The study is Suppes, et al., 2009. Maintenance Treatment for Patietnts with Bipolar 1 Disorder: Results From a Noarth American Study of Quetiapine [AKA Seroquel] in Combination With Litium or Divalproex (Trial 127). American Journal of Psychiatry, v. 166, pages 476-488.
After getting patients with bipolar disorder stabilized for 3 months on seroquel, pts were randomly assigned to either stay on Seroquel, (plus have lithium or valproic acid as augment), or to get placebo Seroquel (along with lithium or valproic acid as augment).
Seroquel was successful at the main outcome: fewer Seroquel-plus-other-med pts relapsed compared to placebo-plus-other-med.
The study humbly presents a full panel of metabolic data, also, relating to body weight.
In the long-term (about two years) follow-up, the group that continued with Seroquel had greater elevations in glucose (not good) compared to placebo, and greater weight gain compared to placebo (not good).
For the glucose issue, it appears that this differs by diabetic status: little difference for those with no diabetes risk, modest but significant difference for those with diabetes risk, and scary changes for those with diabetes. For each of these three sets of people, I am looking at both the average, tellign which group was higher, plus looking at the standard deviation, which tell how orderly or wildly the glucose changes are.
Blood glucose for those with no diabetes risk looks the same between both groups (Seroquel and placebo).
Results diverge for the diabetes-risk group, with greater standard deviations - which means there are a few things going on to make many people have many differnt experiences. This result is probably not statistically significnt, but it is a concernign pattern.
For the diabetes group, the change in glucose is much greater for the group taking Seroquel vs. placebo, and the standard deviation (a measure of how much the typical person has changed) for the Seroquel group is off the chart: Seroquel group has standard deviation of 143 milligrams per deciliter, while the diabetic group taking placebo group has standard deviation that is a third of that (44 mg/dl).
An additional dimension is the effort to manage diabetes in these patients. I am going to talk here about the diabetic group only, so we can maybe figure out what is happening with these people. No "baseline" glucose is given for the diabetic groups (diabetic with Seroquel, and diabetic taking placebo). Now, here is a weird result: at the two-year follow-up, BOTH groups end up with the SAME average glucose level, 125 mg/dl, but how they got there differs:
The diabetic group taking Seroquel have had an average INCREASE (of 30 mg/dl avg) in blood glucose to end up at 125, but the diabetic group taking placebo has had an average DECREASE (17 mg/dl) in blood glucose to end up at 125.
Plus, as I said above, the standard deviation in change across time is much greater for the diabetic group taking Seroquel compared to the diabetic group taking placebo.
Why were baseline glucose measures not reported? DK. That would help make sense of this data. Per the article, it seems that glucose was taken at baseline, since that was used, along with other info, to define the diabetes group.
Demographics are not provided to let us know how many ppl were diabetic (per their defn), and the portion randomized to each treatment group is not reported.
While the reporting of randomization info is barely at the acceptable level for a decent study (there are standards, you know, like CONSORT guidelines), we can't tell if the weird pattern of results is due to very different numbers of ppl in the two groups.
Either way: this study fits in line with other studies, indicating: 1. it may be time, for someone with access to this and other data (i.e., FDA) to declare that diabetics should not be taking Seroquel; 2. somebody needs to figure out what effect Seroquel is having upon the bodies of ppl with high BMI and / or with diabetes.
The results are dramatic, and I don't quite see the fasting issue as explaining the scary difference: if you have diabetes and take Seroquel, it is going to be very difficult to manage your diabetes.
A funny thing about this study:
It is designed to see whether Seroquel-plus-adjunct-med is superior to Placebo-plus-adjunct med.
For the adjunct meds, each was given at the recognized therapeutic dose, not at some sub-optimal dose (as some prescribers probably still give Lithium: typically augmenting some other med with 300 mg lithium - far below clinical dose which will normally start around 900 mg).
But if you flip this study around in your mind, it becomes a test of Lithium versus lithium-plus-Seroquel, along with a test of valproic acid versus valproic acid-plus-Seroquel.
I think that is something to think about: it makes you see: this study is not about efficacy of Seroquel, although that is what the study authors (with a long list of Big Pharma connection$) seem to be trying to convince the FDA.
Update: a bit more info:
was this actually a test of divalproex or lithium, with or without augmentation by quetiapine?
For that, you would actually have to have therapeutic dosing levels of the divalproex and the lithium.
This study notes that blood lithium levels, in both arms of the study, were at mean of about .7 meq/liter. This is the bottom of the therapeutic window range.
Divalproex blood levels were at average of about 70microg/ml for each group. This is the bottom of the therapeutic window.
And these are averages. Standard deviations are not given, so we are lacking a bit of info on what portion of pts may have beem below therapeutic window. It is possible that these levels were controlled pretty tightly - the fact that the averages are similar in both groups would be quite a coincidence, in light of the headroom available that either drug could have been titrated up to - this is both good and bad - if the study researchers actually intently managed blood levels of these to adjunct meds, then they hypothetically have achieved levels that would maximally favor Seroquel - levels of lithium and divalproex that are within the therapeutic window, so as not to draw attention, but at a point that is gonna lack efficay for a decent portion of patients.
Failing to include standard deviations for these values, or ranges, or some measure of the portion of time pts spent within therapeutic range, is a real limit to understanding this study.
I believe it is safe to say that the lithium and divalproex treatment was technically within normal practice, but barely, and it would not be the optimal way to test a trial of either. Contrastingly, quetiapine WAS dosed at a normal, sufficient level - data suggests at least 400 mg for normal tx response, with a window up to 800 (beyond not tested because there is no reaason to - either 400 to 500 seems to work well, or you are not gonna get a response) - and this study had average dose of 500.
So, whether intentional or not, I would say that the dosing of each of these 3 medications happens to be in the pattern that happens to tilt the study to favor Seroquel.