Q How did you develop the sample preparation and liquid chromatography parameters for your first method?

A Our first liquid chromatography-tandem mass spectrometry (LC-MS/MS) method was total testosterone for pediatric patients. I researched the literature and consulted with colleagues, knowing that achieving sufficient sensitivity would be the biggest challenge. From these investigations, I found LC column, LC gradient, mobile phase, sample extraction protocol, and MS/MS parameter examples. Initially I performed infusions of pure testosterone standard into the MS/MS to optimize parameters for our specific instrument.

Major optimization of the LC gradient was necessary as there was a very high baseline signal at the retention time for testosterone on the LC method we initially developed. This was the most difficult part of method development, and eventually we accepted an increase in run time from 4 to 7 minutes. The improved baseline with the longer run time was more important to us than a higher throughput. Additionally, we found with the extended LC gradient, potential steroid interferences were already separated from testosterone, saving us a lot of work.

Although the initial liquid-liquid extraction method effectively cleaned up the sample, the injection volume was so large that only one injection per sample was possible. I was not comfortable with that, as all instruments eventually need some samples to be re-injected. I experimented with reconstitution solution conditions and injection volumes until two injections could be made from each vial while maintaining the required sensitivity. Obtaining a 2 ng/dL lower limit of quantitation (LLOQ) was not too difficult as our MS/MS is highly sensitive and our sample preparation process is very stringent. We ended up with a 14% coefficient of variation and a signal-to-noise ratio >15:1 at the LLOQ, easily meeting our specifications.

In the beginning of method development we had high pressure issues that would shut down the LC system. We tried increasing the column particle size and length but that did not completely fix the problem. In the end, our solution was to increase the internal diameter of the tubing that connects the injection port to the front of the column. So far we have not contaminated our MS/MS, which for us completely justifies time spent on sample cleanup.

Q What was the validation process like for your first method?

A My goal was to make it as similar as possible to other laboratory-developed­ test validations at our facility. Additional requirements for LC-MS/MS assay validations include matrix effect/ion suppression studies, extraction recovery, use of confirmation ion ratios, and ensuring that the internal standard does not contribute to the analyte signal.

I used our standard laboratory validation protocol that incorporates some Clinical and Laboratory Standards Institute method evaluation documents. I reviewed College of American Pathologists checklists. I attended a short course detailing clinical LC-MS/MS method development and validation and through networking at conferences, was able to consult colleagues running clinical LC-MS/MS methods—advice from this community was invaluable.

Before carrying out a full validation, I did what I call a mini- validation—or pre-validation stress testing—to check all the parameters of a full validation, but on a smaller scale. This avoids the pain of performing time-consuming validation testing, only to discover that the method doesn’t meet specifications. If the latter occurs, the lab has to modify methods and start the validation again from the beginning.

For validation, we also purchased the National Institutes of Standards and Technology Standard Reference Material (NIST SRM 971) for testosterone to verify calibration. We compared results from patient samples tested with our method and with two reference laboratory methods. Our results across the linear range of our assay differed from these methods by approximately 20% and 4%, respectively. With the additional assurance that our calibration had been verified against the NIST SRM 971, we deemed these patient comparisons and our accuracy to be acceptable.

In the past 4 years, the method validation protocol hasn’t changed, but with practice, we have become more organized. One pointer is to have as many different people as possible participate in the validation and extend the validation over weeks—not days—to better emulate performance of the assay in production. For one of our analytes, the quality control (QC) range based on the precision from validation was artificially narrow, and we had to widen the acceptable QC range shortly after go-live. I also learned that making good calibrators and QC is technically challenging and labor intensive, so now we purchase them. However, in addition to standard lot-to-lot testing, we check the calibrators for accuracy before implementation and parallel assay new lots of QC.

Q What tips do you have for new users about the go-live and post-live monitoring phases of implementing an LC-MS/MS method?

A I recommend ensuring that a detail-oriented person with LC-MS/MS expertise reviews all results and raw data at least for the initial few days or weeks of production testing. No one in our lab had run an LC-MS/MS before, so at the beginning I reviewed every detail before releasing patient results. Make sure the procedure is specific, thorough, and crystal clear at each step. A step-by-step job aid for sample setup is useful to post at the bench.

Our QC department performs QC chart reviews and we have QC rules implemented in our laboratory information system; however, I also monitor the QC for trends. Posting a chart at the data review bench with the analytical measurement range (AMR) and carryover criteria for each analyte was helpful to us. Most LC-MS/MS software doesn’t flag results above the AMR, so it may not be obvious if samples need to be diluted before reporting. It is helpful to know as soon as possible if samples require additional processing, as it will determine if we can move on to the next batch, or if we have to delay that so we can run dilutions.

I also recommend making a chart with the expiration dates of all reagents. As our calibrators and QC are custom-made, the lead time can be up to 2 months, so we have to make sure they are ordered, received, and assayed before the current lot expires. The last thing a laboratory should do is try to extend the life of reagents.

For post-live monitoring, I ask our technologists to have a very low threshold for referring a sample to me for review or letting me know if something just doesn’t look right. Sometimes an observation that might appear to be trivial can be an early indicator of drift in an assay, such as a deteriorating column or bad reagent. If the retention time is shifting, there is higher background than usual, or an extra peak.

I also personally train every technologist who works on the LC-MS/MS to ensure that training is consistent. If another technologist has developed a bad habit, such as refilling a mobile phase bottle instead of taking a clean one, I do not want the new person to pick it up. Most important is keeping up with instrument cleaning and maintenance: this ensures instrument longevity and minimizes down time. We have daily, monthly, and quarterly maintenance schedules and preventive maintenance carried out twice per year by LC-MS/MS engineers. A service contract is invaluable as these instruments have expensive parts and can take time to fix.

To ensure quality and keep our system performing correctly, we run a system suitability sample before each batch of patient samples. Internal standard peak areas should be consistent within each run. We average the internal standard peak areas in the calibrators and QC, and we have specific criteria that internal standard peak areas of patient samples must fall within in order to be acceptable (e.g. +/-2 standard deviations of the mean). We also monitor confirmation ion ratios during each run. The average ion ratio for both analyte and internal standard is calculated for the non-zero calibrators and QC. Patient sample ion ratios have to be within +/-20% of the average ion ratio or it may be indicative of an interference in that sample. For example, recently the testosterone ion ratios alerted us to a patient sample with an interference that was not at all obvious from looking at the peak shape.

In our experience, the effort spent monitoring these parameters is paid back a thousand fold by less instrument down-time, fewer failed runs, and confidence that our LC-MS/MS assays are functioning appropriately.