Process Validation of Low Volume Production of Medical Devices

K

keizer_soze

Hello,

I work at a small medical device company where we produce a bunch of low-volume devices with a lot of operator-dependent processes, and I would appreciate input on some SPC related questions. Mainly, what is the most appropriate method for a) verifying that the design is acceptable and b) validating processes which have critical impact on the product.

As mentioned, since this is a small company, the production is mostly manual labour with processes such as UV gluing.

Our production quantities for a typical product might be 50 pieces a month, so right away there are limited product available for testing.

We control the material that comes in fairly well as we have good specifications, and the equipment is properly maintained and calibrated. Most of the specs are 100% verified, but some cannot be since they are destructive.

With that background, here is a specific example:

A product (say a needle) is designed. The needle's dimensions and performance requirements are pre-determined. Samples are built with R&D techniques using representative processes and materials for design verification, usually 10 samples. Sticking to the UV example, a performance requirement might be tensile strength of X Newtons.

To transfer the product to production, processes are fine-tunes (as much as we can) and a process validation is done (or at least we try to).

The UV process might look like:
1. Spread UV glue around shaft of needle at about 10mm from the end.
2. Attach hub to needle
3. UV glue hub using UV wand for 45s to 60s.
4. Verify under a microscope that there are no gaps in the UV.

That's it!

So, my specific questions are:

1. What is an appropriate method to analyze results from design verification? For example, should we do a 95/95 test? A t-test? Mean - 3 standard dev > minimum? Or just make sure all test pieces pass the criteria? Or something else? Keeping in mind we have limited quantities to spare due to cost constraints.

2. For transferring product to production, our process validation requires at least 3 lots be validated. We perform an IQ, OQ, PQ, but the processes are extremely dependent on the operator. So I don't know how to analyze the test results, or whether the validation is meaningful. Should we do a Cpk analysis of 1.33? Or some other way to ensure the processes are capable for production in the long-run?

Thanks, any advice is kindly appreciated.
 

Miner

Forum Moderator
Leader
Admin
Re: Use of SPC in low-volume production of medical devices

Your questions appear to be more related to process validation of medical devices than to SPC. Is this correct? And are you building to a specific standard or regulation? This will help us to move it to the appropriate medical device forum where you will get more relevant answers.
 
A

arios

Re: Use of SPC in Low Volume Production of Medical Devices

We control the material that comes in fairly well as we have good specifications, and the equipment is properly maintained and calibrated. Most of the specs are 100% verified, but some cannot be since they are destructive.

With that background, here is a specific example:

A product (say a needle) is designed. The needle's dimensions and performance requirements are pre-determined. Samples are built with R&D techniques using representative processes and materials for design verification, usually 10 samples. Sticking to the UV example, a performance requirement might be tensile strength of X Newtons.

To transfer the product to production, processes are fine-tunes (as much as we can) and a process validation is done (or at least we try to).

The UV process might look like:
1. Spread UV glue around shaft of needle at about 10mm from the end.
2. Attach hub to needle
3. UV glue hub using UV wand for 45s to 60s.
4. Verify under a microscope that there are no gaps in the UV.

That's it!

So, my specific questions are:

1. What is an appropriate method to analyze results from design verification? For example, should we do a 95/95 test? A t-test? Mean - 3 standard dev > minimum? Or just make sure all test pieces pass the criteria? Or something else? Keeping in mind we have limited quantities to spare due to cost constraints.

2. For transferring product to production, our process validation requires at least 3 lots be validated. We perform an IQ, OQ, PQ, but the processes are extremely dependent on the operator. So I don't know how to analyze the test results, or whether the validation is meaningful. Should we do a Cpk analysis of 1.33? Or some other way to ensure the processes are capable for production in the long-run?

Thanks, any advice is kindly appreciated.

For your specific question #1, strictly talking about "Design Verification" I would confirm that there is an output for every input, that is the idea about Design Verification. An output is design drawing, a given specification, or a certain conceived feature on the product. Having said this I don't clearly understand why the proposed analysis tools, but I could be wrong.

For your specific question # 2, you could confirm the consistency of the UV adhesive dispensing process. If you use an automatic or a pedal dispenser I would confirm using a capability analysis that the amount dispensed is uniform at its various rates. As for the product itself, you could validate the operation on the basis operator qualification to perform the operation properly, and for this you use attribute sampling for no leaks, no occlusion, or visual checks with a lense, or other tool. The operaion would remain validated provided you use qualified operators to perform it and provided there are no changes. Via IQ I would confirm the intensity of the UV curing process using a light intensity meeter (don't know the name in English..help!) and have a fixture to insure the light is measured always at the same distance.

I would also establish a frequency of change of the ligth bulbes as part of the PM routines developed during the IQ

Hope this helps:2cents:
 
K

keizer_soze

Re: Use of SPC in low-volume production of medical devices

Hello Miner,

I suppose yes, the question is more related to process validation. I always thought both PV and SPC are heavily interlinked, but that could just be my inexperience.

As far as standards are concerned, yes, the tensile requirements are from a standard - I believe ISO 10555-1 is one of them. And of course also to ISO 13485 and FDA QSR. I suppose the question is more related to implementation of ISO 13485 requirements of properly verifying and validating the design and the processes, and how much is considered adequate depending on the "manual-ness" of the assembly procedures and the number of samples available.

Thanks!

Your questions appear to be more related to process validation of medical devices than to SPC. Is this correct? And are you building to a specific standard or regulation? This will help us to move it to the appropriate medical device forum where you will get more relevant answers.
 
K

keizer_soze

Re: Use of SPC in Low Volume Production of Medical Devices

Hello Arios,

For your specific question #1, strictly talking about "Design Verification" I would confirm that there is an output for every input, that is the idea about Design Verification. An output is design drawing, a given specification, or a certain conceived feature on the product. Having said this I don't clearly understand why the proposed analysis tools, but I could be wrong.

I agree, but what is the output for the following input:
All product handles must have a tensile strength > X Newtons.

There is no way to 100% verify like using a drawing or spec since the only way to measure it is to destroy the product. From what I know, the only way to confirm output meets input is to perform destructive testing on a sample and use stats to show that there is high confidence the design meets its input. Is this not so?

For your specific question # 2, you could confirm the consistency of the UV adhesive dispensing process. If you use an automatic or a pedal dispenser I would confirm using a capability analysis that the amount dispensed is uniform at its various rates. As for the product itself, you could validate the operation on the basis operator qualification to perform the operation properly, and for this you use attribute sampling for no leaks, no occlusion, or visual checks with a lense, or other tool. The operaion would remain validated provided you use qualified operators to perform it and provided there are no changes. Via IQ I would confirm the intensity of the UV curing process using a light intensity meeter (don't know the name in English..help!) and have a fixture to insure the light is measured always at the same distance.

I would also establish a frequency of change of the ligth bulbes as part of the PM routines developed during the IQ

Hope this helps:2cents:

Please elaborate on what type of capability analysis. When I hear capability, I think of Cpk - is that what you mean? On a side note, I would argue to do as you suggest if we had the time and money for all of it, but for the sale of 50 devices a month, the costs would go through the roof!

Thanks.
 

Miner

Forum Moderator
Leader
Admin
Re: Use of SPC in low-volume production of medical devices

Hello Miner,

I suppose yes, the question is more related to process validation. I always thought both PV and SPC are heavily interlinked, but that could just be my inexperience.

As far as standards are concerned, yes, the tensile requirements are from a standard - I believe ISO 10555-1 is one of them. And of course also to ISO 13485 and FDA QSR. I suppose the question is more related to implementation of ISO 13485 requirements of properly verifying and validating the design and the processes, and how much is considered adequate depending on the "manual-ness" of the assembly procedures and the number of samples available.

Thanks!
I moved your thread to the ISO 13485 forum for input from experts in that field. I suspect that SPC is just one of many possible approaches to Process Validation in the field. However, we will let the experts chime in.
 
A

arios

Re: Use of SPC in Low Volume Production of Medical Devices

Hello Arios,

I agree, but what is the output for the following input:
All product handles must have a tensile strength > X Newtons.

There is no way to 100% verify like using a drawing or spec since the only way to measure it is to destroy the product. From what I know, the only way to confirm output meets input is to perform destructive testing on a sample and use stats to show that there is high confidence the design meets its input. Is this not so?

Then you are probably talking either about process or design validation. Not about design verification
 
T

treesei

Something I have been thinking about operators in a process which heavily depends on the operator's performance: Consider operator as a factor of the process (like a piece of equipment) and use the same concept of IQ, OQ, PQ.

The concept of process validation is to ensure the process consistantly produces products that meets the specs so that batch tests can be optimized (I don't want to say minimized). A certain quantity of samples are needed for validation but once the process is validated the small batch size in production is no longer a problem. Process validation is a one-time deal until the process changes and needs re-validation.
 
J

jscholen

Let's start with some fundamentals from my perspective:

process validation is establishing by objective evidence that your process has a high degree of assurance as to its capability.

SPC is statistical process control.

You don't use SPC to establish capability (unless you have a significant amount of collected data to rely on), you use SPC to provide continued assurance that your PV results are still valid.

Process Validation is typically established through some level of confidence in the outcome, ie, 95/95, etc..

On a manual process, you need to establish confidence in the operators ability to manually assemble the needle and hub. If you haven't already thought about it, You should try to dumb down the process to a point where you DON"T rely on the operators judgment (too subjective). Example, if the operator must dispense glue on the needle, don't rely on their judgment on how much glue to apply, use a glue gun that dispenses the same amount each time..either way(judgment or automatically dispensed), this is an example of a process you would provide statistical assurance.

So, my specific questions are:

1. What is an appropriate method to analyze results from design verification? For example, should we do a 95/95 test? A t-test? Mean - 3 standard dev > minimum? Or just make sure all test pieces pass the criteria? Or something else? Keeping in mind we have limited quantities to spare due to cost constraints.

I would recommend you establish a confidence interval and I would bite the bullet on establishing a high confidence if your risk analysis states your process is critical to patient safety. Make your customer/senior management understand the implication of a recall if you don't get this right.


2. For transferring product to production, our process validation requires at least 3 lots be validated. We perform an IQ, OQ, PQ, but the processes are extremely dependent on the operator. So I don't know how to analyze the test results, or whether the validation is meaningful. Should we do a Cpk analysis of 1.33? Or some other way to ensure the processes are capable for production in the long-run?
First, you will need to analyze and establish confidence in the operator(s) ability to meet the process specification if it has been properly translated from design specification.

Second, then you can do your 3 lot PQ to establish confidence in your process. As I understand it, Cpk will be hard to establish on low volume and tight confidence interval...this depends on your design tolerance intervals.

This really depends on how well controlled your process is. If you run a production size of 50 units, then having units that you can do destructive testing on during production, (ie, in your example of needle and hub, sample off the front end, middle, and end of production for a certain number of your production units to do pull testing) to provide some assurance that your production lot, can go a long ways toward long term capability.
 
M

MIREGMGR

...the processes are extremely dependent on the operator.

Which is a less-clear way of saying they're uncontrolled.

First, as others have noted, get a process under the appropriate degree of control. Then begin to consider its validation.
 
Top Bottom