DD0123 Variable/Value Level Length mismatch

Pinnacle 21 Support Define.xml
1

Could you give an example of a “format-12.7” value?
Is it something that you can have “1234.1” but also “0.1234567” ?
I.e. you never have the case that the “total” has more than 9 charcters?
Personally, I think that rule DD0123 (corresponding to CDISC rule 142) does not take care of this situation.

You could set the Length in the define.xml at the variable level (LBSTRESC) to 12, but then it may be that P21 complains about the value is not corresponding to the length in the XPT file, which is 9 (as FDA does not want you to “spoil” bytes due to this miserable format they enforced to us - even CSV would have been a better choice …).
Once another reason we should get rid of XPT format, better yesterday than tomorrow …

Essentially, we would not need to set “Length” for text variables. This idea comes from the time that relational databases had fixed lengths (i.e. “CHAR(N)”). It could however be used to set the value using VARCHAR(N) from it (that’s why I use it for). However, as far as I know, FDA does not use databases to store submissions in, they just use the XPT files “as is”.
Also, it is pretty ridiculous in modern times that numeric values must also be put in --STRESC, as they are already in --STRESN.

The best thing you can indeed do is to explain the occurrence of DD0123 in the reviewers guide.

2

Hi Jozef,

Yes, that’s exactly your example : sometimes ‘1234.1’, sometimes ‘0.1234567’.

I will just document it in the SDRG : I do not see a way to avoid it, except VLM for each parameter, just for this reason, which would be a waste.

Cheers

– Thierry

3

Hello,

I come through a figure case for this define rule, and I would welcome any suggestion.

Here is the situation: I have the variable LBSTRESC declared globally as text, length = 9.

  • where LBTESTCD = HCG, I have VLM text, length = 8 and codelist = POSITIVE/NEGATIVE.

  • where LBTESTCD ne HCG, I have VLM float, length = 12.7, hence the message since “how can 12 digits fit into 9 characters?”

The answer is that inside the numeric values, the longest integer part is 4 digits. The longest decimal part is 7 digits. But not on the same parameters. So globally it can fit in 9 characters, but to print any number, we need the format 12.7 (4 integer + 7 decimal + 1 decimal point).

I do not want to create VLM for each parameter, as it is of little interest.

For the moment, I explain the above in the SDRG, but does anyone have a better idea?

P21, could you check by value, in the data, instead of checking in the define?

Cheers

– Thierry