Individual patient data supplied by investigators

With this approach, trialists are asked to provide individual details on all patients included in their trial. The minimum data that can be collected are the patient identifier,

Box 11.6 Example pages from data extraction form [46]

treatment allocated and outcome, together with the date of randomization and date of outcome, if time-to-event is to be calculated. However, it is often important to collect additional baseline or prognostic variables. As well as providing the opportunity for sub-group analyses and a richer understanding of the trials, these covariates are also extremely useful in checking the integrity of the randomization process. In practical terms there is likely to be a trade-off between requesting a minimal dataset with least inconvenience for trialists and obtaining a wide range of variables for further analysis at the risk of discouraging participation in the meta-analysis. However, most trialists (or their data centres) are usually able to supply information on the most common prognostic variables with little additional effort.

Those conducting an IPD meta-analysis should be prepared to accept data in a variety of high-tech and low-tech formats, ranging from electronic transfer of databases to photocopies of anonymized patient notes. For most trial organizations, recent trials will be supplied on floppy disk or sent by e-mail. However, for old or archived trials it may be easiest for trialists to complete paper data collection forms (Fig. 11.5).

Patient identifier

Date ot birth dd/mm/yyyy or aqe

Sex

Tumour stage

Histology

Pert status

Treatment arm

Randomised dd/mm/yyyy

Survival status

Death/ Last foflow-up dd/mm/yyyy

Cause of death

ool

b*

u

s

1

1

is/os/mi

D

1

ooZ

M

\ll

2.

1

■x.

a2|os)i<m

D

o?lo3/t113

l.

OOS

M

U|

3

2

^

ot>/oi>/i<m

D

n<\z

3

OOif

tz.

M

u

%

1

1

2S]o(>)m2-

A

3<5|u\iOOI

OOS"

bl

F

VM

•3

1

2

\7/o?jmz

D

1

oob

>1

M

HI

■=1

1.

2.

Z,/e>1IM2

D

q

oo?

t>*&

P

HI

2.

Z

1

iwlos/mz

D

Vt IwjW'iif

1

Fig. 11.5 Example data form for collecting IPD.

In our experience, an increasing amount of data is being sent by e-mail (Fig. 11.6), which reduces considerably the amount of resource required for data entry and checking, and makes communicating and querying data considerably easier. Every effort should be made to reduce the burden on the trialist or data centre providing the information.

Fig. 11.6 Method of supplying data for (a) a meta-analysis initiated 1989 [23,47] and (b) a meta-analysis initiated 1994 [36].
0 0

Post a comment