63 Fibrinogen genes and peripheral arterial disease S!R,&mdash;Dr Fowkes and colleagues (Mar 21, p 693) report an interesting association between fibrinogen and peripheral arterial disease. Using the Bcll restriction fragment length polymorphism on the 13-locus of the fibrinogen gene, they found a frequency of 0-097 of the rare B2 allele in controls (n= 126) and 0-197 in cases (n=121) (p<0-01; derived from table u). Not discussed is the frequency of the B2 allele in other studies-ie, 0-247 in a population of 66 men and 25 women in working groups in north-west London;l 0 161 in 96 healthy individuals from London (and 0’ 189 in 45 young patients with peripheral arterial disease and 0.141 in another 85 patients attending a peripheral arterial disease clinic);2 0-174 in 118 unrelated individuals recruited among monozygotic twins in the neighbourhood of Oslo; and 0- 143 in 49 healthy volunteers from the westem Netherlands.4 There is a significant difference (X2 test) between the frequency in healthy volunteers in the study of Humphries et all and the frequencies found by Wiseman et al2 and by us4 (p < 0-05) on the other hand. Wiseman et al did not observe a significant difference between healthy controls and either young or other patients with peripheral arterial disease. 0.00 I -, I I 1 3 5 7 9 11 13 15 17 19 21 Number of Individuals (x 100) Cumulative APOE*2 frequency in healthy Dutch males. Results of three independently obtained randomisation series in a population of 2000. At each randomisation twenty sets of 100 individuals were generated ; individual frequencies of all studies (left) and cumulative allele frequencies (right) are shown. Horizontal bars in left panel represent mean and 95% CL W = first, . = second, and [] = third series. An explanation for these discrepant results does not lie in deviations from the Hardy-Weinberg equilibrium since all studies did not deviate significantly from it. We therefore evaluated the influence of sample size on gene frequencies. All the above- mentioned studies were on samples of only 45-126 individuals. We used data from a study on 2000 healthy Dutch volunteers typed for the apolipoprotein E polymorphism. APOE*2 shows a gene frequency of 0.070-0.100 in most populations studied so far.5,6 By means of a randomisation programme we generated three series of twenty groups of 100 persons, and calculated the APOE*2 frequency in each group. In all groups of 100 the allele frequencies were in Hardy-Weinberg equilibrium. The cumulative apoE2 frequency for each series converges to the mean value of 0-082 (figure). However, before 400-500 individuals was reached, the allele frequency showed large variability-ie, with a population size of about 100 individuals (as in most of the above studies) the estimated gene frequency could deviate strongly from the "real" mean. Our simulation suggests that for allele frequencies of 0. 100 or less a population size of 100 is not large enough. The frequency of the rare allele for fibrinogen determined by the BclI polymorphism in patients with peripheral arterial disease can only be known definitively if populations larger than 400-500 are studied. Gaubius Laboratory (IVVO-TNO), PO Box 430, 2300 AK Leiden, Netherlands, and MGC Department of Human Genetics, Sylvius Laboratory, Leiden University P. DE KNIJFF A. C. W. DE BART R. R. FRANTS L. M. HAVEKES C. KLUFT 1. Humphries SE, Cook M, Dubowitz M, Stirling Y, Meade TW. Role of genetic variation at the fibrinogen locus in determination of plasma fibrinogen concentrations. Lancet 1987; i: 1452-54. 2 Wiseman SA, Jaye PD, Powell JT, Humphries SE, Greenhalgh RM. Frequency of DNA polymorphisms of the apolipoprotein B and fibrinogen genes in young patients with peripheral arterial disease. In: Zilla P, Fasol R, Callow A, eds. Applied cardiovascular biology, 1989, vol 1: International Society of Applied Cardiovascular Biology. Basel: Karger, 1990: 118-23. 3. Berg K, Kierulf P. DNA polymorphisms at fibrinogen loci and plasma fibrinogen concentration. Clin Genet 1989; 36: 229-35. 4. de Bart ACW, de Maat PMP, Hennis BC, Kluft C. Longitudinal stability of plasma concentrations of fibrinogen in healthy volunteers. Perfusion 1991; 4: 418. 5. Smit M, de Knijff P, Rosseneu M, et al. Apolipoprotein E polymorphism in the Netherlands and its effect on plasma lipid and apolipoprotein levels. Hum Genet 1988; 80: 287-92. 6. Davignon J, Gregg RE, Sing CF. Apolipoprotein E polymorphism and atherosclerosis. Arteriosclerosis 1988; 8: 1-21. ** This letter has been shown to Dr Fowkes and colleagues, whose reply follows.-ED. L. SIR,-Dr de Knijff and colleagues are incorrect in their interpretation of tests of significance in relation to sample size. In a case-control study, the main purpose of a significance test is to determine the extent to which an observed difference may have occurred by chance and takes account of random error in the sampling of cases and controls. A statistically significant difference suggests that chance is an unlikely explanation, no matter what the sample size. In fact, too small a sample is a difficulty when a non-significant result occurs because there may be a genuine difference that a larger study might show as statistically significant. This could, for example, have contributed to the lack of significance seen between cases of peripheral arterial disease and controls in Wiseman and colleagues’ study.l The B2 allele frequencies in the studies cited by de Knijff et al are of a similar order of magnitude. By contrast to the Edinburgh artery study, however, these studies were not based on proper random samples in the general population. It is meaningless to calculate significant differences between these studies when the populations differ in methods and criteria for selection. De Knijffs simulation exercise is both unnecessary and inappropriate. Statistical theory will provide the distribution of the sample allele frequency for any sample size, without the need for simulation. It is, in any case, the distribution of the differences between the frequencies in two samples that is relevant here. The usual way to determine the adequacy of sample size is by power calculations? We estimated that to detect a difference in allele frequency of 10% with 90% power and p<0’05 would require about 260 alleles in each group. Our sample sizes were 242 alleles for the cases and 252 alleles for the controls. We conclude that our sample sizes were adequate, but would reiterate the point we made that the relation between fibrinogen genotype and atherosclerosis needs to be investigated in other randomly sampled populations. Wolfson Unit for Prevention of Peripheral, Vascular Diseases and Medical Statistics Unit, University of Edinburgh, Edinburgh EH8 9AG, UK F. G. R. FOWKES P. T. DONNAN R. J. PRESCOTT Department of Medical Genetics, University of Glasgow, Duncan Guthrie Institute, Glasgow J. M. CONNOR 1. Wiseman SA, Jaye PD, Powell JT, Humphries SE, Greenhalgh RM. Frequency of DNA polymorphisms of the apolipoprotein B and fibrinogen genes in young patients with peripheral arterial disease. In: Zilla P, Fasol R, Callow A, eds. Applied cardiovascular biology, 1989, vol 1: International Society of Applied Cardiovascular Biology. Basel: Karger, 1990: 118-23. 2. Hennekens CH, Buring JE. Epidemiology in medicine. Boston: Little Brown, 1987: 260-62. Neutralising antibodies against C3NeF in intravenous immunoglobulin SIR,-C3 nephritic factors (C3NeF) are autoantibodies directed against neoantigenic determinants of the C3bBb convertase of the alternative complement pathway, and are found in the serum of patients with membranoproliferative glomerulonephritis (MPGN) and/or partial lipodystrophy.1-3 The beneficial effect of intravenous immunoglobulin (IVIG) in autoimmune diseases has been