Directiva N°013-2016-CG/GPROD

The m ajor problem at the outset o f this study was the small size o f the potential tw in sample available, if a traditional population-based m ethod was em ployed. T here were two reasons fo r this: 1) the rarity of M ND, with a prevalence rate o f approxim ately 4-7/100,000; (this is obtained from the annual incidence rate of betw een 1.5 - 2.2/100,000 m ultiplied by the mean life expectancy o f 2.5 - 3 years (tables 1 and 3). 2) the proportion o f twins w ithin the general population. The ratio o f m ultiple births to singletons was only 1:80 live births before 1940, w hen the m ajority o f recent M ND sufferers were born. The ratio o f living tw in pairs is reduced by the frequency of in fan t death, which was greater among twins. This ranged from 151 per 1,000 live births in 1901 to 60 per 1,000 live births in 1941 (Nissel 1987), giving a survival ratio o f approxim ately 1 tw in pair per 100 people. T herefore, it can be seen that for a population o f 48.5 m illion (England and Wales), there are betw een 1900 and 3400 people living w ith M ND at any tim e, o f w hich around a hun d red th part will be twins (19 -34), a very small num ber. D ifficulties identifying and tracing the tw ins would reduce this figure fu rth e r and create bias.

To overcom e this fundam ental problem , an alternative and novel m ethodology was devised, term ed the D EA TH DISCORDANT TWIN PA IR M ETH OD. Basically, this involved iden tify in g twins among the people who have already died from M ND , and tracing their co­ twins. Any living co-tw ins could then be approached about participating in the study, and death certificate details obtained for the dead co-tw ins. Inform ation pertaining to both twins would be sought from the co-tw in.

Initial negotiations w ith the O ffice o f Population Censuses and Surveys fo r E ngland and Wales (OPCS) were concerned w ith the availability and access to the necessary docum entation, the logistics o f conducting such a study, and use o f OPCS m edical research services fo r tracing the co-tw ins.

D ata Collection: Interview ing the subjects was chosen as the most ap propriate m ethod for collecting the data. This is a tim e consum ing exercise, but often yields b etter results because o f personal contact, especially when dealing w ith sensitive topics.

Interview ing allows expansion and exploring o f an area from the initial question, enabling m ore detailed inform ation to be gathered. Obviously this is not possible w ith se lf­ adm inistrated questionnaires, w hich need to be very precise and carefully w orded, to avoid confusion and alienation w hich may produce unusable answers (O ppenheim 1966).

It is im portant to ensure that inform ation is gathered from all subjects under sim ilar conditions. This is to reduce the unknow n effects of the environm ent, and interaction w ith the interview er, on the su b ject’s responses. O ften, very little can be done to control the physical environm ent. V isiting the subjects at home how ever, provided a com parable place w hich was the least threatening and hopefully most congenial for the subjects. In an effo rt to control the e ffe c t o f the researcher on the subjects responses, the interview s followed a set questionnaire and the researcher attem pted to behave sim ilarly w ith all the subjects.

V aliditv o f Proxv Data: Since the M ND cases had all died, it was necessary to use proxy data to obtain the required inform ation, but obviously the co -tw in data was given directly. This w ould m ean that the quality o f the data would inevitably be unbalanced. U sually data is obtained directly from both cases and controls, or alternatively, if the case has died, relatives o f both can be used instead to avoid d iffe re n tia l recall. L ittle m ethodological work has been u n d ertak en to validate proxy data against direct inform ation. Linet et al 1989 exam ined the recall o f past m edical history by cases, as com pared w ith that o f a relative surrogate, and v alidated each opinion against hospital records. G enerally, the surrogate recall was inferior, although this depended on the disease. This has also been shown fo r sm oking data. T h e refo re, the general effect o f proxy data collection would appear to be the underestim ation o f actual events (a type II error), rath er than overestim ating any exposure (a type I error). In m ost case-control studies the opposite effects can occur since cases have a tendency for selectively recalling events they feel are im portant to their disease, w hich supports their own theory as to a cause (Sackett 1979).

To im prove the quality of data gathered for the case tw in, it was decided that w here possible, any living spouses could be approached through the co -tw in s a fte r their interview . This

would provide two independent sources of inform ation and thus increase the validity o f the data collected fo r the case twins.