We did not observe a relevant anti-inflammatory or sUA lowering effect of metformin during the first six months after starting ULT in a real-world setting. Although these effects of metformin are supported by pharmacological and empirical evidence, several contextual factors can lead to a null effect when treating gout patients in a real-world setting.
Firstly, the anti-inflammatory effect of metformin might be too weak to have a clinically relevant contribution in gout treatment in a phase where strong anti-inflammatory treatments like colchicine are prescribed as prophylactic treatment . Another explanation for the lack of difference in gout flares in this study is the effect of other possible variables that interfere with the proposed anti-inflammatory mechanism of metformin, for example state of diabetes regulation. Poorly controlled diabetes is described to decrease the risks of gout flares in some studies [14, 15]. This suggested mechanism in diabetes mellitus might counteract the possible effect of metformin, however we did not have the data to correct for this possible mechanism.
The lack of sUA lowering effect of metformin might be driven by differences in study context and design. The study by Barskova et al.  was a small intervention study with metformin in which the included patients did not use ULT. In our study all patients started ULT. Also, in our study only prevalent metformin users were included. It is therefore possible that through index event bias  our sample disproportionally included patients in whom metformin did not have a sUA lowering effect, or not enough to prevent the development of gout. However, index event bias would also reduce the proportion of DM patients and metformin users in our cohort, but with 22% DM patients of which 52% used metformin our cohort stays well within the expected ranges [3, 17]. Furthermore, other anti-diabetic medication may have this sUA lowering effect in gout as well, thus resulting in a net null result. Whether this effect is unique for metformin has indeed not been tested. Of note, previous studies have shown that even drugs within the same class can have different off-target effects, for example in a study comparing losartan and irbesartan, only losartan showed a sUA lowering effect in patients with gout .
This retrospective study might have some general limitations, such as underreporting of gout flares and a possibility of double reported flares. However, firstly we assume that this would be the case in both groups and probably should not result in a biased between group difference, secondly our flare rate is comparable with other studies [19, 20]. Also, we had no data on the type of DM. However, it is likely that most patients have type 2 DM since this accounts for 90 to 95% of all DM, and gout is mainly associated with type 2 DM [3, 21]. Also, we had no data on the state of diabetes regulation, including HbA1c levels, which may interact with the risk of gout flares as well [14, 15]. In the non-metformin group mean age was slightly higher and renal function lower, resulting in confounding by indication. However, our analyses were corrected for these differences when necessary.
Strengths of this study include the considerable sample size, resulting in adequate precision while excluding any relevant effect considering the confidence intervals, and correction for confounders. Due to the non-limiting inclusion criteria, multi-centre data collection and a prevalence of DM in the cohort within the expected range, the generalisability of the study seems solid. Also, the uricosuric effect of metformin was assessed using different outcome measures, including correction for second order effects such as differences in ULT use.
In conclusion, although pharmacological effects of metformin probably include anti-inflammatory and urate lowering effects, we could not confirm a clinically relevant effect in patients starting ULT treatment and receiving usual care flare prophylaxis.