Specifications for tablet weight

Q: I recently read about an FDA observation of a company that, when doing in-process control of its tablet press, weighs 10 tablets at certain intervals and records/plots the average of those 10 tablets. Thus, the company does not record individual tablet weights, which means that it does not know whether any individual tablets are out-of-specification. If an NDA and product specs do not contain tablet weight, then is this a valid FDA or other auditor observation of an in-process control measurement?

A: Fred Rowley responds:

It is probably not possible to have an NDA or even an ANDA approved today without a specification for target weight and range for any given tablet. Even assuming that the product is "grandfathered," the FDA now asks firms to submit a file and that file must have a specification for target and range.

From a compliance standpoint, simply filing a spec for target weight with a range does not relieve the firm from assuring the Agency and the public that all tablets from a batch are within this range. This is usually finalized at product validation.

Going forward, the FDA inspectors simply follow the inspector's guide. This guide is public and available online. One of the key items inspectors watch for is the item mentioned. Expressed differently: You might ask, "How does the firm assure the public and itself that each and every one of the tablets in a batch meets at least the range established?"

The answer to this is complex. With a simple tablet press, one might accomplish this goal through a combination of validation and SOPs together with a set of individual product action limits. This is where the knowledge and application of the CPK index and SPC in general will statistically defend control limits stated on a batch record.

Of course, a computer does this very thing, depending upon the specific set of "rules" the firm places in its brain. Many tablet presses have three limits:

1. Individual weight variation inside limit one (say, +/- 0.5% of target): Take no action.
   
   
2. Weight variation outside limit one but inside limit two (say, between +/- 0.5% and 3%): Adjust the weight and monitor for effect.
   
   
3. Weight variation outside limit two but inside limit three (say, outside 3% but inside 5% of target): Continue to adjust to the target weight but reject all tablets you find that are outside this limit.
   
   
4. Weight variation outside limit three: Reject all tablets and shut down the press.

Other items for the computer brain:

• How often to sample?
  
  
• What quantity to sample?
  
  
• How good is the ability to detect and reject?
  
  
• Does the reject mechanism keep up with the computer?
  
  
• How many reject tablets to tolerate before I shut down?

Bottom line: You cannot depend upon weighing an aggregate of tablets and accept the sample if the mean value is within the established range.