Serum ferritin was the best predictor of iron deficiency in elderly patients with anemia
ACP J Club. 1991 Sept-Oct;115:54. doi:10.7326/ACPJC-1991-115-2-054
Patterson C, Guyatt GH, Singer J, Ali M, Turpie I. Iron deficiency in the elderly: the diagnostic process. Can Med Assoc J. 1991 Feb 15;144:435-40. [PubMed ID: 1993290]
To evaluate the usefulness of the history, physical examination, and laboratory tests in diagnosing iron deficiency in elderly patients with anemia.
Analytic survey of physicians' estimates of iron deficiency compared with bone marrow iron stores.
2 university-affiliated hospitals in Canada.
259 consecutive patients > 65 years of age who had anemia documented on 2 occasions between January 1984 and March 1988. Anemia was defined as a serum hemoglobin level ≤ 120 g/L (men) or ≤ 110 g/L (women). Patients with a recent blood transfusion or documented acute blood loss were excluded.
Description of test and diagnostic standard
The attending physician (an internist, geriatrician, or hematologist) defined a pretest probability estimate of iron deficiency anemia based on the history, physical examination, and the initial complete blood count. Post-test probabilities were estimated with knowledge of transferrin saturation, erythrocyte protoporphyrin, and serum ferritin levels. The diagnostic standard was a blinded interpretation of bone marrow iron level.
Main Outcome Measures
Likelihood ratios for various cut points and pretest and post-test probability estimates and the area under the receiver operating characteristic curve.
235 patients had bone marrow aspirates that were interpretable. Physicians' pretest and post-test probability estimates were obtained for 196 and 173 patients, respectively; approximately one third of the patients were deemed to be iron-deficient by bone marrow diagnosis. Likelihood ratios for patients with pretest probability estimates of ≤ 0.4 and > 0.4 were 0.6 and 6.0, respectively, indicating that physicians' pretest estimates were quite powerful. Prediction of bone marrow diagnosis of iron deficiency was slightly more accurate by serum ferritin results alone than by physicians' post-test estimates (area under the receiver-operating characteristic curves 0.91 vs 0.87, P = 0.03). Likelihood ratios for the serum ferritin level were 0.1 for > 100 µg/L, 0.5 for 46 to 100 µg/L, 4.0 for 19 to 45 µg/L, and 30.6 for ≤ 18 µg/L. Diagnostic power improved, however, when ferritin levels were interpreted in conjunction with physician pretest estimates. For ferritin levels of 46 to 100 µg/L, likelihood ratios were 0.3 and 5.3 for pretest probabilities of ≤ 0.4 and > 0.4, respectively.
Serum ferritin was the most powerful predictor of iron-deficiency in elderly patients, but pretest estimates based on history-taking, physical examination, and the initial complete blood count can provide additional information. The lower accuracy of post-test estimates may indicate occasional misinterpretation of ferritin levels by clinicians.
Source of funding: In part, Ontario Ministry of Health.
Address for article reprint: Dr. G.H. Guyatt, Department of Clinical Epidemiology and Biostatistics, McMaster University Health Sciences Centre, 1200 Main Street West, Hamilton, Ontario L8N 3Z5, Canada.
In our fondness for new tests, we often forget the diagnostic power of basic clinical data. Patterson and colleagues have shown the role of clinical judgment in the diagnosis of iron deficiency anemia. Serum ferritin remains the most useful measure, particularly when levels are ≤ 18 µg/L or ≥ 100 µg/L (1). For patients whose values fall within this range, diagnostic power improves when the history, physical examination, and initial complete blood count are considered in addition to serum ferritin. We should not, however, discount the possibility of iron deficiency in elderly patients without knowing the ferritin level. Even when pretest suspicion was lowest (< 0.2), nearly one quarter of the patients in this study were iron deficient. Interestingly, physicians may occasionally be misled by knowledge of ferritin levels, tending to interpret "low-normal" values (19 to 45 µg/L) as evidence against iron deficiency. Regardless of pretest suspicion, values in this range are almost 4 times as likely to come from patients with as without iron deficiency.
It would be useful to define which elements of the history and physical examination were the best guides of clinical judgment. A high suspicion of iron deficiency (based, for example, on heme-positive stools or recent fall in hemoglobin) appears to be more useful than a low clinical suspicion. Despite these limitations, this work is a valuable attempt to evaluate physicians' most fundamental skill: the use of basic clinical information to diagnose disease.
David Atkins, MD
University of WashingtonSeattle, Washington, USA