Further studies are needed to establish a difference in outcome related to the use of a particular renal replacement technique.
However, it should be kept in mind that our sample size was too small to establish a difference in infrequent risk factors such as invasive diagnostic procedures.
The Cochrane meta-analysis, however, could not establish a difference in the number of patients with hemodynamic instability (however defined) or with hypotension.
A sample size calculation based on SD (standard deviation) values in previous studies, determined that 40 patients were needed to establish a difference in LDL-cholesterol reduction of 15% between the intervention and control groups (power 0.80, significance level 0.05).
However, meta-analyses comparing EN with PN did not establish a difference in mortality and the lower incidence of infectious complications with EN may be explained largely by the higher incidence of hyperglycemia in patients receiving PN [ 36, 39].
However, if further French studies conducted on large populations establish a difference in survival according to type of cancer care health centre, the geographical disparities in access to reference cancer site established in our study would imply geographical inequalities in cancer care management as in a recent English study (Kim et al, 2000), especially for women and elderly patients.
For example, to establish a difference of 2% in postoperative bleeding with a α = 0.05 and a power of 80%—about 500 (inclusion of estimated lost to follow-up) patients are needed per group.
Furthermore, with survival slightly favouring tamoxifen, confidence intervals on the hazard ratio established that a difference in favour of tamoxifen plus prednisolone of even 5% at 5 years was very unlikely (P < 0.02).
Clearly, if firmly established, a potential difference in cure rates of this magnitude is clinically important and this emphasises the need to select the most appropriate model.
