The issue in this situation is not whether the creatinine level actually changes, but whether the test result changes. Assuming that the urine specimen is properly stored, it will not have significant changes over time.
For any lab test, though, there is some variability in the results, even if the exact same specimen is tested multiple times. For screening methodologies, the tests are typically less reproducible, which means that there can be greater variability in the results that can be obtained from each specimen. If the result from a scsreening test is then repeated by the lab, using a methodology that is more accurate, then the lab result is typically much more reliable, so should be the creatinine level that should be used in any decision making.
There are many different brands of tests on the market, but the package insert to whichever brand of test is being used to perform the screen should include documentation of the variability that is seen with that test. Your lab personnel would be able to either review the insert or provide you with the insert, and that would be the best documentation of the variability, rather than a clinical study in the medical literature.
As for when to witness a collection, there are no studies that look at whether there should be concern about recurrent similar values on consecutive specimen. The studies have looked at the threshold of creatinine that should raise concern and justify a witnessed collection. There are varied recommendations, but part of that is the level of acceptance intrinsic to the recommendation. When we define any normal range, it is usually the range that encompasses 95% of the normal population. So, about 5% of normal people will have a result outside the normal range. In studies that have looked at traditional determinations of the normal range for urine creatinine, the threshold that would justify a witnessed collection would be a level of 45 mg/dl. However, there also are recommendations that the threshold should be 20 mg/dl, including federal guidelines. These lower thresholds are not based on studies that show a different normal range, but instead reflect a recognition that many normal people will be slightly outside the normal range, and a level below 20 mg/dl will identify a larger percentage of people that are intentionally attempting to dilute the urine. It is not that either threshold is right or wrong, but the decision of which threshold you want to follow will reflect your goals of testing. If you want to be more likely to detect everyone that is trying to dilute the urine, even if you need to witness far more normal people, then using the threshold of 45 mg/dl would be appropriate. But if you would rather limit the witnessed collection of normal people, then it would be appropriate to use the federal 20 mg/dl threshold.
If I can provide any clarification, please let me know.