metformin and lactic acidosis, again

A community-based cohort study confirmed that metformin was safe to use down to an eGFR of 30 (see dm metf ckd jamaintmed2018 in dropbox, or doi:10.1001/jamainternmed.2018.0292 ).

Details:

--75,413 patients with diabetes within the Geisinger Health System, from 2004-2017; results replicated in 67,578 new metformin users and 14,439 new sulfonylurea users from 2010 (the "active comparator" group, also from Geisinger) and  from 350 private US health systems (replication cohort)

--primary cohort (Geisinger): mean age 69, 51% female, 97% white, BMI 34, hypertension 69%/cardiovasc disease 12%, heart failure 10%, mean sodium bicarb 27, A1c 7.4

--meds: insulin 17%, statin 43%, RAS inhibitor 47%, diuretics 34%, NSAIDs 21%

--primary outcome: association between metformin use and hospitalization for acidosis, stratified by eGFR

--of the 75,413 patients:

    --14,662 had eGFR <60

    --1765  had eGFR<30

--median dose of metformin: 1.35 grams/day, independent of their eGFR grouping

Results:

--Primary cohort:

    --2335 hospitalizations with acidosis over mean 5.7 years; 470,114 person-years of follow-up

        --737 acidosis events in 188,578 person-yrs in those on metformin, and 1598 in 281,536 not on metformin

    --only 29 of these events had an acidosis diagnostic code in the primary position

    --time-dependent metformin use was not assoc with incident acidosis overall: HR 0.98 (0.89-1.08), adjusting for diuretics, RAS inhibitors, statins, NSAIDs, insulin, other DM meds

        --eGFR 45-59: HR 1.16 (0.95-1.41), not significant

        --eGFR 30-44: HR 1.09 (0.83-1.44), not significant

        --eGFR <30: HR 2.07 (1.33-3.22), significant

--in both metformin users and non-users, the lower eGFR itself was associated with more acidosis

    --eGFR 60-89: 4 events/1000 person-yrs

    --eGFR 45-59: 7 events/1000 person-yrs

    --eGFR 30-44: 10 events/1000 person-yrs

    --eGFR <30: 24 events/1000 person-yrs

--in active comparator group at Geisinger (new users of metformin or sulfonylurea, n=12,690, with propensity-scoring to mathematically try to make them look like the other cohort): there was no significant differences in acidosis events between those on metformin or sulfonylureas, with only 3 cases of acidosis among the metformin users

--adjusting for baseline A1c or use of insulin did not change these results

--in the replication analysis form the private US health systems: with medianfollowup of the metformin vs sulfonylurea users of 12 months:

    --238 acidosis events with metformin (2.7 events/ 1000 person-yrs) vs 94 with sulfonylureas (5.0 events/1000 person-yrs)

    --there was a nonsignificant higher risk in patients on metformin and eGFR<30

Commentary:

--one reassuring finding was that in all 3 groups (the long-term Geisinger one, the newly diabetic “active comparator” group there, and the replication group), there was no increase in acidosis in those on metformin with eGFR >30

--the current US guidelines are to preferentially use metformin in those witheGFR>45, but they do not recommend its use in those with GFR 30-44 (though okay to continue if their eGFR decreases to <45 while on metformin after assessing risks/benefits, these being undefined), and avoid using if eGFR<30 (see http://gmodestmedblogs.blogspot.com/2016/04/fda-changes-metformin-guidelines.html)

--there are some evident potential shortcomings to this trial:

    --they looked for codes for acidosis, not lactic acidosis, though they did exclude codes for diabetic ketoacidosis. So, it may not be accurate to translate their data into assessing metformin-associated lactic acidosis

    --and, “only 29 had an acidosis diagnostic code in the primary position” (a really tiny fraction of the patients), suggesting that these people may not have had significant acidosis (did they just have a somewhat low bicarb, perhaps related to their renal disease? Or perhaps a not-so-bad infection associated with a low bicarb??). and, the relevant background here is that metformin's predecessor in the US, phenformin, was outlawed after a striking yet frequent occurence of lactic acidosis which was almost uniformly fatal (ie, one might expect this diagnosis of acidosis to be more in the primary position if the association with metformin were similar....)

    --and, perhaps buttressing this last comment: lower eGFR in both metformin users and nonusers was associated with a higher incidence of acidosis. This was confirmed in their replication analysis from the other database

    ​--also, the broad strata of eGFR may have hidden important differences: did the metformin group have more patients with eGFR in the 20-25 range vs 25-30 in the non-metformin group?? that difference, not picked up by the broad "<30" division could explain the difference in acidosis (having thse 3 groups with similar results does make this less likely but still possible)

--one big  issue in overinterpreting these results is that even in patients with eGFR<30, the median dose of metformin was still 1.38 g/d, and the same as in those in the higher eGFR strata. My experience, with some limited supporting data, is that patients get a large % of metformin effect at very low doses (I have a few on only 250mg/d because of GI intolerance), and I have many patients doing very well on just 500mg once a day.  My own approach is to limit the metformin dose to 500 mg/d in those with eGFR<45 (after all, it is renally cleared)

--one significant positive in the above trial is that they accounted for changes ineGFR over time, important given the typically slow deterioration of renal function in diabetics.

--a 2104 study of metformin in patients with type 2 diabetes and kidney disease reviewed the literature and suggested the lower limit of eGFR=30, given limited data on using metformin in patients with lower eGFRs  (see dm metformin ckdjama 2014 in dropbox, or  doi:10.1001/jama.2014.15298. see blog reviewing this review at http://gmodestmedblogs.blogspot.com/2015/01/metformin-in-renal-failure.html  ). One real advantage of metformin is that it seems to be cardioprotective (data from UKPDS, other secondary analyses), but also in this review they cite a group of 19,691 diabetic patients with renal insufficiency and atherosclerotic disease, finding that overall mortality rates were 30% less in those on metformin, even in the group with eGFR30-60.

--and a rather striking study in Taiwan found that even in patients with stage 5 kidney disease (eg, eGFR<15) and on metformin at max dose of 500mg/d did not find increased mortality, though those on higher doses did not do so well (see http://gmodestmedblogs.blogspot.com/2015/06/metformin-ckd-and-death.html )

So, I realize that I pretty frequently bring up this issue of using metformin in those with CKD, given the many impressive benefits of metformin. This study, though with above limitations, does reinforce the relative safety of metformin in those with CKD and eGFR >30. And, despite the relaxing of restrictions by the USPSTF, and before that by European agencies, I still find many clinicians who are quite hesitant to use metformin in those with CKD or are stopping the metformin with decreasing eGFR levels even above the USPSTF thresholds. I would add that it really does not make sense to use more than 500mg/d in those with more severe CKD, since that dose typically works well and may well continue to be cardioprotective.