The EMEA has re-posted points to consider on missing data, this points to consider was formally adopted in 2001, however the EMEA has chosen to re post this on the website. It does not appear to have changed since its last posting.
The EMEA a considered missing data as a potential source of bias when analysing clinical trials, interpretation of the results of a trial is always problematic when the number of missing values is substantial. There are many possible sources of missing data, affecting either complete subjects or specific items, missing data violate the strict Intend To Treat principals: measurement of patient outcomes regardless of protocol adherence and analysis performed by treatment assigned, regardless of which treatment patients actually received. If missing values are handled simply by excluding any patients with missing outcomes from analysis, the following problems may affect the interpretation of the trial results.
The sample size and variability of outcomes affects the power of the clinical trial, power is greater the larger sample size and smaller variability. The reduction in the number of cases available for analysis, completeness of data add ot the resulting reduction of the statistical power.
Bias is the most important concern resulting from the missing data may affect: Designation of the treatment effect, The comparability of the treatment groups, The representativeness of the study sample in relation to the target population. Bias occurs in the estimation of the treatment effect when the relationship between missing this treatment outcomes exists. In most cases it is difficult or impossible to elucidate whether the relationship between missing values and unobserved outcome variable is completely absent. Thus it is sensible to adopt a conservative approach, considering missing values as potential sources of bias.
A possible way of handling incomplete data is to ignore them and perform statistical analysis with complete data only. However, complete case analysis violates intention-to-treat principal. More importantly it is subject to bias, and thus cannot become recommended as the primary analysis confirmatory trial.
The statistical analysis of the clinical trial requires imputation of values to those data that have not been recorded. Many techniques have been used for the imputation of missing data, but none of them can be considered as the gold standard in every situation. The guidance goes on to discuss the many options available:
To cope with situations where response collection is interrupted at one point, the widely used method is last observation carried forward. This method is likely to be acceptable if measurements are expected to be relatively consistent over time.
Best worst case imputation, assigning the worst possible value of the outcome to dropouts are a negative reason (treatment failed) and the best possible value to positive dropouts (kills), is another approach that can be considered, provided it is applied conservatively.
Another simple approach of inputting missing data is to replace the unobserved measurements by values derived from other sources. Possible sources include information from the same subject, from other subject of similar baseline characteristics, the predictive value from an empirically developed model, historical data, etc.
Most methods faced the risk of bias in the standard error downwards by estimating central value and ignoring its uncertainty. This risk can be avoided by some techniques based upon maximum likelihood methodology and with multiple imputation methods. Maximum likelihood methodologies have been proposed that imputation of missing values, as have multiple imputation methods. Maximum likelihood method strategies fit the model by an iterative process. Multiple input methods generate multiple copies of the original dataset replacing missing values by randomly generated values, and analysing is complete sets.
Unfortunately, there is no universally accepted methodological approach and the missing values.the best process of all is the avoidance of missing data in the first place.
If you would like more detail in this area please get in touch with Damien Bové damien.bove@idaconsultants.com
Damien Bové works as a drug development consultant (pharmaceutical or biotechnology) and regulatory consultant, we work with our clients to define a drug development target, define a drug development strategy, define a regulatory strategy or define a commercial strategy. Our clients are generally raising funds or looking to license out their technology and we help them achieve it. If you want to know more don’t hesitate to get in touch.
As you know this website is a great resourse for keeping up to date with developments and regulations, why not get our FREE monthly regulatory and market round up. You can un-subscribe at any time and we do not share your details with anybody.
