Although basing credentialing on completion of a minimum number of procedures is a convenient and widely used practice, some problems are associated with this approach. Learners progress at different rates, and the number of procedures done is no guarantee of proficiency. Different credentialing bodies may require different numbers for the same procedure. Once a minimum number is specified, it often becomes the de facto standard for training: In effect, the floor becomes the ceiling. In addition, the requisite number has not been established for most procedures.

In the absence of objective data, such recommendations have relied on expert consensus. The few studies that have assessed the amount of training needed have found that the number of procedures that must be done for the trainee to attain competence in the procedure independently is higher than the minimum levels recommended by expert panels. Hawes and colleagues [3] graded the performance of 25 residents as they learned flexible sigmoidoscopy. The investigators found that insertion distance, number of lesions correctly identified, number of correct diagnoses, and management scores continued to improve with experience. Most residents required 24 to 30 procedures to become competent in the technique, twice the number ordinarily recommended for credentialing. Cass and colleagues [4] recorded the progress of trainees learning to do esophagogastroduodenoscopy and colonoscopy. Although technical success appeared to come early, it was a "false summit," due in part to the selection of easy cases for the novice trainees. Many more procedures were needed to achieve similar success with the more difficult cases encountered later in training. According to Cass and colleagues' criteria, more than 100 esophagogastroduodenoscopies and colonoscopies each had to be done to achieve technical competency. This estimate was twice the median of 50 recommended by 510 gastroenterologist members of the American College of Physicians in a national survey [5].

In this issue, Jowell and colleagues [6] report on the competence of gastroenterology fellows in endoscopic retrograde cholangiopancreatography (ERCP) as a function of the number of procedures done under supervision. During a period of 2 years, seven supervising physicians graded the fellows' performance on each procedure using two scales: a global rating of the trainees' competence and a rating of technical proficiency. Fellows' evaluations were considered in blocks of 20 ERCP procedures. To achieve overall proficiency, trainees had to obtain satisfactory competence ratings in all 20 procedures and achieve technical success in 80%. On the basis of actual scores and statistical projections from the scores of fellows who had not yet achieved mastery, the authors concluded that the number of procedures needed for overall competence was between 180 and 200. The authors found, as had Hawes and Cass and their colleagues, that the number of procedures needed for mastery exceeded the estimates of practitioners and expert panels. The median estimate from the national survey of gastroenterologists, for example, was 50 [5].

Were the standards of Jowell and colleagues too high? How expert must a trainee become? Some procedures, such as flexible sigmoidoscopy, may serve as a screening procedure in one setting and as a definitive procedure (including biopsy and therapeutic interventions, such as cautery and lesion removal) in another. However, it is more difficult to make this case for ERCP, which requires a high level of expertise throughout. Several aspects of Jowell and colleagues' study, however, may limit the generalizability of their conclusions about the number of procedures required. The study does not provide information on the reliability and reproducibility of the ratings given by the attending physicians. We do not know what factors made up the subjective ratings of competence and whether they were used consistently, particularly in balancing the importance of cognitive and technical skills. Because many of the fellows did not complete their training in ERCP during the study period, the numbers rely on statistical projections. Nevertheless, the study convincingly shows that ERCP is a technically difficult procedure that requires considerable experience to master.

What do these results mean for ERCP training and performance? If this standard is accepted, some gastroenterology fellowships will not be able to provide a volume of procedures sufficient to train their fellows in ERCP; ultimately, fewer fellows will be trained. There is always some degree of compromise between ideal performance standards and what can be done in the practice environment. The benefit of saving patients from repeated procedures, for example, must be balanced against the cost to a larger group of patients who must have their procedures done at a distant center rather than nearby. If the standards used in Jowell and colleagues' study are appropriate, what percentage of persons now doing ERCP procedures would meet them? In our survey of gastroenterologist members of the American College of Physicians [5], the 337 respondents who said that they did ERCP in their practices performed a median of 30 procedures per year; 25% did 5 or fewer per year [5]. Of respondents who had graduated from medical school before 1970, more than half said they had learned ERCP in practice after completing their fellowships. What do we know of their level of performance and how it compares to the proposed standard? It is difficult to argue for having different standards for trainees than for practicing physicians, and credentialing bodies may, at some point, need to develop equivalent standards.

How do these findings on ERCP apply to other procedures? Endoscopic retrograde cholangiopancreatography and other endoscopy-based procedures are among the most technically difficult procedures that internists do. For this reason, agreement on standards for competency is probably easier because the technical end points can be measured and counted. The endoscopic procedures are atypical in this respect because for many medical procedures, the most difficult skills to master are cognitive-knowledge of complications and their management, recognition of lesions, judgment about whether or not the procedure is needed-although these concerns are also highly relevant to ERCP and other complex procedures. Such procedures as abdominal paracentesis, bone marrow biopsy, thoracentesis, and elective cardioversion can be technically mastered in a short time, but acquiring the knowledge needed for unsupervised performance may take considerably more experience. The problem is that, for most procedures, we do not have standards for competence. Pioneers such as Hawes, Cass, and Jowell and their colleagues have shown that we can measure the acquisition of skill with training and that the results may not be what we expected. Whether these investigators have chosen the right standards is not the issue-that is for the profession to debate and explore. The point is that they have shown that we can set outcome standards and use them to measure competency.

Meeting patients' expectations for uniformly safe and successful procedures requires evaluation and record keeping more intense than that allowed by current practices. Residents learn at different rates, and each resident's progress can be successfully measured against a standard. The average number of procedures needed to achieve competency is helpful in estimating the volume of experience needed for training, but the average is not sufficient for roughly half of the trainees. Certification of competence in procedural skills needs to move beyond experience-based criteria to standards based on proven proficiency, measured outcomes, and evaluation by skilled teachers. Jowell and colleagues are pointing the way.