Specifications: A double edged sword?
As we all know specifications are essential for any product and not only affect the validation and verification of a product but everything that is the product. We saw in the minisim this week one of the set backs of the high specificity in a requirement. Should specifications reflect the limits to which any individual component can be to function properly, with buffers of course, giving more room for certain variations but increasing the risk? Do specifications have to be the guidelines for validation or testing in QC or are they the outer limit? Is it okay for specification to have a one sided bound as opposed to double bound, such as in the case with the better coefficient of friction? (assuming it doesn't affect the product in a negative way)
A metric for a specific feature should always fall within the defined specifications; otherwise the documentation is meaningless and must be changed accordingly. Certain metrics can most likely have only a lower or upper limit if the design calls for it, such as a coating having a curing time of at least 2 hours where any time spent curing after 2 hours does not affect the coatings function in anyway, this would also have to be verified beforehand. The physical limitations of the materials also play a crucial role in limiting their design specification and the specifications of other parts. If a part needs to be made from a specific polymer that cannot be subject to temperatures above a certain threshold that the entire device is affected if temperature is a factor in its function.
when you write the specification sheet, SOP or BOM list. It has to be as accurate as you can. one point of view that it will not affect the product in a negative way, which I agree with you for this product only. However, if you manufacture other product that can be used for different application or you will market that for a different market. the customer would like to have all the right accurate information about the product. From my experience in an Industrial flash and storage business, one product may work with one machine, not the other due to one of the major build component. in case of our mini-simulation, the coefficient of friction was an important factor to help the catheter to slide all the way up to the brain. Also, even it doesn't affect the final product, the company will be reliable for any wrong information listed in such a document. It's better to be safe than sorry.
The specifications should always have bounds, and if there is a bound for the coefficient of friction, then it is because that is what is desirable for the specific function. For example, sometimes you wouldn't want there to be that little amount of friction in a product. However some bounds you can theoretically say would just make the product better and better if there didn't exist a higher or lower bound, but then the cost could possibly be the reasoning behind that. For example, to make a product better may just make it more expensive when not necessary.
I agree with andria93 and cs22 that specifications should be as accurate as you can to make sure all aspects are documented and followed correctly. The limits are an interesting aspect because there are times when there should be upper and lower, or just one, but overall I do think of them as a guideline for validation and testing. In this mini-sim, it was essential to have the COF of a certain degree, thus the guideline was fixed to have a range, through be it small.
We saw this week the difficulty in having a very specific DSD. Once the product component was considered a critical component it left us scrambling to find another replacement that would exactly meet the criteria. When design the specifications it should not be designed towards the supplier but rather what is actually feasible for the product. Making it specific to the one supplier or roughly specific to the supplier leaves the company very vulnerable to needing design changes. Not only is the concern that the design will not meet the criteria during designing but also this document is permanent. This means that if this supplier closes and you can no longer purchase from them you need to go through a massive change to qualify a new supplier.
I think that it may be a good idea to put the extra time into thinking about the specifications of a critical component. Yes you want them to be as specific as possible, but if you know during design specification that you will need to dual source a specific component for your product, maybe it would be a good idea to do some research on what specs are available from the suppliers before making a concrete spec so that you don't have to return to design controls to change the spec once you're near project completion. This is also where experience and superior planning skills make a huge impact.
Specifications should be broader during the beginning of design control, but should narrow down during the later stages. The project team can also refer to previous similar projects and compare previous specs to their estimated ones. Products sensitive to failure and/or will require in vivo testing should have smaller spec widths. It should also depend on how many sources/vendors are needed. If the product is a critical component, then more than one vendor is required. When choosing the alternate vendor, the specs of the product should be similar to the original vendor. As the amount of vendors increase, the spec width will increase to accommodate the different ranges each vendor poses. This will decrease the risk of falling out of spec, but it will also increase cost and time. Therefor, it should also depend on company's capability of delaying the project and increasing the budget.
Using the minisim as an example. The product that was being made required a specific amount of friction in order for the product to work. Specifications should most certainly be followed to a t when a final product is being worked on. Deviations of specifications can cause many issues in a product or even failure; which is not a desired outcome whatsoever. However, it would be possible to play with the specs during testing. For instance, in the minisim if the company was to experiment with different friction coefficients different than the original specs this may be okay. However, one final friction spec would need to be chosen, and used throughout validation, and so on. Not following the spec's in the final SOP, or using different materials than listed on the BOM is likely to cause major issues.
Before solidifying the DSD, you would want to make sure the supplier is capable of meeting the requirements. In the case of the minisim, this is what was done. An issue was discovered due to a lack of following company SOPs. If SOPs were followed, the specs would have already accounted for AquaWoah as an alternate source. This way, we would not have been scrambling to find an alternate source as a backup.
The way that I think about this issue is that having a wider range of specification is good for many reasons. First of all, it gives a lot of leeway in terms of manufacturing the product and using the product. If there is not a strict limitation into the device, then the errors producing during manufacturing wouldn't be as frequent as producing a product that must be produced at a set point (e.g. length, width, radius). I think that it's better to ave a one-sided bound than a two-sided bound because you have more freedom and less stress after the device is introduced to market. If the range is long, then that implies that the consumer will not have a difficult time in using the product because since the specification range is not narrow, then the device should be able to work under more circumstances, such as when it has broken. The specifications should be stretched until the point where the device will no long work if it goes beyond it. The only leeway that you should provide is when you can preserve the functionality of the device.
The specifications that are outlined for a product are the limits of that product. Any output of that product that is outside of those bounds is not supported and is not usually a liability to the company. To avoid customer frustration and risk of failures closer towards the limits of a product, buffers are often put in place which, although seems to increase risk, helps in the long run. Furthermore, with my experience in validation, the limits are often the furthest we will go for testing. In addition, however, we dedicate a good amount of time in attempting to bypass those limits which we always enforce hardblocks on (this is for software). Ultimately, testing the supported limits, as well as the blocks enforced to prevent the user from exceeding those limits, is how validation is typically handled, from my experience at least. As far as one vs two sided limits, it of course depends on the situation. Often times, two sided limits are enforced because it is unreasonable and usually impossible to support a product that can have an infinite spec. However, in some cases, infinitely "better" is better. For instance, certain products must have a lifespan of AT LEAST "X" number of years. Obviously, the longer the better, as long as the product can function properly within the defined specifications during that period.
Generally, specifications for a medical device come from industry standards, such as ASTM or ISO. When designing a device, there will be a handful of standards that should be referenced to determine the specifications for each feature of the device. These limits will be the outermost acceptable limits, as anything outside of these bounds will be non-compliant. If your company chooses to do so, they may even require tighter specifications that what these standards dictate, as these standards are meant to cover a range of devices while your device may be used in a particular way. Although this will increase the cost of the product, quality should be considered first and foremost.
When determining if a one sided or two sided spec limit should be used, it depends on the feature you're applying it to, as well as if there are any similar "legacy/predicate" devices this new device can be compared to. For example, if there's a predicate device that performs in a similar way, your spec could say something like "XXX shall be greater/less than that of [predicate device name]". If an acceptable predicate doesn't exist, you'll need to reference standards to come up with a reasonable two sided spec limit.