CRC 2000 - colorectal cancer - data structure and protocols

Contents List

• CRC 2000 - colorectal cancer
• Master list of all centres/groups, trials/strata and treatments (intranet only at present)
• CRC 2000 data preparation: brief written protocol
• Specification of CRC 2000 'green form' format
• CRC 2000 data form rubric
• CRC 2000 data preparation diktats
• Analysis details: endpoints, steering files and strata for CRC 2000

CRC 2000 data preparation: brief written protocol

All patient records are converted into 'green form' format (described below) if not already supplied in it. Results received as tables are converted into sets of synthetic 'green form' records. The following routine checks (where appropriate) are performed on every 'green form' compilation:

• Duplicate patient entries
• Patient identifier missing
• Randomisation date missing
• Treatment allocation missing
• Surgery date missing
• Tumour site missing
• Tumour stage missing
• Gender missing
• Randomisation age missing
• Recurrence date missing
• Recurrence type missing
• Survival status missing
• Death date missing
• Randomisation date wrong, before 1945 or out of range
• Surgery date wrong or out of range
• Recurrence date wrong or out of range
• Last follow-up or death date wrong or out of range
• Treatment allocation code unknown
• Tumour site code unknown
• Gender code unknown
• Randomisation age not in range 20-98
• Recurrence type code unknown
• Survival status code unknown
• Tumour stage incompatible with metastatic disease status
• Recurrence flag error
• Recurrence type given without event
• Cause of death given when alive
• Died of colorectal cancer without recurrence
• Died of cause other than colorectal cancer but with recurrence

If an event such as the recurrence of disease is reported at a date later than the quoted last follow-up date, the last follow-up date is automatically changed to the later date. The completeness of follow-up is then calculated for the end of each calendar year. The distributions of randomisation dates, randomisation ages and time elapsed since last follow-up are checked for any significant imbalance between treatment groups in two ways as follows. Firstly, a t-test is applied to the difference between the mean value of each distribution for patients in each group with the corresponding mean for patients in the remainder. Secondly, an F-ratio is calculated for each distribution by comparing the variance between the groups with the variance within the groups. The distribution of time elapsed since last follow-up is also checked in these two ways for any significant imbalance between those patients with and those patients without a recorded recurrence of disease. Finally, the distribution of time elapsed since last follow-up is checked in the same two ways for any significant imbalance between patients in two categories of tumour site (colon; rectum), two categories of tumour stage (A/B; C/D) and two categories of gender (male; female).

Where patient serial numbers form an obvious sequence it is checked for missing numbers.

A tabulated breakdown of variables is produced for each trial, together (where relevant) with lists of patients in 'problematical' categories such as those with lapsed follow-up, uncertain death cause or second malignancy site. Graphs of accrual date and the proportion of living patients still on follow-up as a function of time from randomisation by treatment allocation are also produced, together with Kaplan-Meier life-table curves. Before trial data are finally incorporated into the overview, the analyses described above are sent to the participating trialist(s) for checking and approval.

Contact

Vaughan Evans
CCC Secretariat
C.T.S.U.
Radcliffe Infirmary
Oxford OX2 6HE
England
Tel. U.K.(44)-Oxford(1865)-557241; FAX: U.K.(44)-Oxford(1865)-558817

Specification of CRC 2000 'green form' format

CRC 2000 data form rubric

GUARANTEE OF CONFIDENTIALITY OF DATA

ANY INFORMATION PROVIDED OVERLEAF TO THE CCC/LIMAG SECRETARIAT WILL BE HELD SECURELY AND IN STRICT CONFIDENCE.

NOTES ON FORMAT OF DATA REQUESTED OVERLEAF

Special coding conventions

Please accompany these forms by an explanatory letter about any special coding conventions (e.g. on tumour site, tumour staging or cause of death) you have used, plus notes on any special features of the study(s) to which you wish to draw attention.

Dates that are not (or not yet) known exactly

either leave DAY blank, and give (approximate or provisional) month and year;
or leave DAY and MONTH blank, and just give approximate year.

BASELINE DATA

Patient identifier

Any convenient convention you wish, in case any correspondence becomes necessary. (If reporting several trials, please try to use a system that implicitly specifies both the trial and the patient.)

Date randomised

Please describe ALL patients EVER randomised, including even lost, ineligible or withdrawn patients, and ignore all non-randomised patients.

Trt. gp. allocated

Treatment group number: 1 or 2 only, for 2-group trials, or a wider range for trials with more arms, as defined by you at the top of the form. N.B: even if, in reality, some quite different (or even opposite!) treatment was inadvertently given, what is wanted is the originally-allocated treatment. (For patients erroneously entered more than once, give only the first allocation.)

Date of surgery

See note above on approximate dates.

Tumour site

0 = unspecified; 1 = colon; 2 = rectum; 3 = colon and rectum. If you prefer to use your own classification of tumour site (e.g. in order to code sigmoid tumours separately) please do so, and send us details of it.

Tumour stage

Please use your own classification and send us details of it, or use the Dukes classification (A = lesion confined to muscularis propria; B = lesion extends through muscularis propria with negative nodes; C = positive nodes), or any other standard system (e.g. Astler-Coller modification, TNM etc). Extra codes: D = metastatic disease; X = benign tumour (e.g. adenoma); and Y = inoperable disease.

Gender

1 = male; 2 = female.

Entry age

Age at randomisation.

FOLLOW-UP DATA

Recur?

Any recurrence? 1 = none recorded; 2 = some recurrence (local or distant or both).

Approx. date of 1st recurrence

Date of first recurrence: see note above on approximate dates.

Site of 1st recurrence

CCC version:

0 = unknown; 1 = local only; 2 = local and distant; 3 = distant only.

LIMAG version:

0 = unknown; 1 = local only; 2 = liver (± other site); 3 = distant (± local) but not liver.

Warning: recurrence sites received in either of these scenarios are re-coded into the 'new' internal system.

Dead/other

1 = alive when last traced; 2 = known to be dead; 3 = lost despite extensive inquiries, but alive when last traced.

Date died/last traced

Date of death, or date last known to be alive: see note above on approximate dates.

Death cause

If the patient died without reported recurrence, give underlying cause of death. Either state the cause in words, use an ICD code or use your own classification and send us details of it.

CRC 2000 data preparation diktats

1 Local spread found at surgery

Nothing special, unless excision is incomplete.

2 Incomplete excision / residual tumour after surgery

Set recurrence = 'Yes', set recurrence date = surgery date, set recurrence type = 'local' (or as appropriate, if other site(s) already given). If stage is 'A', 'B', 'C' or unknown, set it to 'Y?' and if 'A', 'B' or 'C', check the change with the trialist.

(N.B: formerly, known stages were not changed and unknown stage was set to 'D?'; later, 'Y' was used instead of 'Y?').

3 Local recurrence at surgery

Set stage to 'Y?' and if 'A', 'B' or 'C', check the change with the trialist.

(N.B: formerly, known stages were not changed and unknown stage was set to set to 'D?')

4 Local recurrence reported on days 1-30 after surgery

Put on hold and refer case to RG.

(N.B: formerly, procedure adopted was as 'local recurrence at surgery'; later, scope was reduced to days 1-5 after surgery)

5 Metastases at surgery

Set recurrence = 'Yes', set recurrence date = surgery date (if not given), set recurrence type = 'distant' (or as appropriate, if other site(s) already given). If stage is 'A', 'B', 'C' or unknown, set it to 'D?' and if 'A', 'B' or 'C', check the change with the trialist.

(N.B: formerly, known stages were not changed and unknown stage was set to set to 'D?'; later, only the alterations from 'A' and 'B' were checked with the trialist)

6 Metastases reported on days 1-30 after surgery

Put on hold and refer case to RG.

(N.B: formerly, procedure adopted was as 'metastases at surgery'; later, scope was reduced to days 1-5 after surgery)

7 Recurrence at unknown site at surgery

If stage is 'A', 'B', 'C' or unknown, set it to 'D?' and if 'A', 'B' or 'C', check the change with the trialist.

8 Recurrence at unknown site reported on days 1-30 after surgery

Put on hold and refer case to RG.

9 Conversion into Dukes system

N 'unknown' to be treated as 'N0' (RG)

Note on simultaneity of events

Events within 30 days are taken as simultaneous, e.g. 'local recurrence' followed by 'hepatic metastases' 15 days later is counted as 'local + distant including liver'.

Note on coding of relapse sites

There are up to 65 logical possibilities, depending on coding in the data presented. These have been collapsed to the 12 categories listed in the current form definition.

[Return to the Contents List]

[End of document, updated to 29 November 2000]