点击显示 收起
1 CeSSIAM 01011 Guatemala City Guatemala
2 Walter Straub Institute of Pharmacology and Toxicology Ludwig Maximillians University 00336 Munich Germany E-mail: k.schuemann{at}lrz.uni-muenchen.de
Dear Sir:
The December 2001 issue of the Journal featured the coincident publication of 3 articles related to anemia and its partition between iron deficiency and non-iron deficiency origins among African preschoolers living in areas of endemic malaria (1–3). Our attention was drawn to the article concerning an iron-supplementation intervention trial (3), which reported difficulties and caveats reminiscent of an experience of ours in Guatemala (4).
In the planning of their study, Zlotkin et al (3) approached their power and sample-size assumptions in the following manner: "On the assumption of 90% cure rates in the drops group and 80% in the sprinkles group and with a type I error set at 0.05 and a 0.9 probability of detecting a true difference, the final sample size estimate was 286 subjects per group." They expected up to 90% of anemic Ghanaian infants and toddlers to have iron deficiency and iron-responsive anemias. This expectation raises 2 points in our minds. The first of these points is that such an expectation conflicts with the sense of the introductory section of their paper, where they state, "...we tested the hypothesis that the response to treatment of anemia would be better with 2 mo of treatment with microencapsulated ferrous fumarate sachets daily than with ferrous sulfate drops..." (3). Both of these statements use wording that seems to refer to a one-tailed test, but with the latter statement inclining toward superiority for the sprinkles and the former statement inclining toward a 10% better efficacy for the drops. Thus, it is important that the authors clarify whether the study was powered for a one-tailed or two-tailed test of the hypothesis of difference and that they specify how many degrees of freedom were assumed for the statistical comparison. This is not to question their conclusion of no difference between 58% and 56% recovery rates, but to address a conceptual point. If a one-tailed test were conducted as stated with the power assumptions, then no conclusions of superiority for sprinkles could have been possible; one would be testing an asymmetric hypothesis to find either 1) that the ferrous sulfate was 10% better than fumarate or 2) that it was not.
The second point relates to their assumptions about the magnitude of the response to anemia treatment. The actual findings in their study fell far short of their expectations for a cure rate: in the 2-mo trial, 80 mg microencapsulated ferrous fumarate added to porridge (sprinkles) cured 56.3% of the anemia, and 40 mg ferrous sulfate in elixir (iron drops) cured 57.7% of the anemia. The optimistic interpretation of this finding is that 40% of the children were compromised in their ability to respond to iron because their anemia was due to malaria, another infection, or some other micronutrient deficiency. Thus, in fact, 100% of the "truly responsive" candidates in both wings of the intervention trial were cured by the iron treatments. Such a direct and simple interpretation could be seen as being bolstered by the coincident findings in malarial areas of Kenya (1) and the Ivory Coast (2) that 22% and 44%, respectively, of the subjects with anemia did not have iron deficiency and by the authors’ own findings that malaria-infected children have a 23% lower chance of responding to iron treatments than do children without malaria (3).
Unfortunately, a more complex and pessimistic interpretation seems to be more logical; this interpretation even challenges the possibility of attributing any specific fraction of the anemia response to the iron treatments in the absence of concurrent observations on a third treatment wing, which serves as a placebo control. The authors address this issue in their statement: "Because it would be unethical to provide a placebo to a child with anemia, we did not include a placebo control" (3). Such a proclamation of ethical absolutism could be taken more seriously had it not come from 2 of the institutions that conducted one of the largest placebo-controlled trials of vitamin A and child morbidity and mortality ever conducted (5, 6). Are we to accept that 2 mo of observation of children with hemoglobin in the 70–99-g/L range is unethical, but it is acceptable to follow children at risk of hypovitaminosis A for an entire year in a placebo trial with mortality among the outcome variables? The consequence of this doctrine for anemia research in Ghana would mean that any field team in that country must surely bring liters of iron drops along with the HemoCue apparatus and lancets in their backpacks, because teams would be obliged to treat almost all comers after any survey. This drawback is especially valid if one realizes that newer international standards for "anemia" in this age group have a cutoff of 110 g/L (7). If 65% of Ghanaian infants and toddlers in this region have a hemoglobin below the 100-g/L anemia definition used to enroll a presumably iron-responsive sample (3), over 90% might be classified as anemic with the more conventional 110-g/L criterion. It would be worth knowing what treatment, if any, was offered by Zlotkin et al to those children found on initial screening to be in the < 70-g/L range of hemoglobin (ie, severely anemic) and those found to be in the 100–109.9-g/L range (ie, mildly anemic by international standards).
Without a placebo group, moreover, we are at a loss to assess the attributable efficacy of either form of iron in this study (3). Because all the children were chosen to be in the lower part of the hemoglobin distribution, mathematical regression to the mean would account for some of the test-retest increment in hemoglobin concentration. This regression-to-the-mean term, moreover, might be larger than anticipated, given the fact that capillary finger-stick samples were used. Morris et al (8) examined the reliability (within-individual variability) of measurements of hemoglobin concentrations in capillary blood. From samples taken from the same persons, either concurrently from different anatomical sites or on consecutive days, they showed that concentration measurements in capillary blood have lower reproducibility than do those in venous blood, which is attributable to larger biological variation with peripheral sampling than with venous sampling. Hence, those children admitted into the eligible study pool on the basis of a hemoglobin concentration just below 100 g/L at baseline might have a hemoglobin concentration just above the criterion level at the second measurement merely because of the inherent unreliability of capillary samples (8). Another reason for the improved anemia status of the subjects may simply have been the development of the subjects during the 2 mo of the study; anemia rates tend to improve with age after infancy. Moreover, the effects of regression to the mean or asymmetrical diagnostic misclassification, in addition to developmental changes, could account for some of the observed improvements in the rates of anemia; thus, to determine the true degree of improvement in anemia attributable to the iron treatment, there would have to be a no-treatment (placebo) wing in the study design.
Aside from the issues of quantification with regard to efficacy, we also encountered a problem in the authors’ evaluation of safety. The authors comment on the similarity of the 14.5% rate of diarrhea with the ferrous sulfate treatment and the 12.8% rate with the ferrous fumarate treatment (3). Although we can readily accept the conclusion of no difference between treatments with regard to diarrhea experience, the design does not provide sufficient evidence to safeguard against intrinsic adverse effects of microencapsulated ferrous fumarate. Only by knowing that a no-treatment group also had diarrhea incidence in this range could we exonerate the iron treatments of causing the gastrointestinal symptoms. With regard to the assumptions of safety for a new treatment, moreover, its comparison with a placebo is procedurally indispensable. Even if the authors offer us "efficacy relativism" in the conclusion on the basis of comparative effects with the proven therapy as the standard for anemia cure (3), we simply cannot accept this logic of "safety relativism" with regard to adverse effects. It would not be sufficient to argue that sprinkles are safe on the basis of the fact that they do not have any more adverse effects than the elixir form of ferrous sulfate. The latter is an old medication that may have been grandfathered into the pharmacopeia with safety criteria that we may not now accept for a new agent or format. For new questions of safety, comparison against no exposure would still seem to be needed at some point.
The final conundrum for the ethical imperative to treat all anemias on sight comes from the findings of the 3 previously mentioned African studies in malarial areas. In Kenya, 22% of anemic children with a median age of 19 mo did not have iron deficiency anemia (1); in the Ivory Coast, 44% of anemic children with a mean age of 49 mo did not have iron deficiency anemia (2); in Ghana, 40% of anemic children with a mean age of 13 mo did not have iron-responsive anemia (3). Malaria is not the only possible confounding factor. In our 10-wk, placebo-controlled intervention with heme iron and ferrous sulfate in children with a mean age of 21 mo in the nonmalarial Guatemalan highlands, 43% of anemic children (hemoglobin concentration < 115 g/L) had no evidence of iron deficiency (4). Thus, if we are to believe Zlotkin et al that "it would be unethical to provide a placebo to a child with anemia," the question is what do we treat the child with to cure the anemia? Iron will work for some children, although we do not know a priori which ones; what do we offer to the rest, and when?
This is not just an ethical question for investigators and their populations, who are screened for eligibility on the basis of anemia. It underlies a very important program and policy issue for micronutrient deficiency interventions in developing (and developed) countries: when is an anemia a nutritional anemia, and when is a nutritional anemia an iron deficiency anemia? The results of all of the studies from Africa and Central America cited in this letter seem to converge toward a conclusion that "close to 100%" can no longer be the answer for either question. We may be obliged to do the "unthinkable," namely, to screen and diagnose anemias before assigning treatments both in human research and in the public health domain. The experience reported in the articles from Kenya (1) and the Ivory Coast (2) provides a basis for such diagnostic screening.
Finally, the authors leave some numerical issues to be resolved. Although they reported screening 880 children and enrolling 557, in the side effects subsection of Results, they state, "Seventy-four percent (933 of 1277) of the mothers of children in the drops group, " and "Diarrhea was reported in 76 of 523 ... subjects in the drops group and in 62 of 486 ... in the sprinkles group." These latter figures must represent typographical errors that were not detected in editing.
REFERENCES