42 31 111
10 39 10 18 11 42 20 26 60
15 18 15 0 3 10 19 5 41
3(11) 38 (40) 20 (44) 0(0) 18 (56) 0(0) 3(7) 0 10(9)
Radiography* Mammography Contrast studies of gastrointestinal tract Ultrasonography Computed tomography and magnetic resonance scans Myelography Other (skeletal radiography, intravenous pyelography, angiography) Endoscopies Upper gastrointestinal Lower gastrointestinal Bronchoscopy and mediastinoscopy Tumor markers Isotope scans Protein studies Cytologic examinations! * Numbers exclude chest radiography. t Numbers include bone marrow cytologic examinations.
15 March 1992 • Annals of Internal Medicine • Volume 116 • Number 6
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19746/ by a University of California San Diego User on 06/07/2017
477
Table 5. Yield of Biopsies and Cytologic Examinations Investigation
Number
Definitive Result n(%)
Cytology of body fluids Fine needle aspiration True cut Bone marrow Endoscopic Open biopsy Total
124 53 27 44 51 77 376
24 (19) 41 (77) 13 (48) 6(14) 37 (63) 67 (87) 188 (50)
aimed at properly staging the patient with much higher yield and cost effectiveness. This would decrease the length of hospitalization and distress to the patient and would result in greater cost saving to the health system. Acknowledgments: The authors thank Miss L. Negro for manuscript preparation and Mrs. S. Oakes for maintenance of the patient registration database. Requests for Reprints: Richard M. Fox, MB, PhD, Department of Medical Oncology, Royal Melbourne Hospital, Post Office, Victoria 3050, Australia. Current Author Addresses: Dr. Farag: Department of Medicine, St. Vincent's Hospital, Victoria Parade, Fitzroy, Victoria 3065, Australia. Drs. Green, Sheridan, and Fox: Department of Clinical Haematology and Medical Oncology, The Royal Melbourne Hospital, Post Office, Victoria 3050, Australia.
Dr. Morstyn: Clinical and Medical Affairs, Amgen Center, 1840 Dehavilland Drive, Thousand Oaks, CA 91320-1789. References 1. Ultmann JE, Phillips TL. Cancer of Unknown primary site. In: DeVita VT, Helman S, Rosenberg SA; eds: Cancer: Principles and Practice of Oncology. Philadelphia: J.B. Lippincott; 1987-1843-53. 2. Stewart J F , Tattersall MH, Woods RL, Fox RM. Unknown primary adenocarcinoma: incidence of overinvestigation and natural history. Br Med J. 1979;1:1530-3. 3. Greco AF, Vaughan WK, Hainsworth JD. Advanced poorly differentiated carcinoma of unknown primary site: recognition of a treatable syndrome. Ann Intern Med. 1986;104:547-53. 4. Greenberg BR, Lawrence HJ. Metastatic cancer with unknown primary. Med Clin North Am. 1988;72:1055-65. 5. Ultmann JE, Golomb HM. Principles of neoplasia: approach to diagnosis and management. In: Adams RD, Braunwald E, Petersdorf RG, Wilson JD; eds. Harrison's Principles of Internal Medicine. New York: McGraw-Hill; 1980:1583-97. 6. Kirsten F, Chi CH, Leary JA, Ng AB, Hedley DW, Tattersall MH. Metastatic adeno or undifferentiated carcinoma from an unknown primary site—natural history and guidelines for identification of treatable subsets. OJ Med. 1987;238:143-61. 7. Levine MN, Drummond MF, La belle RJ. Cost-effectiveness in the diagnosis and treatment of cancer of unknown primary origin. Can Med Assoc J. 1985;133:977-87. 8. Nystrom JS, Weiner JM, Wolf RM, Bateman JR, Viola MV. Identifying the primary site in metastatic cancer of unknown origin. Inadequacy of roentgenographic procedures. JAMA. 1979;241:381-3. 9. Simpson GT 2d. The evaluation and management of neck masses of unknown etiology. Otolaryngol Clin North Am. 1980;13:489-98. 10. Nystrom JS, Weiner JM, Heffelfinger-Juttner J, Irwin LE, Bateman JR, Wolf RM. Metastatic and histologic presentations in unknown primary cancer. Semin Oncol. 1977;4:53-8. 11. Moertel CG, Reitemeier RJ, Schutt AJ, Hahn RG. Treatment of the patient with adenocarcinoma of unknown origin. Cancer. 1972;30: 1469-72.
Experience stands on its own dunghill in medicine and reason yields its place. Medicine has always professed experience to be the touchstone of its operations. Plato was right to say that good doctors themselves ought to have had the diseases they want to cure and been subject themselves to the misfortunes and circumstances which they have to diagnose. Let them catch the pox if they want to know how to cure it. I'd trust such a doctor. Michael deMontaigne, Essays III, Chapter 13
478
15 March 1992 • Annals of Internal Medicine • Volume 116 • Number 6
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19746/ by a University of California San Diego User on 06/07/2017
• Objective: To investigate the degree and type of delays in performing diagnostic biopsies in medical patients with suspected malignancy. • Design: Retrospective survey of clinical histories of patients referred between January 1985 and March 1989. • Setting: Inner city teaching hospital internal medicine (nononcologic) services. • Patients: Patients with gastrointestinal and lung cancers, adenocarcinoma of unknown primary site, and lymphomas were referred as inpatients by internists. Two hundred fifty-five patients were eligible, and 177 were evaluable. • Main Outcome Measures: The number, type, and results of tests done before and after biopsy were analyzed. • Results: In 67% of patients the biopsied lesion was detected by the second day of evaluation; however, there was an 8- to 10-day delay before a biopsy was done. This delay was consistent across the four malignancy groups studied. Although logistic and other unavoidable delays occurred in 40% of the cases, in 60% delays could only be attributed to continued, frequently low yield, noninvasive tests. An average of 3.3 tests were made per patient, with only 24% leading to a definitive biopsy. • Conclusion: Because of the performance of many other tests, a substantial delay exists in proceeding to biopsy during the diagnosis of cancer by internists. Annals of Internal Medicine. 1992;116:473-478. From The Royal Melbourne Hospital, Victoria, Australia. For current author addresses, see end of text.
tigations and a delay in diagnosis. Although such a delay will rarely affect treatment outcome, it is likely to increase the length of hospital stay, cause more distress to patients, and increase the costs to health care systems. The importance of securing an early histologic diagnosis in the investigative process of patients presenting with occult primary malignancies has been previously emphasized (1-4). Ultmann and Phillips (1), writing about occult primary malignancy, recommend reversing the traditional approach: . . . The optimal sequence of steps when evaluating the patients with [occult primary malignancies] concentrates on the re-evaluation of the biopsy material. Note that this differs from the traditional order of enquiry (history, physical examination, laboratory studies) used to delineate other clinical problems. This biopsy should be reviewed first . . . any clues that the biopsy provides will facilitate a directed history or physical examination with emphasis on specific areas of interest. This approach will also target the investigations and is applicable to other forms of malignancy. Further, Ultmann and Phillips (5) are even more direct: "Regardless of the difficulty of the procedures involved histological proof of malignancy must be obtained before the risks involved in further staging." To document our impression, we reviewed the clinical histories of 255 patients presenting with four different groups of malignancies referred to a medical oncology service by internists. We attempted to quantitate the extent and cause of delay in obtaining a diagnostic biopsy. Methods
1 he appropriate management of patients presenting with malignancy depends almost entirely on obtaining an histologic diagnosis because this usually dictates the most appropriate form of treatment. Obtaining a biopsy is, therefore, a prime task in the diagnostic process. It was our impression, and that of other medical oncologists, that internists do a large number of noninvasive investigations in an attempt to either diagnose or stage the patient before resorting to biopsy. This appeared to contrast with surgical practice where biopsies are done more quickly. Such a practice appears to originate in the traditional philosophy in internal medicine to proceed from the least to more invasive investigations. In the diagnosis of cancer, where tissue diagnosis is vital, this approach often results in many unnecessary inves-
Four groups of malignancies, including gastrointestinal carcinomas, non-small cell lung carcinoma, adenocarcinoma of unknown primary site and lymphomas (including Hodgkin disease), were considered in this study. Between January 1985 and March 1989, 255 patients with those malignancies were referred by internists from a single general teaching hospital. The patients were worked up as inpatients on general and specialist (nononcologic) internal medicine services. The attending physicians and housestaff, as a team, decided on the details of the investigative process and ultimate referral. An histologic or cytologic diagnosis of malignancy was a requirement for inclusion. The clinical histories were reviewed retrospectively to obtain the number, type, and results of investigations done before and after biopsy, as well as the number and type of biopsies. The times from initial evaluation to suspicion of cancer, to detection of the subsequently biopsied lesion, and to the actual biopsy were also recorded. Specific delays in doing a biopsy were noted. The data was coded and analysed on a VAX computer (Digital Equipment Corporation, Maynard, Massachusetts) using the SPSSX package (SPSS Inc., Chicago, Illinois). ©1992 American College of Physicians
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19746/ by a University of California San Diego User on 06/07/2017
473
Figure 1. Delay in diagnostic biopsy. Time taken to suspect cancer as a diagnosis, detect an ultimately biopsied lesion, and perform the actual definitive diagnostic biopsy. Cumulative frequency plot with time on a logarithmic scale.
Clinical Presentations and Investigations Presenting symptoms and signs were recorded and those unrelated to the underlying malignancy were excluded from analysis. Investigations done before and after biopsy were recorded separately. Investigations done before biopsy were those performed in an attempt to reach or clarify the diagnosis of malignancy or its apparent stage. Investigations done for other reasons, such as management of concurrent illnesses or intercurrent complications, were not included in the analysis. Investigations that were repeated only to verify the initial investigation were included in the analysis. The yield of the investigations done before biopsy was examined according to whether the results led to definitive biopsy, whether they were abnormal but relevant, or whether they were "unhelpful" (including normal results, technical failures, or results irrelevant to the problem). Because the investigations after the biopsy were essentially staging investigations, their results were not analyzed. Biopsies The methods by which the definitively biopsied lesion was detected were analyzed. Because some patients had more than one biopsy, the mode of detection of the definitively biopsied lesion that gave the final diagnosis was the one recorded. Analysis of the methods of biopsy and their yield was also done to determine if a requirement for multiple biopsies and low-yield methods of biopsy contributed to delay in diagnosis. Specific delays other than those due to a prolonged investigative pathway were identified. These fell into three categories: "Patient unfit for biopsy procedure," "logistic problems," and other. Logistic problems included delays in obtaining investigation time or operating room time. A logistic problem was considered to be present when 3 or more days elapsed from the time of the last investigation to the biopsy and where it was clear in the clinical notes that there was an intention to do a biopsy. Results Of the 255 eligible patients, 177 (98 men and 79 women) were evaluable and included in the analysis. The remaining patients were excluded primarily because the diagnosis was made outside the hospital or because the clinical notes were inadequate to allow assessment. The mean age of the study group was 61.4 years (range, 474
15 to 86 years), with patients with lymphoma being significantly younger (mean age, 55 years) than the other groups of patients. One hundred twenty-five patients were referred by general internists and 52, by specialist internists; 168 of these patients were investigated wholly as inpatients. Clinical Presentations Almost all patients were suspected of having malignant disease at the time of the first evaluation (Figure 1). Eighty percent presented with physical signs that led to discovery of the biopsiable lesion. (These signs were either an abnormality that could be directly biopsied or an abnormality that initiated an investigation detecting a lesion amenable to biopsy, for example, signs of pulmonary consolidation leading to performance of a chest radiograph and detection of a biopsiable lung mass.) Eighty-five to ninety percent of patients with lung carcinoma, adenocarcinoma of unknown primary site, or lymphoma had physical signs on presentation that led to detection of a biopsiable lesion. In contrast, 17 of 36 patients with gastrointestinal tract cancer had no such signs. Fifty-nine percent of patients had more than one sign. Table 1 shows the distribution of lesions detected at the initial physical examination that were amenable to direct histologic or cytologic examination without any intermediate investigations. Detection of the Definitively Biopsied Lesion The definitively biopsied lesion was usually first detected either clinically or radiologically (Table 2). Overall clinical detection of the definitively biopsied lesion occurred in 67 (38%) patients and was more common in patients with lymphomas (54%) and adenocarcinoma of unknown primary site (50%) than in patients with lung (30%) or gastrointestinal tract (11%) cancer. Conversely, radiologic detection was more common in patients with gastrointestinal tract (64%) and lung (59%)
15 March 1992 • Annals of Internal Medicine • Volume 116 • Number 6
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19746/ by a University of California San Diego User on 06/07/2017
malignancies. The least common initial mode of detection was endoscopy (10%). Altogether, 52% of definitively biopsied lesions were detected at the initial evaluation and 67% were detected by the second day of evaluation (see Figure 1). Subgroup analysis showed that a greater delay occurred in detecting definitively biopsied lesions in the gastrointestinal tract malignancies compared with the other groups. Twenty-five percent of gastrointestinal lesions were detected at the first evaluation, 47% by day 2, and 67% by day 5; this delay was significantly greater than in the other patient groups (P = 0.005). The delay in the detection of gastrointestinal lesions compared with the other tumor groups occurred despite the fact that no difference between the groups was found with respect to when cancer was first suspected. Presumably, this greater delay reflected the scarcity of physical signs in the gastrointestinal tract group. Time to Definitive Biopsy The median time to definitive biopsy for the whole population was 8 days (see Figure 1). There were no significant differences between median times for patients with lung cancer, adenocarcinoma of unknown primary site, and lymphoma (9.5 days, 8 days, and 9 days, respectively). However, the median time to biopsy of 13 days was significantly longer for the patients with gastrointestinal tract disease (P = 0.03). Delays in Obtaining a Biopsy When cumulative frequency plots of times to suspicion of cancer, detection of the definitively biopsied lesion, and definitive biopsy for the whole study group are compared, it is apparent that a substantial delay existed between when the lesion was detected and when it was biopsied (see Figure 1). Although more than half of the lesions were detected at first evaluation, there was about an 8-day delay to biopsy. In 106 (60%) patients, delay appeared to be caused by a prolonged investigative pathway after the definitively biopsied lesion was detected. Delays other than a "prolonged investigative pathway" were present in 71 (40%) patients. The latter included delays due to logistic factors (38 patients), lack of fitness of patients to undergo the biopsy procedure (6 patients), equivocal biopsy results requiring a second biopsy (13 patients), biopsiable lesions missed at initial evaluation (10 patients), reluc-
tance of patients to have a biopsy (3 patients), and temporary loss of follow-up (1 patient). The Prebiopsy "Prolonged Investigative Pathway" A total of 1079 investigations were done, resulting in a mean of 6.1 investigations per patient. Patients with adenocarcinoma of unknown primary site had an average of 7.0 investigations each. This number was significantly higher than the numbers in the other groups (gastrointestinal tract, 5.9; lung, 5.5; lymphoma, 5.8; P = 0.001). Complete blood counts were done on 176 patients, routine serum biochemical screening (including liver function tests) was done on 173 patients, and chest radiographs were done on 148 patients. Because these three studies were done on almost all patients and could be reasonably considered as minimum investigations for a patient with possible malignancy, they were excluded from further analysis. After correcting for the exclusion of the "minimum" investigations, the mean number of investigations was 3.3 investigations (Table 3); patients with adenocarcinoma of unknown primary site still had a substantially higher number. The number of radiologic and endoscopic procedures in this group (251 and 95, respectively) was striking. Only 156 of the total of 1079 (15%) investigations led to a definitive biopsy. If investigations that yielded abnormal and relevant results but did not directly lead to biopsy are included, then 530 (49%) investigations may be deemed "useful." When the minimum investigations are excluded from analysis, 141 (24%) of the investigations led to definitive biopsy and a further 191 (33%) investigations were useful. There still remained a large percentage of investigations (43%) that were not useful and contributed little to defining the problem. The yield of individual investigations is shown in Table 4. Investigations with the highest yield in detecting biopsied lesions were radiologic (ultrasound, computed tomography, magnetic resonance scans, and contrast studies of gastrointestinal tract and spine) and endoscopic (other than lower gastrointestinal tract endoscopy), for which the overall yields were 35% and 40%, respectively. Biopsies A total of 252 biopsies and 124 cytologic examinations of body fluids were done, with an overall average
Table 1. Distribution of Abnormalities, Amenable to Biopsy, Found at the Initial Physical Examination among Patients with Cancer Physical Abnormality* Gastrointestinal Tract (n = 36) Palpable mass Effusion Superficial lymphadenopathy Total with at least one sign
6 3 1 10
Number of Patients in Each Category Adenocarcinoma Lung of Unknown Primary Site (n - 44) (n = 51) 2 15 6 23
Lymphoma (n = 46)
7 13 6 26
5 5 18 28
* Some patients presented with physical abnormalities from more than one category. 15 March 1992 • Annals of Internal Medicine
• Volume 116 • Number 6
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19746/ by a University of California San Diego User on 06/07/2017
475
of 2.6 tissue examinations per patient. The overall yield was 50% (Table 5), and if cytologic and bone marrow examinations are excluded, the yield increased to 76%. Open and endoscopic biopsies and fine-needle aspiration gave the highest yields. Only 16 patients (9%) required three or more types of biopsies to reach diagnosis, and 13 patients required repeated biopsy of the same lesion because of initially equivocal results.
this accounted for the longer time to biopsy in this group. Once the lesion was detected, however, the same delay in proceeding to biopsy occurred. Although logistic and similar delays in proceeding to biopsy may sometimes be outside the control of the physician, analysis of the causes of delay in our patients revealed that such a situation occurred only in about 40% of patients. In 60% of the patients, delays to biopsy could only be attributed to continued noninvasive investigations. We took care to include only investigations done to diagnose the malignancy, excluding those done to monitor or manage complications or intercurrent problems. Many of these investigations had low yield (see Table 4). In fact, the overall yield in detecting the lesion was 24% (when excluding the minimum investigations of complete blood counts, biochemical screening, and chest radiographs). Overall, 57% of investigations were useful in defining the clinical problem. Protein studies, tumor markers, mammography, skeletal radiographs, intravenous pyelography and angiography, nuclear medicine scans, lower gastrointestinal tract endoscopy, and cytologic examinations had particularly low yields. The lack of gain from lower gastrointestinal tract endoscopy was particularly striking, with none contributing to diagnosis (see Table 4); this has been previously reported (7). Many of these investigations may have been averted if a biopsy had been done earlier. Subsequent investigations may then have been directed by the results of the biopsy, presumably with a higher yield.
Postbiopsy Investigations A total of 89 studies were performed after biopsy. These procedures were essentially staging investigations. The mean number of studies per patient was 0.5; this was similar to the mean number of investigations per patient in the gastrointestinal tract, lung, and lymphoma subgroups (0.5, 0.64, and 0.85, respectively). Patients with adenocarcinoma of unknown primary site had significantly fewer investigations after biopsy (mean, 0.08 per patient; P < 0.001). Discussion In this study of 177 evaluablc patients with various types of malignancies, a definite delay existed in proceeding to biopsy after a potentially biopsiable lesion was detected. These findings confirm our clinical impression that an excess number of noninvasive investigations account in large part for this delay. Although excessive investigations have been previously documented in patients with adenocarcinoma of unknown primary site (2, 6, 7), this possibility has not been formally evaluated in the diagnostic process in advanced cancer per se. The group we studied was selected from referrals by internists and specifically excluded surgical referrals. Apart from an excess of gastrointestinal tumors in men (M:F = 2:1), the sex distribution of our patients and their presenting clinical features were not unusual. Most of the patients (73% to 80%) were suspected of having cancer at first evaluation, and there was no significant difference among the four groups studied. Further, 52% of the lesions that eventually provided the definitive diagnosis were detected at the initial evaluation, with 67% detected by the second day. Despite this high detection rate, overall there was approximately an 8- to 10-day delay before a biopsy was done. This delay was consistent across the four malignancy groups studied. Lesions were more difficult to detect in patients with gastrointestinal tumors (25% at initial evaluation), and
The nature of many of these studies suggested that perhaps the housestaff or internists were attempting to reach a diagnosis by noninvasive techniques before proceeding to biopsy. This conforms to the traditional medical approach of proceeding from the least to more invasive investigations. In managing malignant disease, however, this approach results in overinvestigation, delay in diagnosis, and an overall increase in costs without avoiding the ultimate need for invasive procedures. It also may reflect a lack of appreciation that ultimately a histologic or cytologic diagnosis is usually necessary. Low-yield biopsy investigations also contributed to a lesser extent to a delay in obtaining a definitive result. Overall, 2.6 tissue investigations per patient were done. This excess was largely due to the low yield of cytologic examinations and bone marrow biopsies (Table 5). Open biopsies had the best chance (87%) of giving a diagnosis and, when possible, these should not be delayed by awaiting results of cytologic examinations.
Table 2. Mode of Detection of Definitively Biopsied Lesions Mode of Detection
Gastrointestinal Tract (n = 36)
Number of Patients in Each Category Adenocarcinoma of Lung Unknown Primary Site (n = 44)
476
n (%)
fl
4 23 9 0
Total (n = 111)
(ft = 51)
(
Clinical Radiologic Endoscopic Other
Lymphoma (n = 46)
13 26 5 0
IS 23 0 3
25 12 4 5
67 (38) 84 (47) 18 (10) 8(5)
15 March 1992 • Annals of Internal Medicine • Volume 116 • Number 6
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19746/ by a University of California San Diego User on 06/07/2017
Table 3. Investigations Excluding Blood Count, Biochemical Screen, and Chest Radiograph Investigation Radiography* Endoscopy Tumor markers Isotope scans Protein studies Otherf Total (mean)$
Gastrointestinal Tract 67 33 11 6 2 5 124 (3.4)
Lung
Adenocarcinoma of Unknown Primary Site
Lymphoma
Total
99 21 31 23 11 24 209(4.1)
57 17 2 7 15 36 134 (2.9)
251 95 52 42 31 111
28 24 8 6 3 46 115 (2.6)
* Radiography numbers exclude chest radiographs. t Includes cytologic examination of effusions, cerebrospinal fluid, cervical smears, and urine and bone marrow examinations. $ Numbers in parentheses are the mean number of investigations per patient in each group.
Fine-needle aspiration was also associated with a high yield (77%), and this procedure is easy to do and has minimal complications. Its high diagnostic accuracy when done by experienced pathologists has been previously reported (9). However, it gives no information regarding tissue architecture and so may not always avert the need for an open biopsy, especially in lymphoma. Bone marrow biopsy is more appropriately a staging investigation and thus deferring it until after a diagnosis is obtained may result in a lesser number being performed. Inconclusive results requiring a second biopsy occurred in only 13 patients and thus were not an important factor for delay in diagnosis. Another interesting finding related to the apparent overinvestigation of patients with adenocarcinoma of unknown primary site. Such patients represent about 2% to 9% of patients with cancer (1, 6, 10), and their average survival is 3 to 7 months (6, 11). Many reports have emphasized the low yield and cost ineffectiveness of aggressively investigating these patients in search of a primary tumor (2, 6, 7, 8). Despite this, our study suggested that patients with adenocarcinoma were still extensively investigated. These patients had a significantly higher number of prebiopsy investigations despite the fact that lesions were not more difficult to detect in this group. Further, from the nature of these
investigations and the observation that these patients had fewer postbiopsy investigations, it appears that these patients were extensively studied in search of a primary site before a biopsy was done. In one study, a comprehensive search for a primary tumor was estimated to cost an extra $2 to $8 million for the additional benefit of 6 months' survival for 1000 patients compared with a more limited search based on directed investigations after review of histologic findings (7). It is noteworthy that our population was skewed by the exclusion of nonhospitalized patients and patients from nonteaching hospitals. Records of such patients are often difficult to access, and it is unlikely that their inclusion would have significantly changed our results. We conclude that a substantial delay in proceeding to biopsy of a detected lesion exists in the internists' management of cancer. A large contribution to this delay comes from the performance of a series of noninvasive investigations and low-yield tissue examinations. Although many of these may have been initiated by house officers, it is the responsibility of the attending physicians to direct the housestaff to use the most direct and cost-effective methods of obtaining an accurate histologic diagnosis. We suggest that early biopsy with accurate histologic or cytologic diagnosis of malignancy would result in more directed noninvasive investigations
Table 4. Yield of Investigations Investigation
Number
Normal or Irrelevant
Abnormal and Relevant
Directly Leading to Biopsy (%)
251 11 26 39
66 9 11 6
98 1 4 14
87 (35) 1(9) 11 (42) 19 (49)
142 5
65 0
26 0
51 (36) 5 (100)
28 95 45 18 32 52 42 31 111
10 39 10 18 11 42 20 26 60
15 18 15 0 3 10 19 5 41
3(11) 38 (40) 20 (44) 0(0) 18 (56) 0(0) 3(7) 0 10(9)
Radiography* Mammography Contrast studies of gastrointestinal tract Ultrasonography Computed tomography and magnetic resonance scans Myelography Other (skeletal radiography, intravenous pyelography, angiography) Endoscopies Upper gastrointestinal Lower gastrointestinal Bronchoscopy and mediastinoscopy Tumor markers Isotope scans Protein studies Cytologic examinations! * Numbers exclude chest radiography. t Numbers include bone marrow cytologic examinations.
15 March 1992 • Annals of Internal Medicine • Volume 116 • Number 6
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19746/ by a University of California San Diego User on 06/07/2017
477
Table 5. Yield of Biopsies and Cytologic Examinations Investigation
Number
Definitive Result n(%)
Cytology of body fluids Fine needle aspiration True cut Bone marrow Endoscopic Open biopsy Total
124 53 27 44 51 77 376
24 (19) 41 (77) 13 (48) 6(14) 37 (63) 67 (87) 188 (50)
aimed at properly staging the patient with much higher yield and cost effectiveness. This would decrease the length of hospitalization and distress to the patient and would result in greater cost saving to the health system. Acknowledgments: The authors thank Miss L. Negro for manuscript preparation and Mrs. S. Oakes for maintenance of the patient registration database. Requests for Reprints: Richard M. Fox, MB, PhD, Department of Medical Oncology, Royal Melbourne Hospital, Post Office, Victoria 3050, Australia. Current Author Addresses: Dr. Farag: Department of Medicine, St. Vincent's Hospital, Victoria Parade, Fitzroy, Victoria 3065, Australia. Drs. Green, Sheridan, and Fox: Department of Clinical Haematology and Medical Oncology, The Royal Melbourne Hospital, Post Office, Victoria 3050, Australia.
Dr. Morstyn: Clinical and Medical Affairs, Amgen Center, 1840 Dehavilland Drive, Thousand Oaks, CA 91320-1789. References 1. Ultmann JE, Phillips TL. Cancer of Unknown primary site. In: DeVita VT, Helman S, Rosenberg SA; eds: Cancer: Principles and Practice of Oncology. Philadelphia: J.B. Lippincott; 1987-1843-53. 2. Stewart J F , Tattersall MH, Woods RL, Fox RM. Unknown primary adenocarcinoma: incidence of overinvestigation and natural history. Br Med J. 1979;1:1530-3. 3. Greco AF, Vaughan WK, Hainsworth JD. Advanced poorly differentiated carcinoma of unknown primary site: recognition of a treatable syndrome. Ann Intern Med. 1986;104:547-53. 4. Greenberg BR, Lawrence HJ. Metastatic cancer with unknown primary. Med Clin North Am. 1988;72:1055-65. 5. Ultmann JE, Golomb HM. Principles of neoplasia: approach to diagnosis and management. In: Adams RD, Braunwald E, Petersdorf RG, Wilson JD; eds. Harrison's Principles of Internal Medicine. New York: McGraw-Hill; 1980:1583-97. 6. Kirsten F, Chi CH, Leary JA, Ng AB, Hedley DW, Tattersall MH. Metastatic adeno or undifferentiated carcinoma from an unknown primary site—natural history and guidelines for identification of treatable subsets. OJ Med. 1987;238:143-61. 7. Levine MN, Drummond MF, La belle RJ. Cost-effectiveness in the diagnosis and treatment of cancer of unknown primary origin. Can Med Assoc J. 1985;133:977-87. 8. Nystrom JS, Weiner JM, Wolf RM, Bateman JR, Viola MV. Identifying the primary site in metastatic cancer of unknown origin. Inadequacy of roentgenographic procedures. JAMA. 1979;241:381-3. 9. Simpson GT 2d. The evaluation and management of neck masses of unknown etiology. Otolaryngol Clin North Am. 1980;13:489-98. 10. Nystrom JS, Weiner JM, Heffelfinger-Juttner J, Irwin LE, Bateman JR, Wolf RM. Metastatic and histologic presentations in unknown primary cancer. Semin Oncol. 1977;4:53-8. 11. Moertel CG, Reitemeier RJ, Schutt AJ, Hahn RG. Treatment of the patient with adenocarcinoma of unknown origin. Cancer. 1972;30: 1469-72.
Experience stands on its own dunghill in medicine and reason yields its place. Medicine has always professed experience to be the touchstone of its operations. Plato was right to say that good doctors themselves ought to have had the diseases they want to cure and been subject themselves to the misfortunes and circumstances which they have to diagnose. Let them catch the pox if they want to know how to cure it. I'd trust such a doctor. Michael deMontaigne, Essays III, Chapter 13
478
15 March 1992 • Annals of Internal Medicine • Volume 116 • Number 6
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19746/ by a University of California San Diego User on 06/07/2017
