Individual patient data

The main aims of data checking procedures are to ensure the accuracy of data, integrity of randomization and completeness of follow up. For any one trial the results of all the data checks should be considered together to build up an overall picture of that trial and any associated problems. Where there are concerns about the data supplied, these should be brought to the attention of the trialist and sympathetic efforts made to resolve them.

Range and consistency checks should be carried out for all data irrespective of whether they were supplied electronically or were entered manually into the meta-analysis database (in which case it is important to audit the data entry process). Any missing data, obvious errors, inconsistencies between variables or extreme values should be queried and rectified as necessary. If details of the trial have been published, these also should be checked against the raw data and any inconsistencies similarly queried. All of the changes made to the data originally supplied by the trialists, and the reasons for these changes, should be recorded.

Sun Mon Tue Wed Thu Fri Sat

Fig. 11.7 Example of pattern of randomization anticipated for non-acute conditions. The day of week can be calculated from a date by most standard database packages.

Sun Mon Tue Wed Thu Fri Sat

Fig. 11.7 Example of pattern of randomization anticipated for non-acute conditions. The day of week can be calculated from a date by most standard database packages.

As part of a series of investigations to check the validity of the treatment assignment process (i.e. that the trial is in fact randomized), the distribution of patient-related variables can be checked for balance across treatment arms and across major baseline characteristics. This can be done using a chi-square test. It is, however, important to remember that imbalances may occur by chance alone especially for non-stratified variables and when trials are small. Other checks that can be done include looking at the weekday of randomization (Fig. 11.7). For example, for UK cancer trials we would expect very few randomizations at the weekend. In studies from other countries it is important to appreciate cultural differences in working patterns. Similarly, randomizations in trials of acute disease would be expected on all days of the week.

The pattern of randomization can also be checked by producing simple plots of cumulative accrual. Figure 11.8 illustrates this for a trial with a 1:1 assignment carried out by minimization, showing the numbers allocated to the two treatments to be close throughout and crossing frequently. Figure 11.9 (which has been made public with the trialists' permission [13]) illustrates how illuminating such plots can be. The curve is from an unpublished trial of radiotherapy versus chemotherapy in multiple myeloma.

Here we see that from the start of the trial until early 1985 the pattern is similar to that of Fig. 11.8. However, in the middle of 1985 the curves diverge and remain apart - the cumulative accrual to the chemotherapy arm continues to rise, whereas the radiotherapy arm remains flat for a period. On further enquiry, it transpired that in this trial the radiotherapy equipment was unavailable for six months during the trial, but during this time patients continued to enter the chemotherapy arm. It was only when the individual patient data were provided for a meta-analysis that this problem was brought to the attention of the trialist who agreed that the appropriate solution was to exclude this small number of non-randomized chemotherapy patients from the analysis.

Where survival (or another time-dependent variable) is the primary outcome it maybe important that trial follow-up is as up-to-date as possible since an increased follow-up

200-1

200-1

Date of randomization

Fig. 11.8 Example plot of cumulative numbers of patients randomized to two treatment arms.

Date of randomization

Fig. 11.8 Example plot of cumulative numbers of patients randomized to two treatment arms.

Fig. 11.9 Cumulative randomization plot for a trial with a non-random phase during recruitment. Reproduced with permission from [13].

may see a reduction in treatment effect if the survival curves are converging, or an increased treatment effect if the curves are diverging. Thus, where appropriate, data should be checked to ensure that follow-up is up-to-date and that it is balanced across treatment arms. Balance in follow-up can be checked by selecting all patients outcomefree and using the date of censoring as the event to carry out a 'reverse survival' analysis

(see Section 9.3.4). This produces censoring curves, which should be the same for all arms of the trial. Any imbalance should be brought to the attention of the trialist and updated information should be sought.

As a final stage of checking, each trial should be analysed individually and the trialist sent a copy of the analyses and tables of patient characteristics, together with a printout of their data as included in the meta-analysis database. This allows the trialist to verify that the data being used from their trial are indeed correct.

0 0

Post a comment