Observational studies reproducibility

A recent Editorial in the BM by loannidis et al has questioned the widely held view that the only scientifically acceptable studies are randomised controlled clinical trials, and that observational studies are unreliable and unre-producible. They cite two papers reviewing more than 20 studies which show that the correlation between observational and randomised controlled studies was remarkably close, and that the reproducibility of observational studies was good. 

FIGURE 17.5 Comparison of observed risk of a coronary heart disease event ( ) and predicted risk (lines) for patients receiving either placebo or pravastatin over the 4.4-year duration of the West of Scotland Coronary Prevention Study. (Reproduced with permission from West of Scotland Coronary Prevention Group. Circulation 1998,97 1440-5.)... 

Epidemiologic observations consistendy show lower rates of prostate and breast cancer, and osteoporosis in populations that consume soy (Anderson Garner, 1997 Moyad, 1999 Nagata et ah, 2001a Wu et al., 1998). Intervention studies were conducted to investigate whether these observations are reproducible. These studies will help clarify whether the results reported in epidemiologic studies are causal, and due specifically to soy products (Reinwald Weaver, 2006 Rice Whitehead, 2006). 

Cullis and coworkers studied the oxidation of ethanethiol in aqueous alkaline solution under constant pressure of oxygen. They observed low reproducibility of the oxygen uptake rates even when careful precautions were taken to avoid the presence of adventitious impurities. Under their conditions (EtSH 0-3-0-5 M NaOH 0-5-2 M the base always in excess) the stoichiometry of the reaction was found, in agreement with other authors to be ... 

NMR studies carried out on nuclei such as FI, and Si this does not really impose any limitation on their observation. Broader spectral lines can be reproduced by pulse teclmiques, provided that corrections are made for the RF-irradiation and probe responses but this requires carefiil calibration. Such corrections have been most extensively used for examining satellite transition spectra from quadnipolar nuclei [11]. 

We also have studied fluid distribution in the pore H = 6 (Fig. 12(b)) at Ppq = 4.8147 and at two values of Pp, namely at 3.1136 (p cr = 0.4) and at 7.0026 (pqOq = 0.7 Fig. 12(b)). In this pore, we observe layering of the adsorbed fluid at high values of the chemical potential Pp. The maxima of the density profile pi(z) occur at distances that correspond to the diameter of fluid particles. With an increase of the fluid chemical potential, pore filhng takes place primarily at pore walls, but second-order maxima on the density profile pi (z) are also observed. The theory reproduces the computer simulation results quite well. 

---