Comment on “The Abrupt Cessation of Therapeutically Administered Sodium Oxybate (GHB) May Cause Withdrawal Symptoms”

To the Editor:

We write regarding a recent article entitled “The Abrupt Cessation of Therapeutically Administered Sodium Oxybate (GHB) Does Not Cause Withdrawal Symptoms 1.” The conclusion stated so unequivocally in the title is misleading and is contradicted by the abstract and the text, both of which describe symptomatology consistent with GHB withdrawal. Furthermore, the conclusions themselves are unsupported by the data and the statistical analysis is flawed, ultimately suggesting that abrupt cessation of sodium oxybate may indeed cause withdrawal symptoms.

First, contrary to what would seem to be implied by the title of the article, the study was not a prospective investigation of the incidence of GHB withdrawal symptoms. Rather, the study was solely a retrospective analysis of data prospectively collected for the purpose of measuring the efficacy of Xyrem (GHB) in the treatment of cataplexy. The primary end‐point for the initial data collection was, therefore, the return of cataplexy symptoms following treatment cessation, not the occurrence of withdrawal symptoms. Consequently, there was no prospective determination of clinical parameters by which to measure withdrawal symptomatology, no physical examination of patients, and no application of any standardized withdrawal assessment instrument or scale. Instead, the study is based solely on retrospective interpretation of adverse events noted in daily diaries of cataplexy patients.

Second, a study claiming “no withdrawal symptoms” must ask and report on symptoms specific to withdrawal, providing data on both presence and absence of such symptoms. It is unlikely that such withdrawal data were collected in a retrospective study of cataplexy symptoms. Further, there is no table providing a complete list of adverse events or documenting their occurrence in individual participants and it is, therefore, impossible to know how many patients reported none vs. one vs. multiple adverse events, which symptoms they suffered, nor even what the full range of adverse events was. Although the occurrence of tremor, which is a frequent sign of early or mild GHB withdrawal, was mentioned in an earlier abstract of the same title 2, it was omitted from the current article entirely. Other symptoms consistent with GHB withdrawal, such as diaphoresis and hallucinations, are also absent from discussion; were they even included as potential adverse events? Investigators report that 9 of 29 patients (31%) experienced adverse events, but list only anxiety, insomnia, dizziness, and somnolence. The Results section concludes with the following unsupported statement, that “although these symptoms may possibly be associated with a withdrawal syndrome, they are more consistent with the return of narcolepsy symptoms.” On what is this conclusion based and why is it stated in the Results section?

Third, although the authors state in the Discussion section that “no patients reported craving the discontinued sodium oxybate,” there is no mention of documentation of craving in the Methods section. Again, given that data collection was designed to assess the recurrence of cataplexy symptoms, it is unlikely to have included “craving” as a potential adverse event. How, then, was this purported absence of craving determined? Was a specific question on craving included on a daily checklist and systematically elicited from all patients in the placebo arm? If so, the authors should specify this. Otherwise, the statement is unsupported and should be excluded from discussion.

Fourth, although the authors cite four articles on GHB withdrawal 3-6 and report its occurrence with daily doses of 25–144 g of GHB and dosing every 30 min to 3h, a careful read of the studies reveals that withdrawal was described as occurring with a daily dose of as little as 10 g and dosing every 4.8h 3 (as compared with a daily dose of 9 g of Xyrem). Further, the authors emphasize the occurrence of withdrawal from GHB used “as a drug of abuse” but fail to note that, of a total of 18 cases of GHB addiction described in the four studies that they reference, 10 patients (56%) were using it not as a drug of abuse but for its purported (but unfounded) health benefits, including bodybuilding, to decrease alcohol cravings, and alleviation of panic attacks. Additionally, the authors' statement that GHB suppresses opiate and ethanol withdrawal symptoms is not supported by a critical read of the clinical literature, as we have addressed (regarding ethanol) in another publication 7.

Lastly, the statistics of the study are flawed. There were more adverse events in patients who were withdrawn from Xyrem, but because it did not reach statistical significance the authors claim that “there was no statistically significant difference in the frequency of adverse events” and then conclude that there is no difference. However, to be unable to prove a difference is not the same as proving that there is “no difference.” There was, in fact, a “trend” toward more adverse events in the withdrawal group; the p value (p = 0.108) shows that there is an 89% chance that the greater incidence of adverse events in the withdrawal group is real and not due to chance. In order to prove there is “no difference,” (meaning “no withdrawal symptoms”), the authors would have had to show significantly fewer adverse events in the nonwithdrawal group. That is, the results would have to be the opposite of what they found, and with a p value of 0.05. On the contrary, their data suggest that, had there been a large enough “n,” the authors would have proven that there are indeed withdrawal symptoms, which is the opposite of their purported outcome. In fact, if there had been just one more patient with symptoms (10 of 30 rather than 9 of 29), the p value by their two‐sided (inappropriate; see below) Fischer's exact test would have been 0.05. This would have established a “statistically significantly” greater incidence of withdrawal symptoms in the withdrawal group. Additionally, the authors state that they used ANOVA and Fischer's exact test for statistical analysis, but they present no normally distributed data which could be analyzed by ANOVA. Furthermore, by our calculations, the authors appear to have used a two‐sided Fischer's exact test. More appropriate would be a one‐sided test, because it should be assumed that the group continuing to take GHB will have no more withdrawal symptoms than the group taking placebo. A one‐sided Fischer's exact test (p = 0.08) results in statistics even less favorable to the authors' conclusions; i.e., with this more appropriate test of the data, the probability that the results are real and not due to chance (that is, that there really are withdrawal symptoms) increases from 89% to 92%.

Finally, there is no discussion of limitations or shortcomings of the research and there is no suggestion that prospective investigation specific to withdrawal symptomatology should be undertaken.

In closing, it would appear that a more accurate title, and one consistent with the nature and findings of the study, would be, “The Abrupt Cessation of Therapeutically Administered Sodium Oxybate (GHB) May Cause Withdrawal Symptoms.” The title is the most frequently read and easily accessed element of a clinical article 8. Our past 9 and ongoing research on use and addiction to GHB and its analogs led us to express concerns to Orphan Medical subsequent to publication of the initial abstract of the same title 2, but to no apparent avail. Severe GHB withdrawal may be lethal 4, but its development and early stages are, to date, poorly characterized. In the absence of prospective data gathered with the express purpose of investigating GHB withdrawal symptomatology, we would request that greater care be taken in investigation, reporting, and publication.

Deborah L. Zvosec

Minneapolis Medical Research Foundation

Minneapolis, Minnesota, USA

E‐mail: [email protected]

Stephen W. Smith

Department of Emergency Medicine,

Hennepin County Medical Center

Minneapolis, Minnesota, USA