but subsequent tests will not be at that level, and decisions based on them are likely to be erroneous. Second, Brown refers to a conun¬ drum by O'Brien: "Suppose a clinical investigator at MegaClinic conducts a clinical trial on treatments A, B, and C in three groups of patients. At the conclusion of the trial, the data ana¬ lyst might perform three compari¬ vs C. The sons: A vs B, A vs C, and well-trained analyst would see that the value for each comparison should be adjusted downward from the overall =£.05. That is, he/she would penalize himself/herself for doing three statistical tests." The point is that when the analyst, X, performs the first comparison (A vs B), there is no difficulty. When the second comparison, A vs C, is made, X is using the same group A as in the first comparison, and it is likely that be The extent dence is unknown, and the value effectively decreases by an unknown amount. When X performs the third comparison, vs C, X is performing a test that is likely to have depen¬ dence with the first test, which used group B, and with the second test, which used group C. Hence, the value should be adjusted, to an un¬ known extent, downward from the overall P^.05. The analyst should pe¬ nalize himselfyherself for doing three statistical tests. On the other hand, "suppose that out in Podunk a study was conducted of just A vs B. Another investigator in Boondock studied A vs C. A third re¬ searcher in Foozball studied vs C. They each used, quite appropriately, P«.05 for evaluating their two sam¬ Why should the in¬ ple results. has the most compre¬ who vestigator hensive overview of treatments A, B, and C be punished with an adjust¬ value that is not ment to the the on separate studies done imposed in three less elaborate trials?" The answer is that, in the three separate studies, there is presumably no de¬ pendence between the tests per¬ formed and in the single study, there is likely to be unknown dependence. If each group referred to in the "co¬ nundrum" consisted of 100 patients, altogether, 300 patients are in the first study. In the separate studies, each of the three comparisons involves 200 patients (ie, 600 patients altogether). As fewer patients are involved, we would expect the first investigation to Neonates: the two comparisons dependent. . . . mutually of depen¬ are disadvantaged in some way com¬ pared with the second. Analysts should heed Brown's warning that repeated statistical tests on the same elements or subjects in¬ troduce an increase in the risk of a type I error. In the same vein, Dr David Kleinbaum commented dur¬ ing a course on epidemiology held last year in Sydney, Australia, that "if you torture data sufficiently, they will eventually confess to some¬ thing." TED BYRT, MED Clinical Epidemiology and Biostatistics Unit Royal Children's Hospital Flemington Rd Parkville, Victoria 3052, Australia 1. Brown GW. P values. AJDC. 1990; 144:493-495. Clavicular Fractures in Frequency vs Significance Sir.\p=m-\Iwish to thank Drs Joseph and Rosenfeld1 for their evaluation of the frequency and presentation of clavicular fractures in newborns. This is useful information for "frontline" pediatricians like myself, but I would like to make the following comments. The fact that the occurrence of newborn clavicular fractures has remained stable in the Western world during the past 85 years suggests to me that the New York State Health Department may be on the wrong track if it thinks that this is "an indicator of the quality of care." Also, it seems that clavicular fractures are benign occurrences that virtually always heal well without intervention. The only reason then to diagnose clavicular fracture would be to explain irritability with motion or asymmetric Moro's reflexes. That suggests to me that, in children without these symptoms, there may be no reason to make a significant effort to make this diagnosis. MICHAEL J. O'HALLORAN, MD MIDELFORT CLINIC LTD 733 W Clairemont Ave Eau Claire, WI 54702-1510 1. Joseph PR, Rosenfeld W. Clavicular fractures in neonates. AJDC. 1990; 144:165-167. In Reply. \p=m-\DrO'Halloran points out that the New York State Health Department may be on the wrong track in using frequency of newborn clavicle fracture as a guidelinefor quality of care. By and large, wethe think he may be correct. In addition, requirement for reporting newborn clavicle fractures in New York State may lead to unnecessary documentation and costs without any effect on the quality of care. Dr O'Halloran also points out that the only reason to diagnose newborn clavicle fractures would be to explain irrisymptoms, such as tability. Another reason to diagnose newborn clavicle fractures is to prevent embarrassment at the first office visit when a large callus is felt covering the healthy broken clavicle. Additionally, we would prefer to make the diagnosis rather than to have the parent bring the callus to our attention on or before the first visit. PAUL R. JOSEPH, MD Warren Rosenfeld, MD Department of Pediatrics unexplained Winthrop-University Hospital 259 First St Mineola, NY 11501 Clavicular Fractures in Neonates Sir.\p=m-\Weread with interest the article by Joseph and Rosenfeld1 in the Feb- ruary 1990 issue of AJDC. We have noticed an increased occurrence of clavicular fractures in our institution during the last 6 months (12 [1.48%] of 812 newborns). We looked at different variables, including maternal age, parity, infant's weight, maternal diabetes, length of labor, oxytocin (Pitocin) induction, delivery methods, local anesthesia, episiotomy, delivery time, and delivering physician (house staff or attending staff) to find the common factor(s) in these infants. The only common factor for seven of the affected infants was a shorter second stage of labor (mean, 11.4 minutes). In contrast to the authors' observation that only one of the 18 infants showed "classic" physical findings, we diagnosed fractures in 11 of 12 infants by the palpation of instability or crepitation over the fracture site. Roentgenography was used to confirm our clinical diagnosis. In one case, two of us (G.C. and J.P.) and an orthopedist felt a subtle crep¬ itation at the distal end of one clavicle; however, roentgenography failed to reveal a clear fracture. We noticed that a small number of Downloaded From: https://jamanetwork.com/ University of Nevada - Las Vegas by Huynh Le Thai Bao on 09/16/2021