Abstract
Modern computers have not so much revolutionised medical research as widened its scope and made it rather easier to do. Within living memory, research data would be collected on hundreds if not thousands of pieces of paper or cards and analysed by hand. Statistical tests were done with a pencil, paper and slide rule. Things gradually ‘developed’ and data began to be encoded onto punch cards and feed into computers the size of wardrobes for rudimentary analysis. Only recently have we been afforded the privilege of entering data straight into a portable machine capable of instantaneous analysis of staggering complexity. Younger researchers will take this kind of thing for granted but I for one will probably never stop appreciating the fact that software as quotidian as Microsoft Excel can do in less than a second what would otherwise take me half an hour, two cups of tea and a lie down afterwards.
Computers enable us to analyse data from huge numbers of people; from populations. Two studies in this issue demonstrate the value of this approach. Chin-Kuo Chang and colleagues from the Institute of Psychiatry report on their analysis of suspected neuroleptic malignant syndrome cases from a large, anonymised electronic patient register. There were 183 suspected cases, 43 of which fulfilled at least one of the six diagnostic criteria applied. Interestingly, only one case met all six criteria, highlighting the immense difficulty clinicians face in diagnosing the condition in practice. In addition to this, 23% of the identified cases not meeting any criteria showed grossly elevated serum creatine kinase – the measure most often used in practice to identify the condition.
Using a similar approach, Paul Kurdyak and co-workers examined databases held for residents of Ontario, Canada (with a population of over 13 million). They were testing the possibility that the co-prescription of hepatic enzyme inhibiting antidepressants to people on metoprolol would lead to an increase in admissions for bradycardia. Despite this being a well known adverse interaction, the authors found no evidence for increased admissions with bradycardia. This does not mean that there is not an interaction, only that it does not seem to be severe enough to be clinically relevant, at least at the level of admissions to hospital.
The other two full papers inn this issue do not rely on electronic data but utilise computer technology in their complex analyses. Goekoop and colleagues report on their examination of potential markers for sub-types of depression. Amongst other findings, it was shown that psychotic depression was characterised by elevated levels of plasma norepinephrine and vasopressin, so supporting the contention that psychotic depression is a distinct subcategory of major depression. Using a different set of markers (brain-derived neurotrophic factor, vascular endothial growth factor and leptin) workers from Turkey found no difference in concentrations between those with melancholic depression and a group of healthy controls. There were however some associations with severity of depression and number of depressive episodes.
Each of the papers in this issue were submitted on-line via a computer of a size and sophistication unimaginable even twenty years ago. At that time and before, papers were submitted in triplicate and by post. Pictures, figures and tables were submitted separately; some graphs drawn by hand. Computers have made the whole process easier, quicker and more efficient. But while we dwell on the hitherto undreamt-of complexity of today’s machines, keep in mind that, in the not too distant future, what we have now will be seen as laughably crude and rudimentary.