A patient with severe angina will often be eligible for either angioplasty (PTCA) or bypass surgery (CABG). Results from eight published randomised trials have been combined in a collaborative meta-analysis of 3371 patients (1661 CABG, 1710 PTCA) with a mean follow-up of 2.7 years. The total deaths in the CABG and PTCA groups were 73 and 79, respectively, with a relative risk (RR) of 1.08 (95% CI 0.79-1.50). The combined endpoint of cardiac death and non-fatal myocardial infarction occurred in 169 PTCA patients and 154 CABG patients (RR 1.10 [0.89-1.37]). Amongst patients randomised to PTCA 17.8% required additional CABG within a year, while in subsequent years the need for additional CABG was around 2% per annum. The rate of additional non-randomised interventions (PTCA and/or CABG) in the first year of follow-up was 33.7% and 3.3% in patients randomised to PTCA and CABG, respectively. The prevalence of angina after one year was considerably higher in the PTCA group (RR 1.56 [1.30-1.88]) but at 3 years this difference had attenuated (RR 1.22 [0.99-1.54]). Overall there was substantial similarity in outcome across the trials. Separate analyses for the 732 single-vessel and 2639 multivessel disease patients were largely compatible, though the rates of mortality, additional intervention, and prevalent angina were slightly lower in single vessel disease. The combined evidence comparing PTCA with CABG shows no difference in prognosis between these two initial revascularisation strategies. However, the treatments differ markedly in the subsequent requirement for additional revascularisation procedures and in the relief of angina. These results will influence the choice of revascularisation procedure in future patients with angina.

The observation of an exponential increase in senile dementia prevalence with age has led to the conclusion that this disease may be inevitable in those who live long enough. The alternative view is that at very high ages the prevalence rate levels off. Studies conducted to date have not included sufficient numbers of very old people to resolve this difference of opinion. The question is important both to our understanding of the biological mechanisms involved and for public health planning. We have carried out a meta-analysis of nine epidemiological studies of senile dementia that used DSM III diagnostic criteria and that included samples of elderly people over age 80. The resulting curve was best described as a flattened S curve that fitted a modified logistic function rather than an exponential pattern. The rate of increase in senile dementia prevalence was found to fall in the age range 80-84, and at around the age of 95 prevalence was seen to level off to about 40%. It seems that senile dementia is better conceptualised as an "age-related" (ie, occurring within a specific age range) rather than as an "ageing-related" disorder (that is, caused by the ageing process itself). Very elderly survivors may be at diminishing risk of dementia and this has implications for public health policy.

A systematic overview, or meta-analysis, of the randomised evidence on maximum androgen blockade (MAB) in advanced prostate cancer identified 25 trials that compared conventional castration (surgical or medical) versus MAB (castration plus prolonged use of an antiandrogen such as flutamide, cyproterone acetate, or nilutamide). Individual patient data were obtained from 22 of the trials. Median follow-up was 40 months, during which 57% of patients died (3283/5710). Crude mortality rates were 58% for castration alone and 56% for MAB. Life-table estimates of the corresponding 5-year survival rates were 22.8% and 26.2%, representing a non-significant improvement of 3.5% (95% CI 0-7%). Logrank time-to-death analyses found no significant heterogeneity between trials (or between the effects of different types of MAB) and no significant evidence of additional benefit in an overview of all these MAB trial results (2p > 0.1). The currently available evidence from randomised trials does not show that MAB results in longer survival than conventional castration.

We tested, under independent conditions, the reproducibility of evidence from two previous trials that homoeopathy differs from placebo. The test model was again homoeopathic immunotherapy. 28 patients with allergic asthma, most of them sensitive to house-dust mite, were randomly allocated to receive either oral homoeopathic immunotherapy to their principal allergen or identical placebo. The test treatments were given as a complement to their unaltered conventional care. A daily visual analogue scale of overall symptom intensity was the outcome measure. A difference in visual analogue score in favour of homoeopathic immunotherapy appeared within one week of starting treatment and persisted for up to 8 weeks (p = 0.003). There were similar trends in respiratory function and bronchial reactivity tests. A meta-analysis of all three trials strengthened the evidence that homoeopathy does more than placebo (p = 0.0004). Is the reproducibility of evidence in favour of homoeopathy proof of its activity or proof of the clinical trial's capacity to produce false-positive results?

Nicotine-replacement therapy (NRT) by gum, transdermal patch, intranasal spray, or inhalation is expensive but how effective is it? We have done a meta-analysis of controlled trials to see how effects on abstinence rates are influenced by the clinical setting, the level of nicotine dependency, the dosage of NRT, and the intensity of additional advice and support offered. Published or unpublished randomised controlled trials of NRT that have assessed abstinence at least 6 months after the start of NRT were identified and 53 trials (42 gum, 9 patch, 1 intranasal spray, 1 inhaler), with data from 17,703 subjects, were included in the analyses. Use of NRT increased the odds ratio (OR) of abstinence to 1.71 (95% confidence interval 1.56-1.87) compared with those allocated to the control interventions. The ORs for the different forms of NRT were 1.61 for gum, 2.07 for transdermal patch, 2.92 for nasal spray, and 3.05 for inhaled nicotine. These odds were non-significantly higher in subjects with higher levels of nicotine dependence but they were largely independent of the intensity of additional support provided or the setting in which NRT was offered. We conclude that the currently available forms of NRT are effective therapies to aid smoking cessation.

There is disagreement on the usefulness of comprehensive geriatric assessment (CGA) due to conflicting results from individual trials. We did a meta-analysis on 28 controlled trials comprising 4959 subjects allocated to one of five CGA types and 4912 controls. Published data were supplemented with reanalysed data provided by the original investigators. We calculated combined odds ratios of important outcomes by pooling data from individual trials with multivariate logistic regression. Combined odds ratio (95% confidence interval) of living at home at follow-up was 1.68 (1.17-2.41) for geriatric evaluation and management units, 1.49 (1.12-1.98) for hospital-home assessment services, and 1.20 (1.05-1.37) for home assessment services. Covariate analysis showed that programmes with control over medical recommendations and extended ambulatory follow-up were more likely to be effective. Our analysis suggests that CGA programmes linking geriatric evaluation with strong long-term management are effective for improving survival and function in older persons.

Management of stroke patients in specialist stroke units hastens recovery but is not believed to influence mortality. We did a statistical overview of randomised controlled trials reported between 1962 and 1993 in which the management of stroke patients in a specialist unit was compared with that in general wards. We identified 10 trials, 8 of which used a strict randomisation procedure. 1586 stroke patients were included; 766 were allocated to a stroke unit and 820 to general wards. The odds ratio (stroke unit vs general wards) for mortality within the first 4 months (median follow-up 3 months) after the stroke was 0.72 (95% CI 0.56-0.92), consistent with a reduction in mortality of 28% (2p < 0.01). This reduction persisted (odds ratio 0.79, 95% CI 0.63-0.99, 2p < 0.05) when calculated for mortality during the first 12 months. The findings were not significantly altered if the analysis was limited to studies that used a formal randomisation procedure. We conclude that management of stroke patients in a stroke unit is associated with a sustained reduction in mortality.

We did a meta-analysis of all published polychemotherapy vs supportive care clinical trials in patients with non-resectable non small cell lung cancer. 7 studies with more than 700 patients were selected. We used the number of deaths at 3, 6, 9, 12, and 18 months as the endpoints because we were unable to obtain all the individual data. Our analysis showed a reduction in mortality during the first 6 months with polychemotherapy. Although small, this increase in survival, together with an improved quality of life, suggests that polychemotherapy should be recommended for patients with non-resectable non small cell lung cancer.

Tight blood glucose control has been speculated to reduce late complications in insulin-dependent diabetics but results from individual studies have been inconsistent. We have done a meta-analysis of sixteen randomised trials of intensive therapy to estimate its impact on the progression of diabetic retinopathy and nephropathy and the risks of severe side-effects. In the intensive therapy group, the risk of retinopathy progression was insignificantly higher after 6-12 months of intensive control (odds ratio [OR] 2.11). After more than two years of intensive therapy the risk of retinopathy progression was lower (OR 0.49 [95% confidence interval 0.28-0.85], p = 0.011). The risk of nephropathy progression was also decreased significantly (OR 0.34 [0.20-0.58], p < 0.001). The incidence of severe hypoglycaemia increased by 9.1 episodes per 100 person-years (95% Cl -1.4 to +19.6) in the intensively treated patients. The incidence of diabetic ketoacidosis increased by 12.6 episodes per 100 person-years (95% Cl, 8.7-16.5) in the patients on continuous subcutaneous insulin infusion. Long-term intensive blood glucose control significantly reduces the risk of diabetic retinopathy and nephropathy progression but long-term continuous subcutaneous insulin infusion was associated with an increased incidence of diabetic ketoacidosis, and intensive therapy may cause more severe hypoglycaemic reactions.

To appraise the effectiveness of the pneumonia case-management strategy in improving child survival, we have done a meta-analysis of six published intervention trials. The results of a seventh published study and two unpublished studies were also reviewed. The six published studies satisfied our criteria for methodological soundness. The reduction in mortality rate (control group minus intervention group) was estimated for each study, and for all the studies together. For total infant mortality, the overall reduction was 15.9 (95% confidence interval 10.6-21.1) deaths per 1000 livebirths; infant mortality due to acute lower respiratory infection was reduced by 10.7 (4.8-16.7) deaths/1000 livebirths. Mortality among children under 5 years was decreased by 36 deaths/1000 livebirths. The pooled estimates of relative risk are consistent with a 20% reduction in infant mortality and a 25% reduction in under-5 mortality. There was no clear association across the studies between the effect of the pneumonia case-management and extent of co-interventions such as immunisation and oral rehydration therapy. The consistency of findings of all the studies, despite differences in design and methods, shows that the case-management strategy has a substantial effect on infant and under-5 mortality, at least in settings with infant mortality rates of 90/1000 livebirths or more. It is important to find out the most efficient ways of implementing this strategy and integrating it into primary health care.

Low-molecular-weight heparins (LMWHs) have theoretical advantages over standard heparin as postoperative thromboprophylactic agents. We conducted a meta-analysis of studies reported between 1984 and April, 1991, in which LMWHs were compared with standard heparin for postoperative prophylaxis. We included only randomised studies (reported in English, French, or German) in which investigators compared currently recommended doses of the agents and used adequate screening techniques for deep vein thrombosis. For all surgical studies the relative risk (LMWH versus standard heparin) for deep vein thrombosis was 0.74 (95% Cl 0.65-0.86), for pulmonary embolism 0.43 (95% Cl 0.26-0.72), and for major bleeding 0.98 (95% Cl 0.69-1.40). Comparable relative risks were observed for the general and orthopaedic surgery studies separately. When the analysis for the general surgery studies was limited to those of strong methodology, assessed by eight criteria defined in advance, the benefit/risk ratio was less favourable--relative risk for deep vein thrombosis 0.91 (95% Cl 0.68-1.23), for major bleeding 1.32 (95% Cl 0.69-2.56). There is at present no convincing evidence that in general surgery patients LMWHs, compared with standard heparin, generate a clinically important improvement in the benefit to risk ratio. However, LMWHs may be preferable for orthopaedic surgery patients, in view of the larger absolute risk reduction for venous thrombosis.

The risk of pelvic inflammatory disease (PID) associated with use of an intrauterine device (IUD) has been an important concern that has dominated decisions on its use throughout the world, especially in the USA. Early research that suggested such an association led to both a dramatic decline in use of the method and its withdrawal from the US market by two manufacturers. However, factors other than use of an IUD are now thought to be major determinants of PID risk. To address these concerns, we have reviewed the World Health Organisation's IUD clinical trial data to explore the incidence and patterns of PID risk with use of an IUD. The overall rate of PID among 22,908 IUD insertions and during 51,399 woman-years of follow-up was 1.6 cases per 1000 woman-years of use. After adjustment for confounding factors, PID risk was more than six times higher during the 20 days after insertion than during later times (unadjusted rates, 9.7 vs 1.4 per 1000 woman-years, respectively); the risk was low and constant for up to eight years of follow-up. Rates varied according to geographical area (highest in Africa and lowest in China) and were inversely associated with age. PID rates were lower among women who had IUDs inserted more recently. Our findings indicate that PID among IUD users is most strongly related to the insertion process and to background risk of sexually transmissible disease. PID is an infrequent event beyond the first 20 days after insertion. Because of this increased risk with insertion, IUDs should be left in place up to their maximum lifespan and should not routinely be replaced earlier, provided there are no contraindications to continued use and the woman wishes to continue with the device.