Inter-rater agreement for SCAN interview
All the interviewers from all sites took part in a joint inter-rater reliability exercise (in English) involving both audio-taped interviews with study subjects and videotaped interviews with actors, role-playing a depressed subject. Item by item kappa statistics for SCAN items, were calculated comparing each interviewer's ratings against AF's "master" rating. A mean item by item kappa coefficient across all the sites of 0.77 (range 0.63 – 0.89) was obtained indicating a substantial level of inter-rater agreement.
Number of subjects, age at interview, age at illness onset and gender by site
For inclusion in the first part of the linkage analysis, 944 affected subjects were genotyped from the 8 study sites as follows: Aarhus 48, Birmingham 146, Bonn 110, Cardiff 126, Dublin 154, Lausanne, 56, London 111 and St Louis 193. The age at interview and age of illness onset by gender of the subjects recruited at each site are shown in Table .
Numbers of male and female subjects, age at interview and age of illness onset by site
Mean age at interview for both sexes combined for each site were as follows: Aarhus 44.13 years (standard error of the mean (SEM)1.5), Birmingham 48.44 years (SEM 1.1), Bonn 51.46 years (SEM 1.2), Cardiff 44.06 years (SEM 0.9), Dublin 43.32 years (SEM 1.0), Lausanne 47.21 years (SEM 1.3), London 45.67 years (SEM 1.0), St Louis 47.14 years (SEM 0.9). These mean age differences were statistically significant (Analysis of variance(ANOVA): F = 6.26 degrees of freedom (df) 7, 936, p < 0.001. Tukey Post hoc test: Dublin, Cardiff, Aarhus, London, St Louis, Lausanne, Birmingham < St Louis, Lausanne, Birmingham, Bonn).
Mean age at illness onset for both sexes combined per site were as follows: Aarhus 22.67 years (SEM 1.3), Birmingham 24.30 years (SEM 0.8), Bonn 27.28 years (SEM 1.2), Cardiff 23.47 years (SEM 0.9), Dublin 22.27 years (SEM 0.8), Lausanne 24.85 years (SEM 1.4), London 22.08 years (SEM 1.1), St Louis 18.17 years (SEM 0.8). These mean age differences were statistically significant (ANOVA: F = 9.82 df 7, 841 p < 0.001. Tukey Post hoc test: St Louis, London, Dublin, Aarhus, < London, Dublin, Aarhus Cardiff, Birmingham, Lausanne < Cardiff, Birmingham, Lausanne, Bonn).
However, there were no significant differences between sites for the numbers of men and women recruited (see Table ) (chi squared test = 6.83 df 7 p = non significant (ns)).
Number of probands, siblings and other relatives recruited by site
Although study participants were mainly affected proband/sibling pairs, there were a few families where parents were also included. The numbers of probands, siblings and parents recruited per site is shown in Table .
Number of probands and siblings recruited from each site
In total there were 369 families with 2 affected siblings, 36 families with 3 affected siblings, 7 families with 4 affected siblings, and 2 families with 5 affected siblings. In addition 53 parents were also interviewed and provided blood for DNA extraction. Thus there were 470 affected sibling pairs (calculated as number of pairs per family equals number of affected siblings minus 1).
Gender, age at interview, age at illness onset and marital status for all sites combined
Of the 944 subjects, 670 (71%) were female and 274 (29%) were males and hence, the female/male ratio was 2.45:1.
Mean age at interview for all female subjects was 45.40 years (SEM 0.5) and for all males subjects was 45.69 (SEM 0.8). There were no significant gender differences for age at interview (t = -0.33, df = 477.58, p = ns)
The mean age of illness onset for depressed male subjects was 22.61 years (SEM 0.7) compared to 22.52 years (SEM 0.4) for depressed female subjects. There was no significant sex difference for age of onset.(t = -0.11, df = 443.55 p = ns).
Fifty five percent of male subjects and 61 % female subjects were living with a partner (married or cohabiting), while 45 % male subjects and 39% female subjects were living alone (ie separated, widowed, divorced or never married). Female subjects were significantly more likely to be living with a partner compared to male subjects. (Chi squared test = 26.89 df = 1 p < 0.001).
Gender, age at interview, age of illness onset and marital status for probands, siblings and parents
There were 295 female and 117 male probands, 335 female and 144 male siblings and 40 female and 13 male parents included in the total sample. There were no significantly differences for the gender of probands, siblings or parents (chi squared test = 0.85, df = 2 p = ns).
The mean age at interview for probands was 45.94 years (SEM 0.6) and for siblings was 45.80 years (SEM 0.5). There were no significant differences for age at interview between probands and their siblings (t = 0.17, df = 872.95 p = ns)
Probands gave a mean age of illness onset of 20.22 years (SEM 0.6) while siblings reported a mean age of onset of 21.04 years (SEM 0.6). Again these differences were not statistically significant (t = -0.98, df = 882.93, p = ns)
There were also no significant differences between probands and their siblings for marital status; 161 probands and 170 siblings were living alone while 242 probands and 290 sibings were living with a partner (chi squared test = 0.81 df = 1 p = ns).
Genotyping was carried out by DeCode and the results checked for mis-specified relationships by the programs RELPAIR and Graphical Representation of Relationships (GRR) at the Institute of Psychiatry. RELPAIR compares the multipoint probability of the data conditional on the possible relationships, while GRR calculates the IBS mean and SD for each pair and plots these values, representing each type of relative pair using a different colour. Decisions about each problem pair were made on the basis of the results from both programs, although where there was discrepancy between the programs the GRR results were used.
To check genotypes with Mendelian and other pedigree errors the PEDSTAT and MERLIN programs were used.
These data cleaning processes resulted in 929 individuals being genotyped at 932 autosomal markers and 44 X chromosome markers. Success rates for the autosomal markers were above 61% and for 90% were above 86%. For the X chromosome the success rate was above 66%. For individuals the genotyping success rate was above 73% for autosomal markers and 61% for the X chromosome.