Include all and only randomized patients

All individuals who were randomized in a trial should be included in the analysis, and any patients who were not randomized should be excluded. Failure to do this could result in unpredictable bias. Unfortunately, not all trials follow this policy when reporting their results, and patients can be excluded (sometimes covertly) from published analysis for a number of reasons. Such patient exclusions, if related to treatment, could

Table 11.4 Sensitivity analysis - excluding certain pre-specified types of individual - from an IPD meta-analysis in soft tissue sarcoma. Modified and reproduced with permission from [36]

Patients included

Deaths / Patients

Hazard ratio

p-value

All patients

709/ 1 568

0.90

0.157

All except patients:

Less than 15 years

With metastatic disease

That received induction chemotherapy

691 /1544

0.89

0.121

All except patients:

With locally recurrent disease

Less than fifteen years

With metastatic disease

That received induction chemotherapy

597 / 1366

0.91

0.278

seriously bias the results, for example, if patients are excluded because they are unable to tolerate the allocated therapy or follow the treatment schedule. An empirical study evaluating the impact of patient exclusion using information from several cancer IPD meta-analyses, comprising data from over 100 RCTs and thousands of patients found that meta-analyses that maintained the investigators' exclusions were more likely to favour the investigational treatment than those that analysed all randomized patients [35].

As IPD meta-analyses, or those collecting summary data from source, do not rely on the trialists' original analyses, trialists should be asked to provide data on all patients. Any randomized patient excluded from the original trial analyses can then be reinstated in the meta-analysis and any non-randomized patients can be removed from the dataset (in some cases non-randomized patients maybe included in published analyses, for example, by including patients from a non-randomized pilot phase of the trial). Although not all trials may keep records of excluded patients, experience in the cancer field has been good and it has been possible to recover information from the majority of excluded patients. For example, in an IPD meta-analysis in soft tissue sarcoma [36], data were recovered on 99 per cent of the 344 patients that had been excluded from the investigators' own analyses. In a more recent meta-analysis in high-grade glioma data from 83 per cent of the 253 patients that had been excluded from the original analyses were recovered [37].

There can be good clinical reasons for excluding certain types of individuals from analyses. However, to be unbiased, any exclusions should be pre-specified and applied objectively and uniformly across trials. Ideally, their impact should be assessed by sensitivity analyses (including and excluding patients to determine whether it influences the estimated treatment effect) (see Section 11.5.7). In the soft tissue sarcoma IPD meta-analysis [36], in such pre-specified sensitivity analyses, exclusions made no appreciable impact on the estimated effect of chemotherapy (Table 11.4). A further advantage of having done these analyses is that it re-assures us as to the robustness of the results.

10 Ways To Fight Off Cancer

10 Ways To Fight Off Cancer

Learning About 10 Ways Fight Off Cancer Can Have Amazing Benefits For Your Life The Best Tips On How To Keep This Killer At Bay Discovering that you or a loved one has cancer can be utterly terrifying. All the same, once you comprehend the causes of cancer and learn how to reverse those causes, you or your loved one may have more than a fighting chance of beating out cancer.

Get My Free Ebook


Post a comment