I have my own view of mixing. I think mixing may be "convenient" but it carries some risks. Mixing of Regular and NPH, both bovine/porcine and rDNA human insulin seems to work "ok." It has been studied extensively and the R/NPH formulations interact in a predictable way. But it should not be expected to be the same as two separate injections, as Bernstein notes and has been mentioned, they tend to muddy each other. And in a way that can be variable. Variability is our enemy.
Which brings us to Lantus. Lantus works by having the insulin in a highly acidic formulation and when it is injected, the change to a more neutral pH causes the Lantus to form little crystals. These crystals dissolve over time giving Lantus it's long smooth action. And mixing Lantus can result in messing up with this process. Here is what the prescribing information says:
Do not dilute or mix LANTUS with any other insulin or solution. If LANTUS is diluted or mixed, the solution may become cloudy, and the pharmacokinetic or pharmacodynamic profile (e.g., onset of action, time to peak effect) of LANTUS and the mixed insulin may be altered in an unpredictable manner. When LANTUS and regular human insulin were mixed immediately before injection in dogs, a delayed onset of action and a delayed time to maximum effect for regular human insulin was observed. The total bioavailability of the mixture was also slightly decreased compared to separate injections of LANTUS and regular human insulin. The relevance of these observations in dogs to humans is unknown.
And finally, let me comment on the cited studies. The first study (2006) looked at 110 children with somewhat poor blood sugar control (HbA1c 8.5%) and found that mixing had no significant effect on their blood sugar control. A result that had a P=1, which means the result can entirely be attributed to random chance. Namely, the study didn't resolve anything (many researchers either don't understand statistics or just write up total bull). And even if it did, the fact that poorly controled diabetes didn't get worse with mixing is hardly a resounding endorsement.
The second study was very small (14 total), and looked at three cohorts, one with a normal MDI (Once day Lantus, 3-4 separate rapid injections), one "separate" (Split Lantus, 1 separate rapid injection at lunch) and mixed (split lantus, mixed rapid at breakfast/dinner and separate lunch rapid). The stated results were all pretty good control, all the regimes did "ok." But was it ok? The study couldn't tell, it didn't have enough participants and was not well controlled. But the conclusions are telling.
Lower nocturnal blood glucose concentrations in study mixed versus study separate and basal, although not statistically significant, should alert physicians that the evening dose of lantus may need to be titrated to prevent hypoglycemia. No serious adverse events occurred during the study. Although the mixtures turned cloudy, no complaints of increased pain or injection difficulties were reported.
The authors noted that the "mixed" behaved differently than separate injections. The mixtures turned cloudly, indicating that the Lantus started precipitating immediately, not a good sign.
The prescribing information advises against it and the studies don't provide very compelling support. But everyone can make their own choice. When I used R/NPH, I routinely mixed. And as Jen pointed out earlier, one should always draw the rapid first as any long acting basal can totally contaiminate a rapid with the slightest molelcule.