Rigorous new approach to constructing a gold standard for validating new diagnostic criteria, as exemplified by the eosinophilia-myalgia syndrome.

BACKGROUND: Constructing diagnostic criteria, a common problem in clinical medicine, is particularly difficult for diseases that lack a pathognomonic "gold standard." To develop an improved strategy for constructing such criteria, we used the eosinophilia-myalgia syndrome as an example. The goal, for research classifications, was to construct validated clinically sensible criteria and to develop improved methods that can be used for other disorders. METHODS: Using a "pattern-based" approach with data from several separate sources, a committee of investigators first prepared and informally tested criteria for the diagnosis of eosinophilia-myalgia syndrome. A gold standard challenge set of reports of cases and noncases was independently generated and separately validated by an external panel of clinical experts. The criteria were then tested using the gold standard set, and interobserver variability and diagnostic accuracy were determined. RESULTS: Interobserver variability showed the following mean proportionate agreements: 98.7% for the presence of specific criteria elements, 99% to 100% for diagnosis, and 97% to 98% for diagnostic pattern. kappa Values were correspondingly high. Diagnostic accuracy showed sensitivity at 88%, specificity at 97%, and overall accuracy at 92%. CONCLUSIONS: The proposed criteria are accurate and reproducible, and can be used in future clinical investigations of the eosinophilia-myalgia syndrome. The new strategy and methods developed for this challenge can be valuable for solving analogous problems in constructing criteria for other clinical disorders.

The Blame-X syndrome: problems and lessons in nosology, spectrum, and etiology.

Symptoms of a functional somatic syndrome have been noted in individual persons and groups for more than a century. Often associated with war, the syndrome has received diverse names and many proposed but unproved etiologies, including exposure to trauma, stress, chronic infection, psychosomatic, chemical, or environmental causes. In recent years, when attributed to agent X, the syndrome could be called the Blame-X syndrome. The clinical, legal, and other problems associated with the syndrome are a reflection of nosologic difficulties in identifying and choosing titles for apparently "new" ailments. The difficulties arise from the complex overlap of symptoms, diseases, and laboratory abnormalities found with modern technology, and from the frequent abandonment of pathophysiologic demands for appropriate correlation of symptoms and objective abnormalities. An important principle in naming apparently new ailments is to avoid etiologic titles until the etiologic agent has been suitably demonstrated. A premature causal name can impair a patient's recovery from the syndrome, and impede research that might find the true cause.

Presenting manifestations, cigarette smoking, and detection bias in age at diagnosis of lung cancer.

PURPOSE: To examine the possible role of detection bias in the association between amount of cigarette smoking and age at diagnosis of lung cancer. The bias can occur because primary lung cancer can often escape detection during life and will be found (if at all) as a "necropsy surprise" unless a diagnostic workup is provoked by such presenting manifestations as hemoptysis and a localized chest lesion. The necropsy surprises will be reduced and the reported rates of pre-mortem incidence will be raised if a cigarette smoking history also acts as a diagnostic incentive. METHODS: This possibility was examined in a case series of 1266 patients whose primary lung cancer had been carefully classified according to diverse features at the time of presentation. For the total case group and for pertinent clinical, anatomic, and demographic subgroups, we then examined the trends for age at diagnosis in relation to amount of cigarette smoking. RESULTS: The overall age at diagnosis (median = 63 years; mean = 61.2) remained essentially similar in five ordinal groups of Tumor, Nodes, Metastases (TNM) and four of five Clinical Severity stages, but had an inverse monotonic gradient in six ordinal groups of customary cigarette smoking [from none to >2 packs per day (ppd)]. Because an earlier age of discovery can be explained by either etiologic or detection-bias roles for heavier smoking, its impact was checked in subgroups with and without diagnostically provocative manifestations. In localized lesions, the smoking-age gradient vanished if suspicious "indicator" symptoms were present, but persisted if they were absent. Regardless of symptoms, the age gradient was strengthened in non-localized cancer lesions where smoking might particularly point to a primary diagnostic source in the lung. CONCLUSIONS: Detection bias may play a distinctive, although often overlooked, role in the work-up decisions that precede and lead to a diagnosis of lung cancer.

Relationship between nonphenacetin combined analgesics and nephropathy: a review. Ad Hoc Committee of the International Study Group on Analgesics and Nephropathy.

BACKGROUND: The debate on the association between nonphenacetin-containing combined analgesics and renal disease has lasted for several years. METHOD: A peer review committee of scientists, selected jointly by the regulatory authorities of Germany, Switzerland, and Austria and the pharmaceutical industry was asked to critically review data on the relationship between nonphenacetin combined analgesics and nephropathy. RESULTS: The committee regarded epidemiologic evidence on nonphenacetin combined analgesics as inconclusive because of sparse information and substantial methodological problems. The committee also noted that a diagnosis of analgesic-associated nephropathy (AAN) in clinical practice usually depends on information about exposure before or in the early stages of the disease and is seldom accompanied by specific histologic evidence. The morphologic finding of papillary calcification can arise from other conditions and is not specific for AAN. For these reasons, the identification criteria for AAN should be reappraised with scientific methods to validate the diagnostic procedure. In the limited amount of experimental pharmacological data in humans and animals, the committee found no convincing evidence to confirm or refute the hypothesis that nonphenacetin combined analgesics are more nephrotoxic than single formulations. For caffeine taken with combined analgesics, the currently available information is not sufficient to postulate a harmful toxicological effect. CONCLUSION: The committee's two main conclusions were that sufficient evidence is absent to associate nonphenacetin combined analgesics with nephropathy and that new studies should be done to provide appropriate data for resolving the question.

Do caffeine-containing analgesics promote dependence? A review and evaluation.

OBJECTIVE: Debates about the suspected association between kidney disease and use of analgesics have led to concern about whether caffeine could stimulate an undesirable overuse of phenacetin-free combined analgesics. A committee was asked to critically review the pertinent literature and to suggest guides for clinical practice and for consideration of international regulatory authorities. PARTICIPANTS: A group of international scientists, jointly selected by the regulatory authorities of Germany, Switzerland, and Austria and the pharmaceutical industry. EVIDENCE: All invited experts evaluated relevant literature and reports and added further information and comments. CONCLUSIONS: Caffeine has a synergistic effectiveness with analgesics. Although caffeine has a dependence potential, the potential is low. Experimental data regarding dependence potential for caffeine alone may not correspond to the conditions in patients with pain. Withdrawal is not likely to cause stimulation or sustainment of analgesic intake. For drug-induced headache, no single or combined analgesic was consistently identified as causative, and no evidence exists for a special role of caffeine. Strong dependence behavior was observed only in patients using phenacetin-containing preparations, coformulated with antipyretics/analgesics and caffeine. This finding may have led to the impression that caffeine stimulates overuse of analgesics. SUMMARY: Although more experimental and long-term data would be desirable to show possible mechanisms of dependence and to offer unequivocal proof of safety, the committee concluded that the available evidence does not support the claim that analgesics coformulated with caffeine, in the absence of phenacetin, stimulate or sustain overuse.

Confirmations and surprises in the association of tobacco use with sinusitis.

OBJECTIVE: To generate estimates of sinusitis prevalence for adults in association with the use of tobacco or passive smoke exposure. DESIGN: Analysis of data from the Third National Health and Nutrition Examination Survey, 1988-1994. SETTING: Sample of the noninstitutionalized civilian population of the United States. PARTICIPANTS: A total of 20,050 adults aged 17 years or older. MAIN OUTCOME MEASURE: Presence of self-reported sinusitis or sinus problems. RESULTS: In the United States, 66 million adults, constituting 35% of the adult population, reported having sinusitis or sinus problems at least once during the previous 12 months. Female sex, non-Hispanic white or non-Hispanic black race, higher income levels, and progressively higher educational levels were associated with increased prevalence of sinusitis. The prevalence of both acute and recurrent or chronic sinusitis increased with direct cigarette and other tobacco use but did not rise with passive exposure to cigarette smoke. CONCLUSIONS: Consistent with data for other respiratory ailments, the direct use of tobacco confers a small increased risk of developing sinusitis in the adult population, but contrary to expectation, passive smoke does not. The demographic variables of sex, race/ethnicity, and educational level demonstrated unexpectedly strong associations with the prevalence of sinusitis and should be analyzed and controlled for in future studies of sinusitis.

"Benign" tumors and "early detection" in mammography-screened patients of a natural cohort with breast cancer.

BACKGROUND: Although the cure of breast cancer by "early detection" and prompt treatment rests on the belief that all breast cancers grow at the same rate, many cancers have been shown to grow rapidly and others slowly. In particular, mammography screening may often detect the slow-growing, nonaggressive tumors that might not be found until much later, if at all. METHODS: We reviewed the medical records of a natural cohort of 233 patients. The cohort comprised all women who received their first antineoplastic treatment for breast cancer at Yale-New Haven Hospital during the period from January 1 through December 31, 1988, and had a median follow-up thereafter of 82.4 months. RESULTS: The mammography screen-detected group (MSDG) contained 97 (42%) of the 233 breast cancers. The rates of subsequent freedom from cancer deaths or recurrences were 95% (92 patients) in the MSDG and 79% (107 patients) in all other patients (log-rank 2P<.001). This superiority occurred partly because 90 (93%) of the MSDG were in the good prognosis TNM stages 0, I, and IIA, compared with 92 (68%) of the non-MSDG (chi2 2P = .001). Of the 31 patients with stage 0 (carcinoma in situ), all of whom had disease-free survival, 24 (77%) were found by mammography screening. Even within similar TNM stages, however, the MSDG had distinctly better disease-free survival results than the non-MSDG. For patients in TNM stages I and IIA, the "failure events" had respective rates of 2% and 13% (log-rank 2P = .02). CONCLUSIONS: The results suggest that many of the breast cancers found by mammography screening have excellent prognosis not just because of early detection, but also because many of the cancers are relatively benign, requiring minimal therapy.

Claims of equivalence in medical research: are they supported by the evidence?

BACKGROUND: Most clinical studies are done to show comparative superiority, but many reports now claim equivalence between the investigated entities. These assertions may not always be supported by the methods used and the results obtained. PURPOSE: To assess the justification and support for claims of clinical or therapeutic equivalence in medical journals. DATA SOURCES: A search of MEDLINE for articles published from 1992 through 1996. STUDY SELECTION: From 1209 citations that contained the word equivalence in the title or abstract or contained the Medical Subject Heading therapeutic equivalency, we excluded 1121 studies reporting nonoriginal research, purely laboratory or other nonhuman research, and studies in which equivalence was not the main claim. The remaining 88 eligible papers were evaluated for five methodologic attributes. DATA SYNTHESIS: Only 45 (51%) of the 88 reports were specifically aimed at studying equivalence; the others either tried to show superiority or did not state a research aim. The quantitative distinctions regarded as "equivalent" ranged from 0% to 21% for direct increments and from 0% to 76% for proportionate differences. An equivalence boundary was set and confirmed with an appropriate statistical test in only 23% of reports. In 67% of reports, equivalence was declared after a failed test for comparative superiority, and in 10%, the claim of equivalence was not statistically evaluated. The sample size needed to confirm results had been calculated in advance for only 33% of reports. Sample size was 20 patients per group or fewer in 25% of reports. CONCLUSIONS: Many studies of clinical equivalence do not set boundaries for equivalence. Claims of "difference" or "similarity" are often made not by thoughtful examination of the data but by tests of statistical significance that are often misapplied or accompanied by inadequate sample sizes. These methodologic flaws can lead to false claims, inconsistencies, and harm to patients.

A rating system for prompt clinical diagnosis of ischemic stroke.

BACKGROUND: When a CT scan is not available, an early accurate clinical diagnosis of ischemic stroke is essential to initiate prompt therapy. Our objective was to construct a clinical index that is easy to use when stroke patients are first evaluated at the hospital, to identify those who probably are experiencing an acute ischemic episode. The study was conducted at a university-affiliated medical referral center and two community general hospitals in Mexico. METHODS: Clinical records were reviewed for 801 patients with sudden onset of a focal or global neurologic dysfunction, presumably of vascular origin lasting more than 24 h. Eligibility criteria for this study were admission to the hospital within the first 24 h after symptomatic onset, CT scan diagnosis between 24 and 72 h, and age >45 years. Ischemic stroke included cases of arterial brain infarction, while nonischemic stroke included subarachnoid or intraparenchymatous hemorrhage, mass lesion, venous infarction, and in cases without a CT scan evidence that could explain the clinical manifestations. Data excerpted for analysis were age, sex, history of diabetes mellitus or previous stroke/transient ischemic attack (TIA), time of onset of symptoms, presence of headache, vomiting, neck stiffness, hemiplegia, leukocytosis or atrial fibrillation, diastolic blood pressure, and Glasgow coma scale (GCS) rating. Two multivariable analyses were used: 1) step-wise multiple logistic regression (SMLR), and 2) conjunctive consolidation (CC). RESULTS: After appropriate exclusions, the study proceeded with 83 ischemic and 42 nonischemic stroke patients. With SMLR, six variables were selected as predictive for ischemic stroke, including neck stiffness, diastolic blood pressure, previous history of stroke/TIA, hemiplegia, GCS, and atrial fibrillation. An appropriate sum of weighted ratings had a positive predictive value (PPV) of 100% for ischemic stroke. With consolidated categories, the PPV was 97% when patients had the following: no neck stiffness; no atrial fibrillation but history of stroke/TIA and GCS > or =12, or no neck stiffness but atrial fibrillation. CONCLUSIONS: Among patients with acute stroke, clinical data can be used to identify a group with a high probability of ischemic stroke. There are slightly different results between both methods; while SMLR includes the four variables selected by CC, the latter included neither diastolic blood pressure nor hemiplegia/hemiparesia. However, CC results seem easier to understand and interpret than with SMLR.

Basic biomedical science and the destruction of the pathophysiologic bridge from bench to bedside.

Although formerly a prime intellectual focus of medical practice, research, and education, the pathophysiology of organs and organ systems has become increasingly deemphasized. The field of clinical investigation is thereby left without a sturdy bridge to connect epidemiologic studies of patients with cellular and molecular studies. The decline of pathophysiology has led to major defects in diagnostic reasoning and clinico-pathophysiologic correlations, to isolated accomplishments in molecular research, and to the neglect of many prominent scientific questions that can be asked and answered only at the level of organs and organ systems. An important intellectual source of the problem is the ideologic belief that scientific importance is inversely proportional to the size of the investigated entities. With this belief, the title of "basic" is excluded from fundamental scientific questions in any research that does not occur at the level of cells and molecules. Although recent changes in National Institutes of Health policy may augment the decreasing number of physician-investigators, the more serious intellectual problems of constrained scientific creativity will continue until the current ideology is revised.

Indexes of contrast and quantitative significance for comparisons of two groups.

Although boundaries for 'large' and 'small' differences are needed to plan research and interpret results, the diverse indexes of descriptive contrast for the central indexes, A and B, of two groups, have not received intensive attention. For two means, the increment of mid R:A-Bmid R: reflects the slope of a line showing the 'effect', but is altered by different units of measurement. Division of mid R:A-Bmid R: by the common standard deviation produces the standardized increment (SI), which is sometimes called the 'effect size'. Despite many advant ages, it does not contrast the relative magnitudes of A and B. For the latter contrast, the relative change or proportionate increment (mid R:A-Bmid R:/B) is particularly easy to understand, and the relative translocation (mid R:A-Bmid R:/[A+B]) produces a bounded range from -1 to +1. Nevertheless, all indexes of relative magnitude ablate the scales of measurement, thereby increasing difficulty in interpretation. Although seldom applied, proportionate reduction in overall system variance can be highly useful. Its square root leads to eta, the analogue of a correlation coefficient, which corresponds to a standardized slope for the direct increment. The values of eta usually approximate (SI)/2. Although arbitrary levels have been proposed for 'quantitative significance' of the SI, the proportionate reduction in overall system variance is often regarded as ineffectual unless >/=10 per cent. With this belief, minimum boundaries for quantitative significance can often be set at eta>/=0.3 and SI>/=0.6. In indexes of relative magnitude for two proportions (or rates), p(A) and p(B), confusion is produced if q(A) and q(B) are alternatively chosen for the denominators. The odds ratio, (p(A) q(B)/p(B) q(A)), avoids these choices, but is often difficult to interpret. For easy understanding and communication, the preferred index is NNE, the number needed to produce one excess effect, calculated as the inverse of the direct increment, that is, 1/mid R:p(A)-p(B)mid R: The standardized increment, mid R:p(A)-p(B)mid R:/ radical(PQ), (where P is the average of p(A) and p(B) and Q=1-P) could offer a single index applicable to both dimensional and binary data, but when P becomes quite small, that is, <0.01, radical(PQ) requires special calculations and also approaches the value of radicalP. Boundaries of 'quantitative significance' are particularly difficult to establish for comparisons of two rates, because of additional consequences in populational extrapolations and clinical implications. Nevertheless, the principles of quantitative significance can aid the ad hoc construction of boundaries that must be set for medical importance when sample sizes are calculated and when results are interpreted for studies of either efficacy or equivalence.

The impact of ethnicity, family income, and parental education on children's health and use of health services.

OBJECTIVES: This study characterized ethnic disparities for children in demographics, health status, and use of services; explored whether ethnic subgroups (Puerto Rican, Cuban, and Mexican) have additional distinctive differences; and determined whether disparities are explained by differences in family income and parental education. METHODS: Bivariate and multivariate analyses of data on 99,268 children from the 1989-91 National Health Interview Surveys were conducted. RESULTS: Native American, Black, and Hispanic children are poorest (35%, 41% below poverty level vs 10% of Whites), least healthy (66%-74% in excellent or very good health vs 85% of Whites), and have the least well educated parents. Compared with Whites, non-White children average fewer doctor visits and are more likely to have excessive intervals between visits. Hispanic subgroup differences in demographics, health, and use of services equal or surpass differences among major ethnic groups. In multivariate analyses, almost all ethnic group disparities persisted after adjustment for family income, parental education, and other relevant covariates. CONCLUSIONS: Major ethnic groups and subgroups of children differ strikingly in demographics, health, and use of services; subgroup differences are easily overlooked; and most disparities persist even after adjustment for family income and parental education.

Clinical epidemiological quality in molecular genetic research: the need for methodological standards.

CONTEXT: A genetic basis has been identified for many medical conditions and some molecular tests have been commercialized. However, little attention has been given to the quality of clinical epidemiology in molecular studies. OBJECTIVE: To examine the clinical epidemiological quality of recent publications on molecular genetic analysis. DESIGN: Cross-sectional study of original research articles published in 1995, identified by manually searching 4 general clinical journals. Of 83 articles identified, 40 were selected for analysis; these 40 discussed molecular genetic techniques, studied 10 or more patients, and had inferential conclusions. MAIN OUTCOME MEASURE: Compliance of the selected articles with 7 methodological standards for clinical epidemiological science (reproducibility, objectivity, delineation of case group, adequacy of spectrum in case group, delineation of comparison group, adequacy of comparison group, and quantitative summary of results). RESULTS: Among the 40 inferential articles that studied 10 or more patients, only 5 (12.5%) complied with all 7 applicable standards, and 10 (25.0%) complied with all but 1 standard, whereas 25 articles (62.5%) failed to comply with 2 or more standards and 9 (22.5%) failed 4 or 5 standards. Most articles did not comply with standards for reproducibility (n = 25, 62.5%) or objectivity (n = 27, 67.5%); however, the majority of articles did comply with standards for adequacy of case group (n = 35, 87.5 %), adequacy of comparison group (n = 35, 87.5%), and quantitative summary of results (n = 36, 90%). CONCLUSIONS: Despite major laboratory advances in molecular genetic analysis, our data suggest that reported applications in clinical journals often have troubling omissions, deficiencies, and lack of attention to the different, but necessary, principles of clinical epidemiological science. Without suitable attention to fundamental methodological standards, the expected benefits of molecular genetic testing may not be achieved.

Early postpartum discharges. Impact on distress and outpatient problems.

OBJECTIVE: To determine the impact of shortened postpartum hospital stays on common clinical phenomena in a sociodemographically diverse, unselected group of general maternity patients. DESIGN: Observational cohort study in which the preapproved hospital stay duration of either 1 or 2 nights was set by third-party payers before each mother's admission. SETTING: Yale-New Haven Hospital, New Haven, Conn, from June 19 through August 10, 1995. PATIENTS: Two hundred forty-four volunteers from among 400 eligible deliveries. MAIN OUTCOME MEASURES: Readmission within 1 month of hospital discharge, report of outpatient morbidity and use of outpatient health services within 1 week of discharge, status of breast-feeding during the first post-discharge week, and patient satisfaction. RESULTS: At discharge from the hospital, the hospital stay was regarded as "too short" by 80 (47%) of 171 mothers and 19 (26%) of 73 mothers in the 1- and 2-night groups, respectively (P = .002). Although readmission rates were similar (5% vs 3%, P = .48), the 1-night group reported significantly more morbidity in the newborns (31% vs 16%, P = .03) and averaged more pediatric visits (96 vs 54 per 100 newborns, P = .002). Mothers in the 1-night group also reported more fatigue (49% vs 29%, P = .001) and more worries about their newborns' health (24% vs 11%, P = .02). They were less likely to start breast-feeding (64% vs 77%, P = .06), and, if they started, were somewhat more likely to stop prematurely (14% vs 8%, P = .43). A series of disturbing events was reported only in the 1-night group. CONCLUSIONS: In a relatively unselected group, mothers who stayed 1 night after routine vaginal delivery reported more distress and more pediatric problems and had greater use of outpatient health services than mothers who stayed 2 nights.