| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Editorial |
1 From the Department of Diagnostic Imaging, Providence Everett Medical Center, Everett, Wash (R.L.B.); and Department of Radiology, University of Michigan Medical Center, Ann Arbor (R.C.C.). Received December 10, 2001; revision requested December 11; revision received December 14; accepted December 14. Address correspondence to R.L.B., Radia Medical Imaging, 728 134th St SW, Everett, WA 98204 (e-mail: rbree@radiax.com).
Index terms: Economics, medical, 854.1261, 854.12989 • Uterus, biopsy, 854.1261 • Uterus, endometrium, 854.1261, 854.12989 • Uterus, US, 854.12989
Radiologists are relatively new in their involvement in evaluation of patients with postmenopausal bleeding (PMB). Most patients who present with this perplexing symptom have a benign cause, yet expenditures on costly examinations and indirect costs related to concern about endometrial cancer takes a substantial toll. Until recently, patients with PMB would undergo gynecologic testing, including endometrial biopsy, dilatation and curettage, or hysteroscopy to evaluate for the cause of bleeding. In the gynecologic literature, investigators in most studies (1) have pointed out the relative infrequency of cancer (10%) and the relative frequency of endometrial atrophy (40%–50%) as causes of PMB. More recently, radiologists and gynecologists performing transvaginal ultrasonography (US) and saline-infusion hysterosonography have discovered large numbers of polyps and leiomyomas in these patients, which suggests that the previous techniques were not able to accurately depict the exact cause of PMB (2–6).
In this issue of Radiology, Medverd and Dubinsky (7) model cost minimization by using US techniques versus endometrial sampling in patients with abnormal bleeding after menopause. In choosing to study this vexing clinical problem with decision modeling techniques, the authors have raised many questions addressed by an important consensus conference held in October 2000 and sponsored by the Society of Radiologists in Ultrasound, or SRU (8). At that conference, participants used consensus-building techniques to attempt to answer some important questions about PMB. Important consensus statements were:
1. Transvaginal US is the imaging procedure of choice to evaluate patients with PMB. Transabdominal US alone is not adequate to evaluate the endometrium.
2. Transvaginal US must be performed with meticulous technique to be considered adequate for evaluation.
3. By using the meta-analytic summary of Smith-Bindman et al (9), it was determined that with an endometrial thickness of 5 mm or less, the chance of a patient having endometrial cancer is about 1%. The conferees also noted that there is no acceptable upper limit of normal for endometrial thickness for any patient, and further evaluation is warranted when there are symptoms and the endometrium is more than 5 mm thick.
4. The conferees also agreed to standard terminology for two of the important US examinations, namely, transvaginal US and saline-infusion sonohysterography.
In their analysis, Medverd and Dubinsky (7) discuss the accuracy of the standard evaluative techniques for PMB. It is notable that efforts to ascertain the sensitivity and specificity of the competing technologies have been fraught with difficulties attributable to the inaccuracy of the reference standards with which the newer technologies have been compared. At the consensus conference, it was pointed out that endometrial biopsy has a high rate of inadequacy of sampling, necessitating performance of other examinations such as saline-infusion hysterosonography, hysteroscopy, or dilation and curettage. On the other hand, when separating endometrial cancer from nonmalignant causes of PMB, endometrial biopsy performs well (10). Because of a paucity of literature on the subject, the consensus conference was not able to develop consensus about the relative cost-effectiveness of US versus biopsy techniques. This analysis was left as a question to be placed on a research agenda. Medverd and Dubinsky (7) nicely point out the lower cost of US techniques with all scenarios. As they suggest, however, the analysis does not include consideration of any clinically relevant outcomes such as the number of cancers or polyps detected or the number of life-years saved, making it a cost minimization analysis only.
Cost minimization analysis, as the investigators have conducted, is used to compare the net cost of interventions (broadly defined to include diagnostic testing) that achieve the same net cost, assuming that the two interventions achieve the same outcome, for example, cancer detection or additional life expectancy. Cost-effectiveness analysis incorporates the costs and outcomes (effects) of an intervention and at least one alternative. The effects measured may be terminal outcomes, such as additional life expectancy (life-years saved) or intermediate outcomes, such as additional cancers detected or false-negative cancers avoided. The results are presented as a ratio of incremental costs to incremental effects; for example, the additional cost per additional cancer by using a given intervention that would not otherwise have been detected by using the other intervention. Cost-utility analysis incorporates patient preference into the analysis by using utility as the denominator, where utility represents the patient’s preference for avoiding a particular disease state. Utility is most commonly evaluated as quality-adjusted life-years (QALYs), where each year lived with a disease, such as endometrial cancer, is decrementally adjusted, as compared with a year lived without cancer. The cost-utility of an intervention is the additional cost of the intervention divided by the additional QALYs lived. Cost-benefit analysis estimates the net social benefit of an intervention as the incremental benefit of the intervention less the incremental cost, where benefits and costs are measured in dollars. To generate a cost-benefit analysis from a cost-utility analysis, the additional cost required to save a QALY will be subtracted from the monetary value of an additional QALY. The result reflects the additional resource expenditure (when the intervention costs more than a QALY) or the resource savings (when the intervention costs less).
Although the investigators are careful to state that their purpose was to perform a cost minimization analysis and therefore did not account for indirect costs, cost minimization analysis usually compares the net costs of interventions, where the net costs include indirect costs. There are, however, reasons to exclude indirect costs from the current analysis. The most compelling reason is the population itself. Incorporating indirect costs, such as lost days of work due to symptoms or being subjected to an additional diagnostic test, tends to undervalue older individuals, in particular, older women. If these women did not work, then there is no net loss of income (an indirect cost). The authors further state that the perspective of the analysis is a societal perspective, where Medicare reimbursements are used as a proxy for cost to society. However, the presentation of costs in the absence of outcome overlooks society’s willingness to expend more resources for interventions that yield a valued effect, as long as the magnitude of the effect is commensurate with the expenditure. In short, the question "How much bang for our buck?" cannot be answered if the "bang" is missing from the analysis.
Indeed, low-level cost-effectiveness analysis by using an intermediate outcome can be performed from the data presented by the authors. If clinically relevant outcome desired is the minimization of false-negative diagnoses, by using the data from the second and last columns from Table 3, an incremental cost-effectiveness ratio (ICER) can be calculated, where the ICER describes the additional cost per additional false-negative case avoided. Comparison of models 1 and 6, the most expensive and least expensive algorithms, respectively, yields the following ICER: Cost (model 6 - model 1)/outcome (model 6 - model 1) = ($449,807 - $437,497)/(2.2 - 2.4) = $81,698 per additional false-negative result averted.
The authors do not subdivide the number of missed cases into that of benign and malignant causes. If we assume that only 31% are malignant (authors’ base-case scenario), then a majority of missed abnormalities are benign. The additional cost seems high, considering that the long-term effect of diagnosis of a benign abnormality is still unknown. However, if we compare models 6 and 4, the ICER decreases to $1.23 per additional false-negative finding averted, a potentially reasonable cost. Society may be willing to forgo the 3% total cost reduction in using model 6 rather than model 4 in exchange for finding more cancers. In short, outcomes matter. Furthermore, these outcomes matter differently to different individuals, namely, the patient, physician, and third-party payer.
A cost minimization analysis is best undertaken from the perspective of a third-party payer, such as a health maintenance organization, or HMO, that may be interested in only total cost reduction (11). As such, HMO-specific costs may be used instead of Medicare reimbursements. The central weakness of performing cost minimization analysis, particularly from a societal perspective, is that the authors have not presented the reader with data showing that the net effect across the algorithms examined is comparable. Indeed, in the one clinical outcome presented in the investigation, the predicted number of patients remaining undiagnosed differs. Although the percentage difference is small, the total number of cases is potentially substantial if applied toward the population of women with abnormal uterine bleeding. Furthermore, the consequences of missing a cancer are potentially much more severe than those of missing a polyp. Data have not been presented that discuss the ability of each algorithm (not each test) to detect additional cancers (or other lesions), compared with the other algorithms and the additional cost of that incremental ability.
Even though the authors do not present the data described previously, the strength of the investigators’ analysis is the consideration of malignant and nonmalignant abnormalities and the evaluation of a series of algorithms that more closely simulate clinical practice in their evaluation of cost. The current study improves on that of Feldman et al (12), who compared a single test, endometrial biopsy, against the base case of no testing, and the sole clinical diagnosis relevant to the analysis is endometrial cancer. The current analysis also improves on the study of Weber et al (13), who tested two competing diagnostic strategies again, with the end point of cancer diagnosis. In another decision analysis, Carlos et al (14) considered a limited number of diagnostic algorithms and accounted for malignant and nonmalignant causes by framing the economic analysis in the context of the intermediate outcome, with the ICER explicitly stated.
To help put the current analysis in perspective, it is useful to refer to one of several technology assessment models similar to those of Fryback and Thornbury (15) and Littenberg (16). By using Littenberg’s hierarchy for technology assessment (16), investigations that support technology diffusion can be classified into one of five levels. Level I assesses biologic plausibility and compares the technology’s proposed mode of action with current biologic information and theory. Level II, technical feasibility, determines whether the technology can be delivered to the target population. Level III, intermediate outcomes, assesses whether the technology has a short-term effect on the biologic or physiologic process that is diseased. Level IV, patient outcomes, investigates the overall medical, psychologic, and financial effects of the technology on the patient, including unintended side effects and long-term morbidity and mortality. Level V, societal outcomes, measures the cost of the technology to society in terms of resource use, ethical issues, and social and political hazards.
In Medverd and Dubinsky’s discussion (7), it is important to note that they were trying to make a diagnosis in every patient with abnormal uterine bleeding. The SRU consensus conference addressed that point and was unable to ascertain the value of finding nonmalignant causes of PMB. Additionally, Bree et al (2) discovered that with saline-infusion hysterosonography, almost one-third of patients with an abnormality had endometrial thicknesses of 5 mm or less, so it is even more important to evaluate the outcomes of finding benign lesions. It is well known that many patients who begin hormone replacement therapy, or HRT, will not be taking that medication after 12 months. A substantial number of patients stop HRT because of bleeding (17). Since anatomic reasons for bleeding are common, it makes some sense that all patients with PMB should undergo either saline-infusion hysterosonography or hysteroscopy. The panelists at the consensus conference were unable to find enough evidence that supports this approach. Instead, they looked at that issue as an important research question. It is not so much the short-term effect of treating the offending lesion but the intermediate term outcome of patient anxiety about the bleeding, and the longer-term outcome that cessation of HRT might have on the cardiovascular, musculoskeletal, and central nervous system health of the patient.
Although the consensus conference was unable to recommend the optimal diagnostic strategy in the evaluation of PMB, implicit in the debate is the increasing pressure to provide as much "value" as possible while maintaining a reasonable cost of evaluation (18). It is important to recognize that information derived from testing has value, even in the absence of an underlying abnormality. Prognostic information may be valuable, independent of its usefulness in determining management strategies (18). This information has been termed "nondecisional" products of testing. For the physician, even if results do not alter management, testing conveys a substantial degree of diagnostic certainty and perhaps increased confidence in the current clinical course. For women with PMB, diagnostic certainty may provide relief of anxiety. Even individuals without a specific clinical complaint requested checkups to express concern about their health and seek reassurance that they had "no disease" (19). Similarly, women with PMB may have major health concerns as a result of their symptoms and may benefit from reassurance. Women with PMB may value other "products" of diagnostic testing beyond determining the presence or absence of endometrial cancer, such as knowing the cause of her symptoms and understanding PMB’s natural history, treatment, and prognosis. In previous work in individuals with multiple sclerosis (20), most patients reported that they were better off having received diagnostic information, although the degree of anxiety reduction depended on the "certainty" of diagnosis. Those in whom a definite diagnosis of multiple sclerosis could be made or excluded were reassured more than those in whom the examination was not definitive. Similarly, contingent on the type of abnormality diagnosed, women with PMB may have positive or negative experiences that alter lifestyle or health behaviors. These women may also value the participatory decision-making experience of choosing the sequence of testing, independent of the results. Incorporation of these patient preferences will elevate the level of technology assessment to level 4 of Littenberg’s hierarchy (16).
The consensus conference was challenged with the task of developing a clinical scenario for the evaluation of women with PMB given the many considerations discussed previously. As a result, they prepared two algorithms that are illustrated in Figures 1 and 2. These algorithms are useful if the patient evaluation begins with endometrial biopsy or transvaginal US. Since there was no consensus on the best approach in a given patient, these scenarios can be of help to the clinician or radiologist who has to evaluate patients with PMB.
|
|
FOOTNOTES
See also the article by Medverd and Dubinsky in this issue.
REFERENCES
Related Article
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
