Last Observation Carried Forward (LOCF) analysis carries forward the results of all patients who dropped out of the trial to every subsequent data collection point until the end of the study. In contrast, Observed Cases (OC) analyses only include the results collected from patients still participating in the study at the given data collection point. It is important to be aware of these contrasting analysis techniques, as they can introduce different biases to the results, depending upon the type of data being studied.

For example, in a progressive disorder such as AD, if a treatment merely slows the rate of deterioration rather than actually generating improvement, LOCF analysis can indicate an arti?cially favourable treatment outcome. This is because patients who prematurely discontinue will appear to have less deterioration, and therefore the number/timing of discontinuations will contribute to the outcome, potentially creating a bias towards one of the treatment arms. Therefore, for clinical studies of patients with AD, European guidelines promote the use of OC analysis in an endeavour to minimise or eliminate such bias.²⁴

As a result, to provide a complete picture in AD studies, both LOCF and OC analyses are often presented.

A conventional meta-analysis creates an overall result from two or more independent studies by statistically combining standardised summary statistics from each individual study. The methodology takes into account the statistical strength of each study dataset and, if the number of included studies is large, it is likely to provide a test with more power than is provided by the individual studies themselves. An advantage of the meta-analysis technique is that, while providing an overall result, the outcomes from individual studies are also re?ected. The main dif?culties with this methodology lie in selecting the relevant and comparable studies to be included and in recognising the population to which the overall result actually applies.

A pooled analysis is a type of meta-analysis in which individual patient data from two or more studies are pooled to form one large data set. Therefore, in a pooled analysis, the unit of observation is the individual patient, whereas in the conventional meta-analysis it is the individual study. So, while pooling data provides an overall outcome, the results of individual studies can not be interpreted. As for the conventional meta-analysis, the main dif?culties in pooling data lie in selecting appropriate studies for inclusion and in recognising the population to which the overall result applies. However, the selection of appropriate studies for a pooled analysis is more restricted, as studies must share more design similarities. For example, pooling of data can only be done if the individual studies have used the same outcome scale at the same assessment time points.

When compared to a conventional meta-analysis, a pooled analysis will usually provide more statistical power, even when the number of included studies is small.

 

末次观察结果转接分析将试验中所有脱落的受试者最后一次观察结果转接到随后每一个数据采集点，直到研究结束。相反，可获得数据分析仅仅包括在研究指定的数据采集点仍然参加研究的患者的结果，重要的是应该意识到取决于研究数据的类型，采用这些对比分析方法分析试验结果可以得出偏向不同组别的结果。

例如，对某一进行性疾病，比如阿尔茨海默病，如果治疗仅仅延缓恶化程度而不是真正能改善患者病情，LOCF分析可能会产生错误的有利治疗结果。这是因为患者提前中断研究通过末次观察结果转接后数据将显示患者病情未恶化，因而中断研究的时间将促成这一结果，潜在地导致对一个治疗组的偏倚。因此，对于阿尔茨海默病患者的临床研究，欧洲指南鼓励采用OC分析，以尽量减少或者排除上述偏倚。²⁴

为了提供AD研究的全景图，一并呈现了LOCF分析及OC分析。

传统的Meta分析方法通过联合对单一研究的标准化汇总统计，从两个或者更多的独立研究中得出一个总的结果。该统计方法考虑了每个研究数据集的统计强度，如果所包含的研究数较大，与单个研究相比，Meta分析具有更强的检验效能。Meta分析的一个优势是，在得出一个总结果的同时，单个研究的结果也会得到反映。Meta分析的主要困难在于选择被纳入的相关及可比较的研究，并确定总结果真正适用的群体。

汇总分析是Meta分析的一种，在该分析中，两个或者多个研究中单个患者的数据被汇总形成一个大数据集。因此，在汇总分析中，观察单位是单个患者，而在传统的Meta分析中，观察单位为单个研究。因此，汇总分析在得出总结果的同时，不能对单个研究的结果进行解释。对于传统的Meta分析，收集数据的主要困难在于选择被纳入的研究，并确定总结果所适用的人群。然而，选择合适的单个研究进行汇总分析更受限，因为研究必须具有更多的设计相似性。例如，如果单个研究在相同的评价时间点使用了相同的结果标尺，才可以对数据进行汇总。

与传统的Meta分析相比，汇总分析通常具有更高的统计效能，即便是在所包含的研究数较少的情形下。

 

原件下载：