Computer printout versus operator recording data in field

COMPUTER SYSTEM VALIDATION (CSV)
#1

Hi,

I wanted to get the forums opinion on using ‘computer generated trend data report from a SCADA’ (which has been fully validated) for a given equipment, versus recording data in person in the field?

Scenario

We have a bioreactor skid connected to a computer system that logs data every one (1) minute on Temperature, DO, agitation, gas flow and pressure over time. This is the design of the system. My colleague developed a test script for the OQ on bioreactor that requires the executor to record temperature data every 30secs for a duration of 10mins.

My argument is that we do not need to record this data every 30 sec for a duration of 10 mins, since we already have a validated computer system. I plan to use the trend data from the computer system which logs data every 1 minute but over a longer duration in place of an executor recording data in the field. However people are uncomfortable because they do not think it is appropriate to just use the validated computer printout.

I have challenged them on why they think it is inappropriate and I get answers such as:

  1.   FDA will always like to see a handwritten data (I disagree because 21 CFR part 11 came out for a reason) 

  2.   They will like to see what happens between the one minute when computer is not logging data (I reminded them that we have a life cycle approach on this project, which also include but not limited to commissioning, tunning etc. In addition we have alarms that have been validated on this system, so if temp excursion is an issue it will trigger an alarm anyway. Also if what happens in between the one minute is an issue then we should change our requirements and spec to sample data at shorter intervals).

So what are your thoughts on this matter? Do you think my ideas are bizzare?

Mr Stephen N. U Nwodo PMP, ASQ CSQE
Member of Biochemical Engineering SG of IChemE
Affiliate Member Institute of Chemical Engineering (IChemE)
Cell phone:909-732-2885 (USA)

#2

[quote=Stephen]Hi,

  1. FDA will always like to see a handwritten data (I disagree because 21 CFR part 11 came out for a reason)

  2. They will like to see what happens between the one minute when computer is not logging data (I reminded them that we have a life cycle approach on this project, which also include but not limited to commissioning, tunning etc. In addition we have alarms that have been validated on this system, so if temp excursion is an issue it will trigger an alarm anyway. Also if what happens in between the one minute is an issue then we should change our requirements and spec to sample data at shorter intervals).

So what are your thoughts on this matter? Do you think my ideas are bizzare?

Mr Stephen N. U Nwodo PMP, ASQ CSQE
Member of Biochemical Engineering SG of IChemE
Affiliate Member Institute of Chemical Engineering (IChemE)
Cell phone:909-732-2885 (USA)[/quote]

Your opinions to me seem very logical and make alot of sense the key point here is that you are using a fully validated computer system, do the people concerned know what this means and it’s validity.

I would think that it’s actually a more comphrehensive test to test over a longer period than to test at more regular intervals, there is more room or something going wrong over a longer period.

If they are worried about the one minute that is not logged, reassure them that your testing is comphrehensive, sometimes when individuals involved don’t understand the process sufficiently they tend to over-analyse, or over-test.

As long as the ‘computer generated trend data report from a SCADA’ is fully validated then in my opinion you are going about this in the right way.

Best of luck!

#3

I would think that using the SCADA data instead of handwritten data that could have errors from collection (writing wrong number, seeing one number and writing another, guessing what the number was, missing the number completely, etc). If the concern is in the interval, shorten the interval of the historian. As long as the alarms delays are not longer than the collection interval, any issue with being out of spec in between should not be an issue.