MacMahon S, Collins R
Institute for International Health, University of Sydney, New South Wales, Australia.
Lancet. 2001 Feb 10;357(9254):455-62. doi: 10.1016/S0140-6736(00)04017-4.
Observational studies and randomised trials can contribute complementary evidence about the effects of treatment on mortality and on major non-fatal outcomes. In particular, observational studies have an important role in the identification of large adverse effects of treatment on infrequent outcomes (ie, rare, but serious, side-effects) that are not likely to be related to the indications for (or contraindications to) the treatment of interest. Such studies can also provide useful information about the risks of death and disability in particular circumstances that can help to generalise from clinical trials to clinical practice. But, due to their inherent potential for moderate or large biases, observational studies have little role in the direct assessment of any moderate effects of treatment on major outcomes that might exist. Instead, sufficiently large-scale evidence from randomised trials is needed to assess such treatment effects appropriately reliably. Wider appreciation of the different strengths and weaknesses of these two types of epidemiological study should increase the likelihood that the most reliable evidence available informs decisions about the treatments doctors use--and patients receive--for the management of a wide range of life-threatening conditions.
观察性研究和随机试验可以为治疗对死亡率和主要非致命结局的影响提供补充证据。特别是，观察性研究在识别治疗对罕见结局（即罕见但严重的副作用）的重大不良影响方面具有重要作用，这些不良影响不太可能与所关注治疗的适应症（或禁忌症）相关。此类研究还可以提供有关特定情况下死亡和残疾风险的有用信息，有助于从临床试验推广到临床实践。但是，由于观察性研究本身存在中度或较大偏倚的可能性，它们在直接评估治疗对可能存在的主要结局的任何中度影响方面作用不大。相反，需要来自随机试验的足够大规模证据来可靠地适当评估此类治疗效果。更广泛地认识这两种类型的流行病学研究的不同优势和劣势，应该会增加最可靠的现有证据为医生用于治疗——以及患者接受治疗——以管理各种危及生命的疾病的决策提供信息的可能性。
