National Comprehensive Cancer Network

Clinical Recommendations

About the NCCN Clinical Practice Guidelines in Oncology™

The NCCN Clinical Practice Guidelines in Oncology™ – the recognized standard for clinical policy in oncology – are the most comprehensive and most frequently updated clinical practice guidelines available in any area of medicine. Covering 97 percent of all patients with cancer and updated on a continual basis, the NCCN Guidelines are developed through an explicit review of the evidence integrated with expert medical judgment by multidisciplinary panels from NCCN Member Institutions. Treatment recommendations are specific and are implemented through performance measurement. NCCN Guidelines Panels address cancer detection, prevention and risk reduction, workup and diagnosis, treatment and supportive care.

The NCCN Clinical Practice Guidelines in Oncology™

Rodger J. Winn, MD, and Joan S. McClure, MS, NCCN Senior Vice President, Clinical Information & Publications, February 2003
Revised January 2008

• Overview
• Guidelines Infrastructure
• Development Process
• Continuous Review
• Guidelines Methodology
• Summary
• References

---

Overview

In 1995, the National Comprehensive Cancer Network (NCCN) began a program to develop a comprehensive set of diagnostic, treatment, and supportive care guidelines (NCCN, 1996; 1997; 1998). The NCCN Guidelines, developed as algorithms, encompass 97% of the tumors encountered in oncology practices. In addition, NCCN Guidelines have been developed to provide recommendations for managing the major symptoms experienced by patients with cancer and a set of pathways detailing the major early diagnostic steps for breast, colon, and prostate cancers. Each NCCN Guideline consists of an algorithm or decision pathway outlining care management, a manuscript discussing important issues related to the algorithm, and references providing data on which recommendations are based.

The NCCN Guidelines are developed and updated by 44 individual panels, comprising of more than 800 clinicians and oncology researchers from the 21 NCCN Member Institutions and their affiliates. The NCCN Guidelines are composed of recommendations based on the best evidence available at the time they are derived, but it is essential that they be continuously updated and revised to reflect new data and new clinical information. The aim of the NCCN Guidelines is to assist oncologists in making the major clinical decisions encountered in managing their patients.

In order to make the best use of the NCCN Guidelines, it is helpful for oncologists to understand the methods underlying the development process. This introduction reviews the NCCN Guidelines infrastructure and details the steps in the development process. The most important precept to keep in mind as the NCCN Guidelines are being used is that they are dynamic documents.

NCCN Clinical Practice Guidelines in Oncology™

Clinical practice guidelines are defined as “systematically developed statements to assist practitioner and patient decisions about appropriate health care for specific clinical circumstances” (Field and Lohr, 1990). This astute definition provides the conceptual framework for developing a guideline. The purpose of clinical practice guidelines is to enhance clinical decision-making. A major task of the NCCN Guideline Panels is to develop pathways reflecting the major step-by-step management decisions required to provide care for patients with this chronic illness. Because the disease process is often complex, the pathways may be complex. Otherwise, the utility of the guidelines would be limited, and important aspects of care might not be covered. Once the sequence of decision points has been defined, specific recommendations are provided to assist the practitioner and patient.

Another major element of the definition is that the guidelines provide information regarding appropriate care. The determination of what is appropriate care involves assessing the balance between the sum of the benefits compared with the sum of the risks (Park et al, 1986). It is therefore important to make subjective judgments about relative weights. A procedure with a high mortality rate (e.g., allogeneic bone marrow transplantation) may be considered appropriate if it cures patients with a rapidly progressing fatal illness, but inappropriate in patients who may live for many years before they succumb to their illness. There are no “cookbook” formulas to determine the boundary between appropriate and inappropriate care, and a major role of guideline developers is to provide an evaluative function to the analysis of data.

The third important component of the definition of clinical practice guidelines is that they must be systematically developed. The NCCN evidence-based approach is described below to allow the user to understand how the algorithm recommendations are derived and how evidence from important studies, as well as expert clinical experience, is incorporated into the algorithms.

NCCN Guidelines Infrastructure

Guidelines Steering Committee

The NCCN Guidelines Steering Committee supervises the guidelines program and reports directly to the NCCN Board of Directors. Each NCCN Member Institution has a representative on this committee who is responsible for coordinating the guidelines activities at his or her institution. The responsibilities of the NCCN Guidelines Steering Committee include:

• Selection of Topics for Guidelines Development: NCCN Guideline topics are based on prevalence of disease, relevance of the area to oncology practice, need for guidelines, and availability of appropriate NCCN expertise. The NCCN Guidelines Steering Committee also selects panel chairs and provides lists of possible candidates from the NCCN Member Institutions.
• Promulgation of NCCN Guidelines Development Policies: Although each panel determines the specific recommendations of a pathway, the NCCN Guidelines Steering Committee is responsible for developing policies to ensure consistency across disease sites.
• Coordination of NCCN Guidelines Activities within NCCN Member Institutions: The NCCN Guidelines Steering Committee members are responsible for naming the institutional representative to each NCCN Guideline panel, overseeing the institutional review process (see below), and coordinating implementation activities.

NCCN Guideline Panels

A panel of experts with members from each of the NCCN Member Institutions develops each guideline. Currently, 44 panels meet at least annually. The NCCN Guideline Panel Members are chosen by the panel chair and the chair of the NCCN Guidelines Steering Committee from a list of institutional nominees. In some instances when special expertise exists, several representatives from an individual NCCN Member Institution may be included. In several NCCN Guideline Panels (e.g., myelodysplasia, distress management), outside consultants have joined NCCN Panel Members to contribute their unique expertise when it was not available within the organization.

Specialties that must be included on a particular panel are identified before that panel is convened. Surgical oncology, radiation oncology, medical oncology, and important specialties, such as gynecological oncology, neuro-oncology, and gastroenterology, are included as indicated. This multidisciplinary representation varies from panel to panel (e.g., the Bone Cancer Panel required a physiatrist, the CNS Panel required a neuropathologist). The participation of nurse oncologists, social workers, and other health professionals was essential to develop recommendations for symptom management and support services pathways. Many of the panels also include a patient representative, especially when issues of long-term care and patient preference are paramount in the panel's considerations.

The inclusion of different management philosophies is equally important in approaching specific tumors. All major strategies of disease management must be considered so that the NCCN Guideline will be acceptable to the broad NCCN membership. The NCCN Guideline Panel Chairs are charged with ensuring that representatives of all treatment strategies are included. This often leads to protracted and even intense panel discussions, but the process has been designed to foster dialogue and, if necessary, accommodate disagreement (See the NCCN Categories of Evidence and Consensus below).

NCCN Guideline Staff

As the program has expanded to include both an increasing number of guidelines and new electronic and paper formats including the JNCCN – The Journal of the National Comprehensive Cancer Network, the staff dedicated to the program has also grown. Developing and maintaining the program requires a diverse skill set and a staff that can be counted upon to roll up their sleeves and do whatever is necessary to develop, update, produce, and publish the NCCN Guidelines. Tasks juggled by staff include: scheduling panel meetings, attending NCCN Guideline Panel meetings and recording NCCN Guideline Panel decisions, developing algorithms that reflect NCCN Guideline Panel consensus, supporting manuscript production, obtaining several levels of approval on each NCCN Guideline, formatting and proofreading guidelines, and handling requests for permission to use the published NCCN Guidelines. And this is without any other “duties as assigned!” It is only through the unstinting efforts of this dedicated staff that the enormous demands of the program can be accomplished.

They are:

Joan S. McClure, MS
Senior Vice President, Clinical Information & Publications

Kristina M. Gregory, RN, MSN, OCN
Vice President, Clinical Information Operations

Dorothy A. Shead, MS
Associate Director, Clinical Information Operations

Nicole McMillian, MS
Guidelines Coordinator

Mary Dwyer Rosario, MS
Guidelines Coordinator:

Maria Ho, PhD
Oncology Scientist/Senior Medical Writer

Miranda Hughes, PhD
Oncology Scientist/Senior Medical Writer

Rashmi Kumar, PhD
Oncology Scientist/Senior Medical Writer

Susan Moench, PhD
Oncology Scientist/Senior Medical Writer

Hema Sundar, PhD
Oncology Scientist/Senior Medical Writer

Kimberly A. Callan, MS, ELS
Director of NCCN Publications

Kerrin Robinson, MA
Medical/Scientific Editor

Genevieve M. Hartzman, MA
Editorial Associate

Jean Marie Dougherty
Administrative Coordinator

NCCN Guidelines Development Process

Given the scope of the NCCN Guidelines — the entire spectrum of clinical decisions for virtually all tumor sites — the NCCN Guideline Panel decisions are based upon evaluation of scientific data integrated with expert judgment. Many of the NCCN Guideline Panels incorporate formal presentations of data for specific controversial issues into the NCCN Guideline meetings. Because the NCCN Guideline Panels are composed of tumor specialists from each of the NCCN Member Institutions, NCCN Panel Members have in-depth knowledge of the literature and awareness of, if not actual participation in, the trials that provide the evidence for the formulations (Winn, 2001).

Initial Development

The NCCN Guidelines development process relies heavily on structured feedback. The NCCN Guidelines are developed by one group of experts, then reviewed by another group of experts, which leads to revisions and modifications, reinitiating the feedback process. Figure 1 depicts the NCCN Guidelines development process, incorporating the features of iteration and feedback.

Figure 1
NCCN Guidelines Development Process

• Derivation of the preliminary pathway: A panel derives the preliminary guideline based on its interpretation of the data and discussion of relevant issues. The NCCN staff are present to provide input about the NCCN Guidelines methodology. The guidelines coordinators collect and review the meeting notes to construct the first draft of an algorithm that conforms to the standard NCCN format. The preliminary guideline, designated Version 1.0, is reviewed with the panel chair to ensure that it reflects the panel's deliberations.
• Institutional review: The preliminary version is then distributed to each NCCN Panel Member who in turn circulates the document internally for institutional review. This local review involves sending the algorithm to experts in all relevant disciplines. Institutional reviews are performed according to the processes and organization particular to each center. In many instances, multidisciplinary teams evaluate the guidelines to ensure comprehensive review. The NCCN Panel Member is responsible for collecting institutional comments and returning them to the guidelines coordinators for collation.
• Collation: The NCCN staff collects all reviews and collates them into a standardized form so that all comments relevant to a particular recommendation can be reviewed together.
• Guideline modification: The NCCN Guideline Panel reconvenes to discuss the comments. Discussion is usually accomplished via a conference call, but if major substantive issues are raised, another meeting is arranged. If a meeting is required, those individuals who expressed major dissenting opinions are invited to join the NCCN Guideline Panel discussions. At this meeting, a revised version of the NCCN Guideline is derived. Minor revisions are the prerogative of the NCCN Guideline Panel and are not reviewed at the institutional level until the next annual review. If the changes are uniformly accepted by the NCCN Guideline Panel members, this version then becomes the final version for that year. If the revisions require major substantive changes, especially if they are not uniformly accepted, the process is repeated with another round of institutional reviews, collation, and modification. On occasion, secondary revisions may be sought from a subset of the NCCN Guideline Panel (e.g., radiotherapists). Once the NCCN Guideline Panel has made its final modifications, the NCCN Guideline becomes the official NCCN version for that year.

Continuous Review

A major feature of the NCCN Guidelines program is that each NCCN Guideline is reviewed continuously to incorporate change as new evidence or innovative therapies become available. If clinical practice guidelines are to remain relevant, they must be up-to-date (Browman et al, 1995). New studies may have a major impact on practice recommendations, and if mechanisms do not exist to rapidly incorporate this information into the pathways, the entire program will lack relevance.

The process of continuous review mirrors the developmental process in a slightly abbreviated form:

• Institutional review: Each NCCN Panel Member is sent copies of the current year's guideline for distribution to institutional experts. Comments are received and forwarded to the NCCN staff offices.
• Collation: Comments are collated by category as with the developmental process.
• Panel review: The scheduled annual reviews consist of a full-day meeting every 3 years to ensure comprehensive review of the entire algorithm. The annual reviews for the intervening 2 years may consist of a conference call or a full-day meeting, depending on the NCCN Guideline Panel chair's assessment of the amount of new material to be covered. At the annual review, the NCCN Guideline Panel discusses any new data that might lead to modification of the pathway recommendations. Comments or information received during the year from institutional faculty members, extramural oncologists, patient advocacy groups, pharmaceutical companies, and other organizations are also discussed. The entire NCCN Guideline is re-reviewed to determine whether additional refinement or explanation is necessary. Any substantive changes may necessitate a second round of institutional reviews. Once the revisions have been uniformly accepted, the new version becomes the official NCCN Guideline for that year.

Guidelines Methodology

The basic format of an NCCN Guideline is an algorithm. The purpose of an algorithm is to define recommendations across the entire spectrum of the oncologic disease; therefore, the construction of a pathway is an orderly presentation of the clinical decisions encountered in managing the tumor and the specific recommendations for those decisions (Margolis, 1983). Figure 2 depicts the NCCN algorithm models.

Figure 2:
NCCN Algorithm Models

For treatment guidelines, the pathways build on a template progressing from diagnosis and staging to primary intervention, adjuvant treatment, surveillance, and management of recurrent or disseminated disease. This template is then modified to fit the management requirements for each disease. Thus, a midtreatment restaging is mandated in some lymphomas, and primary treatment in acute leukemia is divided into induction, consolidation, and maintenance recommendations.

Given the heterogeneity of clinical scenarios, the template for the supportive care guidelines is not as well-defined as the standard treatment pathway, but most have utilized a progressive schema leading from screening to risk/severity assessment, to triage for specialized care, to specialized evaluation, to specific interventions, and finally to reevaluation and maintenance.

Two steps are required to construct an algorithm: (1) determine the logic of the pathway and (2) fill in the appropriate guideline elements (recommendations). Pathway logic determines the branching pathways that incorporate sequential management decisions.

• Developing pathway logic: The development of a pathway's logic is often the most difficult task. The NCCN Guideline Panel must define the relevant clinical groupings along which to expand the pathway. For example, the breast cancer algorithm immediately breaks into paths based on stage, because a staging assessment is made at the initial visit and further management is based on this evaluation. In contrast, the pancreas algorithm is based on anatomical location of the lesion and an assessment of whether or not the tumor is resectable. Thus, the pathway mirrors the way in which the clinician approaches management of the specific tumor. Within each tumor, the group must also decide how many clinical presentations must be accounted for. As a rule of thumb, if a clinical scenario represents less than 5% of presentations for a particular tumor, the NCCN Guideline Panel does not attempt to derive a unique pathway for that scenario unless there are very significant clinical implications (e.g., resection of hepatic metastases in colon cancer even though it occurs in only 3% of cases).
• Identifying Appropriate Guideline Elements: After the basic logic is developed, the NCCN Guideline Panel identifies the appropriate guideline elements. These are the recommendations for specific tests and treatments. The process is best characterized as consensus-driven with explicit review of the scientific evidence by multidisciplinary panels of expert physicians from NCCN Member Institutions. When possible, recommendations are based on high-level (high-powered, randomized trials or meta-analysis) evidence. Frequently, panels must rely on lesser degrees of evidence (i.e., phase II studies with surrogate end points, such as response). In certain areas, the data may be extremely scanty, and the panel must rely on its own clinical experience and judgment. In these instances, while recognizing that future data may lead to significant modifications of the recommendations, the panel offers its opinions about reasonable approaches, based on analogy to other scenarios, anecdotal experience, or even theoretical considerations. The NCCN Guideline Panels are entirely cognizant of the limitations of this approach, and a frequent major discussion point is the clinical trials that are necessary to elucidate these ill-defined areas.

The focus of the NCCN Guidelines program is the development of recommendations that apply across NCCN Member Institutions. Although NCCN Guidelines may be implemented nationwide, there may be instances when the guidelines reflect the expertise and capabilities of larger, comprehensive centers. The individual practitioner must determine whether referral is advisable.

In many disease entities, the data do not permit a single definitive recommendation. In these instances, therefore, a broad list of options must be presented. This is in contrast to pure evidence-based guidelines, which do not make recommendations in the absence of conclusive evidence. Each NCCN Guideline Panel recommends reasonable approaches to disease management based on their experience and knowledge of the biology of the disease. Because absence of treatment inevitably leads to poor results, a list of reasonable choices may provide support for decision-making in difficult clinical situations.

When attempting to define the benefits of therapy, panels are asked to assess the outcomes of a diagnostic test or therapeutic intervention. Whenever possible, evidence for improvement in one of the four major outcomes (ASCO, 1996) in oncology: (1) survival, (2) quality of life, (3) patient satisfaction, and (4) cost-effectiveness are used as primary determinants for the recommendations.

Safeguards for Eliminating Biases: The difficulty inherent in guidelines that are based in part upon consensus is that the biases of the experts may shape the guideline and either exclude reasonable choices or incorporate personal favorites as preferred options (Fink et al, 1984). To minimize this possibility, the NCCN Guidelines process includes the two safeguards listed above: panels reflecting all schools of thought for a particular tumor, and the process of iteration and feedback.

A third important vehicle to address bias is the presentation of new NCCN Guidelines at the annual meeting, where meeting attendees are invited to provide both oral and written comments on the NCCN Guidelines. This broad-based feedback has been of inestimable value in testing the fairness of the recommendations.

Conflict of Interest: Relationships between developers of NCCN Guidelines and the pharmaceutical industry have been viewed as a potential source of bias in guideline development (Choudhry et al., 2002). NCCN has recognized that as faculty members of large academic cancer centers, most NCCN Guideline Panel Members are involved in clinical research and many have active relationships with sponsors developing new therapies. At the beginning of each NCCN Guideline Panel meeting, individual NCCN Guideline Panel Members disclose any potential conflicts of interest orally and in writing. Research support, service on an advisory committee or speakers bureau participation, and patents or equity holdings are examples of relationships that require disclosure. An NCCN Guideline Panel Chair may decide to exclude an NCCN Panel Member from discussions of areas in which a major conflict of interest exists. Aggregate panel disclosures are published with each NCCN Guideline, both online and in JNCCN.

While any pharmaceutical company or other interested party is permitted to submit scientific data and comments for NCCN Guideline Panel consideration, they have no control over the NCCN Guideline Panel's recommendations. Industry representatives do not participate in NCCN Guideline Panel deliberations.

Although, as an organization, NCCN accepts general sponsorship grants from multiple pharmaceutical and biotechnology companies and grants for the support of specific projects that are part of NCCN's ongoing activities, no pharmaceutical industry support is accepted for any component of NCCN Guideline development. Pharmaceutical industry support is accepted in the form of educational grants for presentation of guidelines updates at NCCN meetings and for dissemination of the guidelines (Winn, 2003).

NCCN Categories of Evidence and Consensus:

When studying the recommendations on a pathway, the NCCN Guideline user should be given information about how the recommendation was derived. In order to provide this critical information, the NCCN Guidelines Steering Committee has devised a set of Categories of Evidence and Consensus. These annotations contain two dimensions: the strength of the evidence behind the recommendation and the degree of consensus about its inclusion.

Category of Evidence and Consensus	Quality of Evidence	Level of Consensus
1	High	Uniform
2A	Lower	Uniform
2B	Lower	Non-uniform
3	Any	Major disagreement

• Category 1: The recommendation is based on high-level evidence (i.e., high-powered randomized clinical trials or meta-analyses), and the NCCN Guideline Panel has reached uniform consensus that the recommendation is indicated. In this context, uniform means near unanimous positive support with some possible neutral positions.
• Category 2A: The recommendation is based on lower level evidence, but despite the absence of higher level studies, there is uniform consensus that the recommendation is appropriate. Lower level evidence is interpreted broadly, and runs the gamut from phase II to large cohort studies (Cook, 1992) to case series to individual practitioner experience. Importantly, in many instances, the retrospective studies are derived from clinical experience of treating large numbers of patients at a member institution, so NCCN Guideline Panel Members have first-hand knowledge of the data. Inevitably, some recommendations must address clinical situations for which limited or no data exist. In these instances the congruence of experience-based judgments provides an informed if not confirmed direction for optimizing patient care. These recommendations carry the implicit recognition that they may be superseded as higher level evidence becomes available or as outcomes-based information becomes more prevalent (Baillard, 1993).
• Category 2B: The recommendation is based on lower level evidence, and there is nonuniform consensus that the recommendation should be made. In these instances, because the evidence is not conclusive, institutions take different approaches to the management of a particular clinical scenario. This nonuniform consensus does not represent a major disagreement, rather it recognizes that given imperfect information, institutions may adopt different approaches. A Category 2B designation should signal to the user that more than one approach can be inferred from the existing data.
• Category 3: Including the recommendation has engendered a major disagreement among the NCCN Guideline Panel Members. The level of evidence is not pertinent in this category, because experts can disagree about the significance of high level trials (McNeill, 2001). Several circumstances can cause major disagreements. For example, if substantial data exist about two interventions but they have never been directly compared in a randomized trial, adherents to one set of data may not accept the interpretation of the other side's results. Another situation resulting in a Category 3 designation is when experts disagree about how trial data can be generalized. An example of this is the recommendation for internal mammary node radiation in postmastectomy radiation therapy. One side believed that because the randomized studies included this modality (Woolf, 1997), it must be included in the recommendation. The other side believed, based on the documented additional morbidity and the role of internal mammary radiation therapy in other studies, that this was not necessary. A Category 3 designation alerts users to a major interpretation issue in the data and directs them to the manuscript for an explanation of the controversy.

The methodology of NCCN Guidelines presupposes a dynamic process. The NCCN Guidelines may be useful only if the pathways are continuously updated to incorporate new data, and emerging evidence drives the degree of specificity of the recommendations.

Summary

The NCCN has embarked on an ambitious project to develop comprehensive diagnostic and treatment algorithms for the broad spectrum of oncologic diseases and support modalities. The process relies on the expertise of the NCCN Member Institutions and follows a formal procedure for guidelines development to ensure that the recommendations offer the cancer patient the best chance for a successful outcome.

References

American Society of Clinical Oncology: Outcomes of cancer treatment for technology assessment and cancer treatment guidelines. J Clin Oncol 14:671-679, 1996.

Baillard DJ, Duncan PW. Role of population-based epidemiologic surveillance in clinical practice guideline development. In: Clinical Practice Guideline Development: Methodology Perspectives. Rockville, MD: Agency for Health Care Policy and Research, Public Health Service; 1993:93-104.

Browman GP, Levine MN, Mohide A, et al: The practice guidelines development cycle: A conceptual tool for practice guidelines development and implementation. J Clin Oncol 13:502-512, 1995.

Choudhry NK, Stelfox HT, Detsky AS. Relationship between authors of clinical practice guidelines and the pharmaceutical industry. JAMA 2002;287:612-617

Cook DJ, Greengold NL, Ellrodt AG, Weingarten SR. The relationship between systematic reviews and practice guidelines. Ann Intern Med 1997;127:210-216.

Field MJ, Lohr KN (eds): Clinical Practice Guidelines: Direction for a New Program. Institute of Medicine, Committee on Clinical Practice Guidelines. Washington, DC. National Academy Press, 1990.

Fink A, Kosecoff J, Chassin M, et al: Consensus methods: Characteristics and guidelines for use. Am J Public Health 74:979-983, 1984.

Glaser EM: Using behavioral science strategies for defining state-of-the-art. J Appl Behav Sci 16:79-92, 1980.

Leape LL: Practice guidelines and standards: An overview. Qual Rev Bull 16:42-49, 1990.
Margolis CZ: Uses of clinical algorithms. JAMA 249:627-632, 1983.

McNeill BJ, Shattuck L. Hidden barriers to the improvement in the quality of care. N Engl J Med 2001;345:1612-1619.

National Comprehensive Cancer Network: Oncology Practice Guidelines. Oncology 10 (suppl 11):3-317, 1996.

National Comprehensive Cancer Network: Oncology Practice Guidelines, Volume 2. Oncology 11 (suppl 11):3-346, 1997.

National Comprehensive Cancer Network: Oncology Practice Guidelines. Oncology 12 (suppl 7A):3-271, 1998.

Park EP, Fink A, Brook RH, et al: Physician ratings of appropriate indications for six medical and surgical procedures. Am J Public Health 76:766-772, 1986.

Stiell IG, Greenberg GH, Wells GA, et al: Prospective validation of a decision rule for the use of radiography in acute knee injuries. JAMA 275:611-615, 1996.

Weeks JC: Outcomes assessment in the NCCN. Oncology 11 (suppl 11):137-140, 1997.

Winn RJ. NCCN Guideline development: Some lessons learned. Cancer Control 2001;8:11-14.

Winn RJ, McClure, JS The NCCN Clinical Practice Guidelines in Oncology: A Primer for Users. JNCCN 2003; 1:5-13.

Woolf SH: Practice guidelines, a new reality in medicine: II. Methods of developing guidelines. Arch Intern Med 152:946-952, 1992.

Woolf SH, Lawrence RS. Preserving scientific debate and patient choice: Lessons from the Consensus Panel on mammography screening. JAMA 1997;278:2105-2107.

Back to top