Statistics from Altmetric.com
Editor,—It is somewhat self contradictory to suggest that “a transferrin saturation of <30% may help the diagnosis” if there is still doubt about validation of iron deficiency after receipt of the serum ferritin result, the authors having previously acknowledged that the latter is “the most powerful test for iron deficiency” (Gut2000;46(suppl IV):iv1–5). Statistical considerations which dictate that serum ferritin will always outrank transferrin saturation in predictive power have their basis in the comparison between the receiver operating characteristic (ROC) curves for serum ferritin versus transferrin saturation, yielding values of 0.91 versus 0.71 (p<0.001) for the area under the curve.1Statistical considerations also dictate acknowledgement of mean corpuscular haemoglobin (MCH) as a predictive entity in its own right following documentation than an MCH of <27 pg was superior to a mean corpuscular volume (MCV) of <77 fl in predicting serum ferritin levels of <20 μg/l.2 All low MCV values had low MCH values but nine hypoferritinaemic patients with low MCH had MCV within the normal range.2
In my own study, comprising 201 subjects with iron deficiency (characterised by serum ferritin <18 μg/l), the MCH conferring optimum trade off between sensitivity (65.2%) and specificity (65.9%) for iron deficiency was <24 pg, and this yielded a positive predictive value of 70%. By contrast, for MCV, optimum trade off between sensitivity (61.7%) and specificity (59.1%) was obtained with a cut off level of <77 fl, giving a positive predictive value of 65%.3 There were 31 patients with an MCH <26 pg in the presence of an MCV >80 fl compared with only four with an MCV <80 fl in the presence of an MCH >26 pg3 and, among these, four had an MCH <24 pg in the presence of an MCV >77 fl in contrast with only one with an MCV <77 fl in the presence of an MCH >24 pg. In my study, the most stringent cut off diagnostic level for iron deficiency was a serum ferritin level <10 μg/l found in a subgroup of 145 subjects. At this level, the MCV characterised by optimum trade off between sensitivity (65%) and specificity (66%) was <76 fl (identical with the cut off level in the guidelines), and this yielded a positive predictive value of 55%. Correspondingly, the optimum MCH was either <24 pg, characterised by sensitivities, specificities, and positive predictive values of 74%, 59%, and 80%, respectively, or <23 pg, characterised by sensitivities, specificities, and positive predictive values of 58%, 75%, and 62%, respectively.
Editor,—Suggesting both that transferrin saturation may help in the diagnosis and that ferritin is the most powerful test for iron deficiency anaemia (IDA) is not contradictory. Being the most powerful test does not mean it is always reliable. For example, in inflammatory conditions such as rheumatoid arthritis, ferritin may be normal even if there is iron deficiency.
We find the reference to the greater reliability of mean corpuscular haemoglobin (MCH) compared with mean corpuscular volume (MCV) in diagnosing IDA interesting. We agree that MCH can be useful in the diagnosis of iron deficiency. However, none of the papers quoted takes account of the red cell distribution width (RDW). We wonder if Dr Jolobe would still be able to demonstrate the superiority of MCH compared with MCV if anaemic patients with a normal MCV but raised RDW were excluded. We explain in our guidelines that combined deficiency (that is, iron deficiency together with B12 and/or folate deficiency) may be associated with a normal MCV and may be recognised by a raised RDW.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.