Medical practice can be very hard to change. I have previously covered a paper from 2003 that showed that the “pink lady” or “GI cocktail” provides no benefit to patients, but lidocaine is still widely used for epigastric pain in emergency medicine. Of course, that prior paper was somewhat messy, as it included an anticholinergic in the GI cocktail, so maybe people were just waiting for a cleaner RCT to change their practice. If that was the case, here is the trial you have been waiting for.
Warren J, Cooper B, Jermakoff A, Knott JC. Antacid monotherapy is more effective in relieving epigastric pain than in combination with lidocaine. A randomized double-blind clinical trial. Acad Emerg Med. 2020;10.1111/acem.14069. doi:10.1111/acem.14069 PMID: 32602148 Australian New Zealand Clinical Trials Registry ACTRN12619000928112
This is a partially blinded single centre RCT.
Adult patients with epigastric pain or dyspepsia presenting to an emergency department, in whom the treating physician prescribed an antacid/lidocaine mixture.
10 mL of antacid (Gastrogel: 500mg AlOH, MgSO3 240mg, MgOH24mg per 500ml) plus 10 mL of lidocaine 2% solution (the same 2% lidocaine you use for injection).
10 mL of antacid plus 10 mL of lidocaine 2% viscous gel.
20 mL of antacid alone.
The primary outcome was the change in pain score at 30 minutes, with a predetermined minimally important difference of 13mm on a 100mm visual analogue scale. (Huge props to the authors for defining not only statistical, but also clinical significance before collecting data.)
Unfortunately, the registered primary outcome was change in pain at 60 minutes, and the authors don’t explain why they changed the primary outcome. (Journal editors really cannot let this happen. If a trial is changed from the registered protocol, that change absolutely must be explained in the manuscript.)
Out of 219 potentially eligible patients, they enrolled 89. The median age was around 40, and more than 25% had pre-existing gastrointestinal issues. About 25% of each group had previously used a PPI. Between 10 and 20% of each group received a prehospital antacid, and about ⅓ of patients had taken other prehospital analgesics.
There were no statistical differences in pain scores. At 20 minutes, the median pain score in both the antacid alone and the lidocaine solution groups had passed the predefined minimally important difference, but the score in the viscous lidocaine group hadn’t. By 60 minutes, the median pain score in all groups decreased by the minimally important difference, although none of the changes are all that impressive. The antacid alone group looked better at both time points, and although that doesn’t mean it is better (these are very small, non-statistically significant differences), it makes it very unlikely that it is worse.
Participants preferred the antacid alone. There were statistically significant differences in taste, bitterness, and overall acceptability, all favouring the antacid alone group.
Adverse events are poorly reported, but are stated to be higher in both the lidocaine groups.
|Viscous Lidocaine||Lidocaine Solution||Antacid alone||p-value|
|Initial pain score||64 (36-81)||65 (31-78)||69 (57-80)|
|Change in pain score at 30 minutes||9 (3-26)||17 (7-27)||20 (7-36)||0.30|
|Change in pain score at 60 minutes||21 (3-31)||26 (9-41)||32 (12-42)||0.18|
|Overall acceptability of medication||50 (32-79)||57 (50-73)||75 (50-89)||0.01|
This is a small, single centre trial that was imperfectly blinded. There is certainly a risk that the results are shaped by bias. However, prior research has indicated that these ‘GI cocktails’ don’t help, so there was little reason to be prescribing lidocaine in the first place, and this paper certainly doesn’t support the practice. (Berman 2003)
This paper was published as a research letter, which means that we get less detail than in standard publications. It isn’t clear to me why we only get a limited manuscript, as it is a reasonably well done, pre-registered RCT. It is certainly of higher scientific value than a lot of the crap published with incredible fanfare in the New England Journal of Medicine. Unfortunately, our assessment of studies published as research letters should probably be downgraded somewhat, simply because we are not able to see as much detail, and we might be missing important factors that could influence our interpretation of the study.
The real question is why was this only a research letter and not a full publication. I think this is probably another example of how broken our current publication system is. Publications are chosen based on the results section (focusing primarily on exciting results), whereas scientific quality is found entirely in the methods section. I firmly believe that papers should be accepted by journals based on their methods section before any data is collected, so that the decision to publish is not swayed in any way by the results, but only by the quality of the methods.
The trial was only partially blinded. Investigators and patients were not told what group they were in, but the nurses were unblinded, which means that patients could very easily become unblinded, and could be influenced by the behaviours of their nurses, especially if the nurses had pre-existing beliefs about the benefit of the GI cocktail. Furthermore, given the awful taste of lidocaine, it is highly likely that the patients became unblinded. This matters a lot when the primary outcome is entirely subjective.
Selection bias is also a concern here. Patients were only enrolled if the treating physician ordered an antacid/lidocaine mixture. However, the best evidence to date was that such a mixture is unhelpful (and probably harmful), so patients treated by physicians who were aware of and practicing based on the best available evidence would have been excluded from this trial. Furthermore, although viscous lidocaine was accessed 219 times during the study period, only 120 patients were approached for enrollment, and only 94 were actually enrolled, and then 5 more were excluded for incomplete data.
As an aside, this study serves as a great reminder of the danger of anchoring on a GI diagnosis after giving a GI cocktail. A final diagnosis of cardiac pathology was made in 14% of the enrolled patients! Of course, this also contaminates the results somewhat, as we wouldn’t expect lidocaine or an antacid to improve the pain from an MI.
Ultimately, although I believe these results are probably true, there are some pretty big red flags for bias in this study, so we probably can’t make definitive claims either way. However, there is no evidence of benefit, and some evidence of harm, so until we see more studies there is no reason to mix lidocaine into antacid when treating patients with epigastric pain.
In this single centre RCT, adding lidocaine to an antacid provided no extra pain relief, but increased adverse events and decreased the palatability of the medication. The trial is imperfect, but there is no reason to prescribe the “GI cocktail” or “pink lady” at this time.
Berman DA, Porter RS, Graber M. The GI Cocktail is no more effective than plain liquid antacid: a randomized, double blind clinical trial. The Journal of emergency medicine. 2003; 25(3):239-44. PMID: 14585449
Warren J, Cooper B, Jermakoff A, Knott JC. Antacid monotherapy is more effective in relieving epigastric pain than in combination with lidocaine. A randomized double-blind clinical trial. Acad Emerg Med. 2020;10.1111/acem.14069. doi:10.1111/acem.14069 PMID: 32602148
Justin Morgenstern. Don’t use lidocaine for epigastric pain, First10EM, 2020. Available at: