The company that I am currently with (about a year now) only runs singlet injections. The company I worked at prior to this injected everything in triplicate. We recently had a customer audit and they did not like the fact that we only do a single injection. How many injections is considered "good" lab practice?

I think you can go down the middle of the road without causing too much pain. What I do is this: After purging the injector, conditioning the column, equilibrating (briefly), and injecting a blank, I run a system suitability, injecting my 100% dose standard 5X. I will then run a second blank. If I have many samples (>25), I will make replicate sample preps on every fifth sample with duplicate injections on every third. If I have fewer samples (5-15), I'll replicate every third sample prep and do duplicate injections on every second sample. If I have less than 5 samples, I make duplicate sample preps and duplicate injections on every one. I inject my standards spaced evenly in time through the run with the second to last injection being the highest standard and the last injection being a blank. Sometimes if I'm betting neurotic, I spike my a few of samples and calculate recoveries, too.

I have few difficulties with defending my data this way and if there are problems with any sample that was not replicated, I simply run it again. Usually this doesn't happen, however.

Hope this helps.

I think you will find single injections common practice in industrial labs, and other places were costs are under scrutiny.. Personally, once I established calibration the only "extra" sample I would run would be a known/standard at the end to demonstrate things were working at the end as well as the beginning.


You could answer the question with data. How reliable is your method and technique? If duplicate, triplicate or what ever redundant work does not add to the value of the data then don't do it. A lab doing duplicates either is running an unreliable method or the pepople in the lab lack confidence. Either, is an opportunity for cost saving through method development or training.

I agree with Bill's opinion. I do not do multiple injections of one sample. If it is necessary to do multiple I prepare 3 replicates for each sample type.

Hi Bill,

Occasional replicate sample preparations and a system suitability test do add to the value of the data as far as I'm concerned. When I'm asked how confident I am in my data, I can answer with data. My peak area RSD's obtained from my system suitability test are directly relevant to that confidence. Furthermore, I've seen systems fail system suitability (with a peak area rsd of >2%)and still have a marvelously linear calibration curve.

When I have a situation in which a sample comes back with an unexpected result and I have a replicate of that sample that shows the same result along with system suitability data that show good precision and a calibration that that is nice and linear - with the standards interspersed throughout the run, I can defend myself very easily. I can prove that my preparation technique is good via the replicated preps, I can prove that my system is precise via the system suitability, and I can prove that my system is stable at all points throughout the run because all samples are bracketed by standards.

Yes, it takes more work and costs more money up front, but I very, very rarely have to repeat any analysis due to an odd result and I can spot problems early on. I don't think the way I do it shows poor training or lack of confidence, though I will agree that anyone who injects everything in triplicate is nuts.

Am I the only one who's had a less than full injection, presumably from the injector needle plugging for a single injection? If you don't have an internal std & you don't know what peak area to expect (ie the sample is really an experimental sample) can you do just single injections?

And then there is also the complicated, difficult, and variable matrix which causes overlap. etc., and whose totally wrong results are perfectly repeated by multi-injections.

First and foremost "What are your customer's expectations?" I analyse the same products over and over so I feel comfortable with single injections. If we started doing unknowns or pharmaceutical quality I may feel the need for more replicates as per Bill Tindal.

Like most of the folks have stated you can generally get away with single injections if you demonstrate system suitability at the beginning and end of a run and sometimes throughout if it is a long run. We run our assay's under the assumption that if our single injections are bracketed by acceptable S.S. results then we are comfortable with the data. This is not to say that if we get a funny result we won't go back and re-inject it, but re-injecting one or two vials every so often is alot more cost/time effective than running duplicate/triplicate? injections. When running the system suit. I generally inject it 5x depending on the method and set a %RSD criteria for area and retention time.

Conrad Grob wrote an article (which of course I can't find right now!) a few years ago about method development that addressed this issue as part of the total QA plan.

If I remember correctly, he suggested that appropriate compounds be added at each major step of the analysis. Surrogates, internal control and quantitation standards, and particularly a compound added for evaluation of the injection process all add QA value to the analysis. This compound should be extremely reliable on the system, unlike surrogates, which should mimic the behavior of unknowns. This might mean it has a very different retention time or higher concentration, or not be chemically related to the unknowns. Because of these characteristics, it may be simple to include in an existing method.

While in general we only do a single analysis like most people, quite often I simply cut/paste parts of my runs to do duplicate injections, but at much different times. If the system is running overnight or over the weekend, there is usually plenty of time to do this at virtually no cost.