You don't need to be signed in to read BMJ Blogs, but you can register here to receive updates about other BMJ products and services via our site.


Primary Care Corner with Geoffrey Modest MD: Insulin pumps in type 1 dm, not the best solution

11 Apr, 17 | by gmodest

by Dr Geoffrey Modest

A recent trial looked at the effectiveness of insulin pump treatment versus multiple daily injections in patients with type I diabetes (see doi: 10.1136/bmj.j1285). Prior studies have suggested that pumps work better, but it may have been that those patients on pumps had received more intensive training and education than those on multiple daily injections. So, this study looked at patients given similar education, finding that the benefits of education/training outweighed the advantage of using the continuous subcutaneous insulin infusion (the pump) over multiple daily injections.



— 317 adult participants in the UK from multiple sites ​with type I diabetes were randomized to insulin pump therapy versus multiple daily injections

— Both received structured education: 267 attended one week DAFNE skills training courses (Dose Adjustment for Normal Eating), with a further visit at 6 weeks. This training stresses flexible dose adjustments according to eating, physical activity, and blood glucose level, and was slightly different for those on multiple daily injections vs pumps, to emphasize the specific use and problems with each.

— Mean age 41, 60% male, 91% white, BMI 27, mean duration of diabetes 18 years, 55% with macrovascular complications/43% retinopathy/7% neuropathy/19% nephropathy, 12% with at least one episode of severe hypoglycemia in the past year, mean hemoglobin A1c 9.1 with a range 5.7 to 16.7 and only 9% had a hemoglobin A1c < 7.5%

— Main outcome: changes in hemoglobin A1c at 2 years. Secondary outcomes included body weight, insulin dose, and episodes of moderate or severe hypoglycemia. They also looked at quality-of-life and treatment satisfaction



— Mean change in hemoglobin A1c at 2 years:

–decreased 0.85% with pump treatment

–decreased 0.42% with multiple daily injections

— with adjustment for missing values etc, the A1c difference was 0.24% between the therapies, which is neither clinically nor statistically significant (0.5% being considered clinically significant)

— on a per protocol analysis, the mean difference favoring pump treatment was 0.36%, which did have a p=0.02, still not clinically significant.

— But at 24 months, combining both treatment groups, there was a hemoglobin A1c decrease of 0.54%. In those with an A1c initially >7.5%, the A1c decrease was 0.64%. These decreases, presumably attributable to the education and training prior to beginning each of the drug regimens, were clinically significant.

— secondary outcomes:

— hypoglycemia: 49 episodes in 25 patients over 24 months, did not differ between groups. The incidence of severe hypoglycemia decreased by about half for both groups as compared to baseline.

— No statistically significant difference in body weight, but there was a slight increase in HDL cholesterol and decrease in total cholesterol in both groups without a difference. Insulin dose decreased in both treatments, a little greater in those on the pumps (0.07 IU/kg). No difference in the odds of proteinuria.

— Diabetic ketoacidosis: this was greater in the pump group compared to the multiple daily injections group (17 versus 5 episodes), most related to infections, and 18% by technical failures in those using pumps

— psychosocial questionnaires: no difference between groups in generic quality-of-life status instrument. Improvement in both groups in the overall diabetes-specific quality-of-life questionnaire, though this was greater in the pump group though not always reaching statistical significance. Pump users showed greater improvement in treatment satisfaction as well as more dietary freedom and less daily hassle at both 12 months in 24 months

— other findings: those on pumps had twice the number of contacts with diabetes professionals, especially during the 1st year. There also were more face-to-face contacts and of longer duration in the 2nd year of the study.



— Pumps are used less frequently in the UK, an estimated 6% of type I diabetics use pumps there versus 40% in the US (which may be related to differences in the medical cultures between the 2 countries, with us going more quickly/easily to high tech fixes).

— Pumps are clearly more expensive than multiple daily injections: the pumps cost £2500 in the UK plus an additional £1500 for consumables (cannulas, reservoirs, batteries). And this does not include the increased number of office visits noted above.


— As per the authors, “These results do not support a policy of using insulin pumps in  those with poor glycemic control until the effects of training on participants level of engagement in intensive self-management have been determined”.  I personally support a strong effort to encourage a healthier lifestyle for both type 1 and type 2 diabetics (and pretty much everyone else), for its myriad of positive health effects.  However, diabetes raises particular challenges, since dosing of insulin in particular is so dependent on consistency in diet/exercise as well as on other events that change insulin effectiveness (eg infections, which increase insulin resistance). There may certainly be some advantages of the pump in some patients, with the potential for having more variations in life (different foods, even a small piece of cake on a birthday; doing less exercise some days when not feeling well or the weather is bad; UTIs, etc) and more flexible dosing to compensate. But this study in type 1 diabetics does point out the primacy of structured education to improve glucose control, and then considering technological fixes in some cases on an individual basis. And I think the lessons are more broadly applicable to type 2’s and beyond…​

Primary Care Corner with Geoffrey Modest MD: Review of diabetes care guidelines 2017

4 Apr, 17 | by gmodest

By Dr. Geoffrey Modest
The annual update of the American Diabetes Assn Standards of Medical Care in Diabetes had a few changes over last year’s (​see Diabetes Care, Janurary 2017; 40, suppl 1). I will highlight those changes (see the text for the overall recommendations).
promoting health and reducing disparities in populations (a major update): 
–they acknowledge (finally) that psychosocial care is important in the overall care of diabetics, including self-management, mental health, communication, and life-stage considerations. 
–specific recommendations to look at the patients’ social context (including assessing food insecurity, housing stability and financial barriers) and make use of local community resources and support for self-management​
classification and diagnosis of diabetes
–having a baby ≥ 9 pounds was bumped from the list as an independent risk factor for development of prediabetes or Type II diabetes. and for women with gestational diabetes, they extended the earlier time to check postpartum from 6-12 weeks to 4-12 weeks with oral glucose challenge test
comprehensive medical evaluation and assessment of comorbidities (a new section):,
–recommends a patient-centered communication style using active listening; eliciting patients preferences and beliefs; and assessing literacy, numeracy, and potential barriers to care (but not mentioning motivational interviewing.  perhaps next year???). 
–they also recommend “considering” screening for anxiety in patients who exhibit anxiety or worries, “consider” annual screening for depression as well as at the diagnosis of diabetic complications, consider screening for eating disorders (esp if hyperglycemia and weight loss are unexplained), annual diabetes screening for patients on atypical antipsychotics
lifestyle management:This section was renamed to focus on the importance of lifestyle management. There are few changes this year:
— nutrition therapy was updated to those on flexible insulin therapy to include counting fat (in particular increasing mono-unsaturated fats which may improve glucose metabolism and lower heart disease risk) and protein (protein seems to increase insulin response without increasing plasma glucose concentrations), as well as counting carbohydrates, to reflect the effect of these dietary factors on insulin dosing and blood glucose levels. They also suggest eating foods rich in long-chain omega-3 fatty acids, though the beneficial role of dietary supplements is not supported by the evidence.
— they recommend that prolong sitting be interrupted every 30 minutes with short bouts of physical activity, such as briefly standing, walking, or performing other light physical activities..
— they also added a recommendation stressing the importance of balance and flexibility training in older people
— they highlight the importance of the psychosocial issues: they should be integrated with a collaborative, patient-centered approach and provided to all people with diabetes, with the goal of optimizing health outcomes and health-related quality of life.
— Consider screening for cognitive impairment and depression in those >65 yo
Prevention or delay of type II diabetes
— more emphasis on screening for prediabetes using an assessment tool or informal assessment of risk factors. Patient should receive intensive behavioral lifestyle interventions
— metformin should be considered as preventive therapy in those with prediabetes especially if the BMI is >35,if they are <60 years old, or in women with prior gestational diabetes and/or those with rising A1c despite lifestyle intervention.
— There is also a recommendation that vitamin B12 levels be measured periodically, given the newer data linking metformin use to vitamin B12 deficiency
Glycemic targets
— they redefined clinically significant hypoglycemia as a glucose <54 mg/dl (3.0 mmol/L)
— no significant changes in the overall A1c goals, noting that a reasonable goal for many nonpregnant adults is <7%. Consider <6.5% for selected individuals (e.g. those with short duration diabetes, diabetes that is treated with lifestyle or metformin only, long life expectancy, no significant cardiovascular disease) who can achieve it without significant hypoglycemia or other adverse effects. And less stringent goals such as <8% may be appropriate for those with a history of severe hypoglycemia, limited life expectancy, advanced microvascular microvascular complications, extensive comorbid conditions
obesity management
— for overweight and obese patients with type 2 diabetes who are ready to lose weight, there should be a combination of diet, physical activity, and behavioral therapy to achieve >5% weight loss. Such an intervention should be high intensity (16 sessions in 6 months), focusing on a 500-750 kcal/d energy deficit. It’s okay to use very low-calorie diets (<800 kcal per day) in carefully selected patients, monitored closely, for a three-month intervention only
— weight loss medications may be effective as an adjunct in those with a BMI >27
— metabolic surgery (a.k.a. bariatric surgery) “should be recommended” in patients with a BMI >40 (BMI >37.5 in Asian Americans), and in adults with BMI 35-40 (32.5-37.4 in Asian Americans) when hyperglycemia is inadequately controlled. There needs to be long-term lifestyle support and routine monitoring of micronutrient and nutritional status after surgery.
Pharmacologic approaches
— consider empagliflozin or liraglutide in patients with established cardiovascular disease to reduce mortality risk
— they comment on the noninferiority of basal insulin plus glucagon-like peptide 1 (GLP-1)  receptor agonist as compared to basal insulin plus rapid acting insulin
— they modified their therapy pyramid to include the costs, especially since the cost of insulin has gone up so dramatically (mean AWP of > $300 per vial). However, they continue suggesting that after metformin monotherapy, one could choose any of the following agents without specific preference: sulfonylurea, thiazolidinedione, DPP-4 inhibitor, SGLT2 inhibitor, GLP-1 receptor agonist, or basal insulin.
Cardiovascular disease and risk management
— for patients without albuminurea, can use any of 4 classes of antihypertensives: ACE inhibitors, angiotensin receptor blockers, thiazide-like diuretics, or dihydropyridine calcium channel blockers. The general target blood pressure is <140/90 mmHg (as with JNC8), though lower target such as 130/80 may be appropriate for those at high cardiovascular risk and this can be achieved without undue treatment burden
— lipid management: they do focus on lifestyle modification. For patients <40 years old with additional atherosclerotic risk factors, consider using moderate or high intensity statin along with lifestyle changes. In those 40-75 years old without additional atherosclerotic risk factors, consider using moderate or high intensity statin, high intensity statin if they have additional atherosclerotic risk factors. And consider using a moderate intensity statin in those >75 years old, and high intensity statin if additional risk factors. Also should consider adding ezetimibe to moderate intensity statin in those with acute coronary syndrome and LDL >50 and those who cannot tolerate a statin.
— Consider aspirin therapy for primary prevention in those with diabetes, including most women and men > 50 years old with at least one additional cardiovascular risk factor but who are not at increased risk of bleeding.
— they changed the target for pregnant women with diabetes and chronic hypertension, suggesting a blood pressure goal of 120-160/80-105 mmHg. They also highlight the increased risk of retinopathy in women with pre-existing diabetes who are planning pregnancy or are pregnant
Microvascular complications
— not much change overall. They do state that if there is no evidence of retinopathy and the glycemia is well-controlled, ophthalmologic exams can be done every 2 years. They do reinforce that retinal photography is an acceptable screening tool but is not a substitute for comprehensive eye exam. For neuropathy, see recent blog specifically on those recommendations.
–personally, I have always felt that diabetes management is probably the most difficult of “medical” diseases, in part because of the complex array of psychosocial issues at all stages of diabetes (and beginning in pre-diabetes), and in part because of the necessity yet complexity of working with patients to make major, consistent changes in lifestyle (diet, weight control, exercise, etc), more difficult in a society which does not prioritize or even clearly articulate these generally healthful patterns in a meaningful way: witness how difficult it was to deal with trans fats, where the data against them have been compelling for decades; the push-back, specifically from the perspective from industry have been shown to be minimal/insignficant; and many in public health have gone against the tide/status quo over years to remove trans fats formally from the diet, still with incomplete success.
–it is pretty remarkable to me that the psychosocial issues have taken so long to bubble up to the being recognized and emphasized, along with suggestions for action. the suggestions are still pretty reserved: i personally feel that we should always have been looking at patient education as a 2-way street (and the backbone of treatment), incorporating patients values and understanding, and providing motivation for lifestyle changes mainly by involving the patient in the discussion as the main driver (vs lecturing them on diet and exercise). and given the diverse perceptions on what diabetes is and the potentially devastating effects of its complications, i personally would do more than just “consider” screening for depression and anxiety…
— as per many of my blogs, I remain quite impressed with the GLP-1 receptor agonists, which can be given as a simple once a week injection, and I regularly prescribe them now as my 2nd medication after metformin. They do not cause hypoglycemia, usually are very well-tolerated, help with weight loss, and I have seen quite dramatic effects. Last week I saw 2 patients on long-standing metformin and insulin, and when a GLP-1 agonist was added, their A1c in both cases went from about 9.5 to 6.5, in one case with the patient stopping their insulin. And they have the benefit of likely decreasing cardiovascular events (eg, see​ ). A recent report on empagliflozin found benefit for those with established cardiovascular disease, though I have concerns about the quality and conclusions of that study (see )
— these recommendations still do not deal with my concerns about using A1c as the target of therapy, which the FDA has allowed since the mid-1990s. Some medications that lower A1c pretty well have significant serious adverse reactions (the most obvious example being rosiglitazone and its attendant increase in cardiovascular disease), whereas others lower cardiovascular risk (such as the GLP-1 agonists), and others seem to be pretty neutral (although a lot of these are newer agents and we don’t really know their long-term adverse consequences). So, I do disagree with the approach of ADA that all of the second-tier drugs, after metformin, are equal options.
see here for many blogs on diabetes
see here for recent blog on B12 deficiency in diabetes
see for blog on ADA recommendations for neuropathy sent out a few days ago but not yet posted

Primary Care with Geoffrey Modest MD: Diabetic peripheral neuropathy, and more drug company shenanigans

4 Apr, 17 | by gmodest

By Dr Geoffrey Modest

There was a recent systematic review in the journal Neurology on pharmacotherapy for diabetic peripheral neuropathy pain and its effect on quality of life (see doi. org/ 10. 1212/ WNL. 0000000000003882). This appeared right after the recent guidelines from the American Diabetes Association (see link below), but had several eye-opening differences.


— the researchers updated a 2014 systematic review of 57 studies (see  Griebeler ML. Ann Intern Med 2014; 161:639, and prior blog on this  here) with 24 additional published studies and 25 unpublished studies. All trials were published between 1990 and 2015. The number of participants range from 20 to 804. They referenced the website clinical on March 9, 2016 to get data on unpublished studies.


Results, from placebo-controlled RCTs (those with moderate or large effect are highlighted):

— for anticonvulsants:

   — pregabalin, 16 RCTs, small effect, with low strength of evidence (SOE)

   — gabapentin, 5 RCTs, not effective, low SOE

— serotonin-norepinephrine reuptake inhibitors/antidepressants:

   — duloxetine, 7 RCTs, large effect, moderate SOE

   — venlafaxine, 2 RCTs, large effect, moderate SOE

   — tricyclic antidepressants, 4 RCTs, moderate effect, low SOE

— opioids

   — typical opioids (eg, oxycodone), 4 RCTs, not effective, low SOE

   — atypical opioids (tramadol, tapentadol), 5 RCTs, moderate effect, low SOE

— topical agents

   — capsaicin cream 0.075%, 5 RCTs, not effective, low SOE

— other agents

   — dextromethorphan, 3 RCTs, not effective, low SOE

   — mexiletine, 5 RCTs, not effective, low SOE

   — botulinum toxin, 2 RCTs, moderate to large effect, low SOE

–there were not enough data to evaluate the effect of these medications on quality-of-life.

–adverse effects: varied by agent. Dropout rates in the studies varied from 2.5% to 70% for oral agents, very low for topicals



— Unfortunately, few of the studies extended beyond 3 months (mean follow-up was 8.8 weeks with a maximum of 22 weeks), limiting their clinical utility some.

— The presumptive difference between the typical and atypical opioids as above is that the atypical ones have activity as norepinephrine reuptake inhibitors as well as mu agonists, with moderate effectiveness (vs oxycodone, a mu agonist, which was no more effective than placebo)

— I was a little surprised that capsaicin cream had no benefit, given the initial studies. And I have had patients do pretty well, with essentially no adverse effects. But, even if this is mostly a placebo effect, it is a pretty harmless med and if it works…

— The American Diabetes Association review  gave primary billing to pregabalin and duloxetine (see here  )

— The ADA review also did find a small benefit for oxcarbazepine, with low SOE, but not for other anticonvulsants such as topiramate, valproic acid, or lamotrigine.

— So, why are the above conclusions so different from those of the American Diabetes Association?  Of the 25 additional studies in which were unpublished, 18 were actually completed. The researchers were able to include 7 of these 18 completed studies. Of note, though pregabalin seemed to be effective, with a moderate reported effect size, when the 6 additional unpublished studies were included (all with negative results), the effect size decreased to “small”. In fact an analysis which included these negative studies found that the number of patients who would need to be treated with pregabalin for one patient to achieve a 50% pain relief was 7.7, so pretty few patients seem to benefit from pregabalin (Finnerup NB. Lancet Neurol 2015; 14:162)

–I suspect that the reason that gabapentin was considered ineffective but pregabalin not is mostly related to the serendipity of the studies done. Their anticipated physiologic effect is essentially the same, though gabapentin may be more slowly and erratically absorbed. In both cases there have been published studies suggesting positive efficacy of the medications, but on digging a tad under the surface, there are several large unpublished studies for each finding no benefit.  My guess is that they really do have similar and probably low levels of effectiveness. As with most drugs, some patients may benefit more than others. Is this from genetic differences in receptors or other drug mechanism?  Or maybe placebos work better in some people when they have mild to moderate adverse effects, and these meds both certainly do…(I have never seen anything written on this, but I suspect that meds with more systemic adverse effects may have more placebo effect as well, since some patients may internalize that meds with adverse effects must be stronger meds???…).The recent negative study of pregabalin in sciatica does also raise the question of its overall utility (see here  ). So, I’m not convinced that we should use pregabalin (only available as an expensive brand-name drug) over gabapentin (widely available generic), if we decide to use one of these types of drugs.

— One important finding in the current Neurology review was that venlafaxine, available as an inexpensive generic, was found to have a large effect size, with moderate SOE. This was also found for tricyclic antidepressants, though with moderate effect size. I am a bit hesitant to endorse venlafaxine as the single best agent (largest effect size of meds, somewhat larger than duloxetine which has no generic), since there were only 2 studies published on it. But I will add it to my short list of initial meds to try, along with tricyclics.

–so, bottom line here: the American Diabetes Association’s top 2 drugs, pregabalin and duloxetine, have both been knocked down by this more extensive Neurology review….. and, I think all systematic reviews should include looking at to find potentially pivotal negative studies which have not been published.


But, I think there are a few pervasive generic issues that this systematic review brings up:

–one of the biggest issues raised by this systematic review is: what should we trust in the medical literature? Clearly meta-analyses and systematic reviews can come to different conclusions based on their study inclusion criteria (see here for some of my thoughts on this). But additionally, we have moved to a place where the majority of large-scale medical studies are industry-funded. And they have a very clear bias to publishing the most positive results and putting the most positive spin on their results. And negative results are often not reported (this was the reason the feds required new studies to be logged into when they are initially approved, a law enacted in 1997). And the finding above about pregabalin is a case in point (a similar issue prevailed in the initial gabapentin studies, where there were also large, unpublished negative studies).  So, though pregabalin was highlighted in the recent Am Diabetic Assn recommendations 2 weeks ago, when the unreported negative studies are included, it really does not look so good…

–And,  as per here  companies often do not conform to government regulation/requirements, including rarely providing post-marketing data as required by the feds (their compliance rate was on  the order of 15%!!!!!). And, and, and, this whole situation is very likely to get whole lots worse with Trump, as the FDA moves to earlier licensing of drugs and medical devices, with less rigorous attention paid to their actual clinical benefit. So, the really difficult challenge is “what should I believe in the medical literature??”  I’m not sure how to answer that. But we need to be diligent, skeptical, and even less ready to jump on the bandwagon for new meds and medical devices. (One of my real purposes in writing these blogs is to critique studies or guidelines, and to assess their real generalizability and utility in the context of primary care clinical practice. And I hope they provide some context to help assess applicability of the studies/guidelines…..)


Primary Care Corner with Geoffrey Modest MD: creatinine increases after ACE/ARB may not be so good

15 Mar, 17 | by EBM

By Dr. Geoffrey Modest

A recent article in the BMJ challenged the long-held belief that ACE inhibitor/ARB related increases in creatinine were actually renoprotective (see Schmidt M. BMJ 2017;356:j791).


  • Observational study of 122,363 patients starting treatment with ACE inhibitor (ACE-I) or ARB from 1997 to 2014
  • They assessed the rates of end-stage renal disease, myocardial infarction, heart failure, and all-cause death among patients whose creatinine increased 30% or more after starting treatment, and also assessed the effect of each 10% increase in creatinine above the patient’s baseline.


  • 2078 patients (1.7%) had a creatinine increase of 30% or more.
  • Comparing those with a creatinine increase of > 30%, vs those with < 30%:
    • 56% female vs 46%
    • Median age 68 vs 63
    • Myocardial infarction in 10.5 vs 4.5%
    • Heart failure in 19 vs 4.8%
    • Arrhythmia in 17.2 vs 6.8%
    • Peripheral arterial disease in 6 ​vs 2.5%
    • Stage 3b CKD in 6.9 vs 3.7%/stage 4 in 2.0 vs6%
    • Beta blockers in 23.7 vs 17%
    • Loop diuretics in 28.6 vs 7.2%
    • Potassium sparing diuretics in 8.8 vs 2.0%
    • NSAIDs in 34.0 vs 23.5%
    • Underweight in 2.3 vs 0.9%, healthy weight in 26.9 vs8%, overweight in 34.5 vs38.4%, and obesity 29.0 vs  33.4%
    • So, basically those with greater creatinine increases had more baseline characteristics associated with increased morbidity/mortality
  • Creatinine increases of 30% or more were associated with (adjusted for age, sex, calendar period, socioeconomic status, lifestyle factors, chronic kidney disease, diabetes, cardiovascular morbidities, and use of other antihypertensive drugs and NSAIDs):
    • 43 times the rate of end-stage renal disease, incidence rate 3.43 (2.40-4.91)
    • 46 times the rate of myocardial infarction, IR 1.46 (1.16-1.84)
    • 37 times the rate of heart failure, IR 1.37 (1.14-1.65)
    • 84 times the rate of all-cause death, IR 1.84 (1.65-2.05)
  • There was a greater increase for all outcomes as creatinine went from an increase of <10%, to 10-19%, to 20-29%, to 30-39%, and to > 40%
  • These results were consistent across calendar periods, subgroups of patients, and among those continuing to use ACE inhibitor/ARB’s
  • There were much more dramatic increases in the rate of renal failure during the 1st year after starting ACE-I/ARBs (12.2-fold increase) versus in the 2nd year (3.7-fold) versus 2nd to 5th year (1.7- fold), but then increase to 2.5-fold from 5 to 10 years. However the numbers were small and the trend was nonsignificant. However, there were similar trends for heart failure and mortality which were significant. Heart failure was initially 1.9-fold increase in the 1st year but then settled in at 1.5-fold increase.


  • One major concern is that only about 10% of patients receive the recommended monitoring of serum creatinine soon after starting ACE-I/ARBs and only 20% of those with an increase of >30% discontinue the drugs as is recommended.
  • It has been widely held that larger increases of creatinine after taking these medications (up to 30%) were in fact renoprotective, supported theoretically/mechanistically that by decreasing intra-glomerular pressures, we were sparing the fragile glomeruli frombarotrauma. These data were not terribly rigorous. For example, there was a small study (Apperloo AJ. Kidney Intl 1997; 51(3): 793), which did find that in 40 nondiabetic patients with impaired renal function prior to therapy, those with a greater GFR decline after ACE-I had more stable renal function, and this decline was completely reversible after stopping ACE-I therapy at 4 years.) And this study has been cited in subsequent reviews as clear evidence that the higher the creatinine increase (up to 30%) the better…
  • The benefits of the study are its huge size, its real-world outcomes data and the fact that it represented the general UK population in terms of age/sex/ethnicity, the fact that they only looked at patients who had at least one year of being continuously in the registry before they were started on an ACE-I/ARB, and that they had long-term follow-up until the 1st diagnosis of end-stage renal disease, myocardial infarction, heart failure, and all-cause mortality.
  • However, the negatives of the study are that the patients who had more significant creatinine increases were clearly sicker and had more inherent likelihood of getting these clinical endpoints. Specifically, these patients were older, had more underlying chronic kidney and heart disease, and had more drugs that were potentially nephrotoxic. In particular the use of potassium sparing diuretics suggests the possibility that these patients had more severe hypertension or heart failure requiring these drugs (and the study did not stratify the degree of these conditions at baseline). The increased use of NSAIDs might signal that these patients had more pain, were less ambulatory, perhaps more overweight (though the BMI of those with more a bump in creatinine was actually lower, the specific individuals who went on to renal failure may well have been those with a higher BMI and on NSAIDs, but these data not available), and were less able to have a healthy lifestyle, which has repeatedly been associated with increased morbidity/mortality. And, though they did mathematically model to compensate for the array of potential adverse biases, there was such a divergence in the baseline characteristics of the 2 groups (>30% versus <30% creatinine increase) that I do not trust this mathematical manipulation to compensate for the real potential biases between the groups. (In addition, they did not comment on the underlying clinical conditions of the patients comparing those with <10% increase in creatinine vs those with >30%, but only for those <30% vs >30%)
  • It was also notable that pretty much all of these outcomes were much more dramatic in the 1st year after starting ACE-I/ARBs, suggesting that we should be doing increased surveillance particularly in that 1st
  • There are some perhaps relevant prior studies which suggest that increased creatinine is associated with cardiovascular disease, found in patients with mild to moderate renal dysfunction. However, these were patients with intrinsic renal disease as opposed to medication-induced increases of creatinine. So, not sure this is directly applicable to ACE-I/ARB-induced creatinine increases, but the above results are consistent with this.
  • We also did not know the levels of proteinuria of these patients, so there could be an important unaccounted for bias here. The studies suggesting the renoprotective effect of ACE-I/ARBs in diabetics found renal protection in those with proteinuria, even at low levels of albuminuria.

There have been other studies showing that patients with very significant proteinuria, especially those with greater than 1 g of albumin per day, do have renal protection by ACE inhibitors (see GISEN group. Lancet 1997; 349: 1857, which found that in 352 nondiabetic patients with proteinuria of 3 g of more than 24 hours, ramipril led to significant decreases in proteinuria as well as the rate of GFR decline, and a subsequent study of 186 patients with 1 to 3 g of proteinuria also had renal protection, but to a lesser degree, see Ruggenenti P. Lancet 1999; 354: 359.)

So, this article does give some pause. Perhaps our model of renoprotection is not so accurate. It is notable that the new JNC 8 guidelines do not recommend using ACE-I/ARBs as the primary treatment for hypertension in diabetic patients. Also the new American Diabetes Association guidelines (blog to come out soon) comment that those with diabetes should be treated with ACE-I, ARBs, thiazides, or dihydropyridine calcium channel blockers (no preference, except they recommend ACE-I/ARBs in those with albumin to creatinine ratio is greater than 300 mg/g, level A recommendation, or in those with 30 to 300 mg/g, level B recommendation). So, what is the bottom line here? I am really not sure, pending other studies (and the best being an RCT). But this study does bring up the thinness of the prior assertions that those with ACE-I/ARB induced creatinine increases do better, and also reinforces the importance of checking creatinine (and lytes) after starting these meds, and stopping them if >30% increase. Otherwise, I am hesitant to change current practice.

Primary Care Corner with Geoffrey Modest MD: Diabetic Neuropathy Guidelines

14 Mar, 17 | by EBM

By Dr. Geoffrey Modest

The American Diabetes Association just came out with their position paper on diabetic neuropathy (see DOI: 10.2337/dc16-2042). I will limit my points to type 2 diabetes, though type 1 is covered in this paper

Summary of points:

  • Diabetic neuropathy is a diagnosis of exclusion: diabetic patients may well have non-diabetic causes of neuropathy that should be pursued.
  • 50% of diabetic neuropathies are asymptomatic. It is important to assess for them, for example, to decrease the likelihood of significant foot trauma, improve symptoms and quality of life, and decrease sequelae
  • Prevention of neuropathy:
    • Data are largely for distal symmetric polyneuropathy (DSPN) and for cardiovascular autonomic neuropathy (CAN).
    • Best evidence is for those with type I diabetes where it is important to optimize glucose control as early as possible: 78% relative risk reduction with enhanced diabetes control
    • Risk reduction seems to be less in glucose control with type 2 diabetes, perhaps in part reflecting the different pathophysiology and comorbidities: type 2 diabetes tends to be associated more with overweight, polypharmacy, older age; but also many patients with type 2 diabetes have been prediabetic or diabetic for many years prior to diagnosis. In fact 10-15% of newly diagnosed diabetes already have evidence of DSPN.
  • Distal symmetric polyneuropathy (DSPN)
    • Most common (75% of all neuropathies). Defined clinically by symptoms or signs. Electrophysiologic testing or neurology referral are rarely needed.
    • 50% of people have DSPN after 10 years of disease, and is associated with levels of glycemia, height (perhaps as a proxy of nerve length), smoking, blood pressure, weight, and lipids
    • Those with predominantly small-fiber neuropathy present with pain, burning or tingling feeling, sometimes with a shooting sensation. There may also be hyperalgesia. And this may be found in 10 to 30% of patients just with impaired glucose tolerance. Those with large fiber involvement have numbness, tingling without pain, and loss of protective sensation
    • DSPN is associated with foot ulceration/amputation risk (important to assess regularly and refer to podiatry early), Charcot neuro-arthropathy, unsteadiness and falls (should assess gait/balance, though minimal data to support), and quality of life (DSPN can really affect quality of life, and is associated with depression, anxiety, medication nonadherence)
    • Patient should be assessed annually, those with type I diabetes should be assessed starting 5 years after the diagnosis. Consider assessing those with glucose intolerance as well. Assessment should include temperature or pinprick sensation (small fiber function), vibration sense with a 128 Hz tuning fork, proprioception, ankle reflexes, and 10-g monofilament (large fiber function), and the 10 g monofilament also helps assess risk for ulceration and amputation
    • Important to rule out the myriad of other causes of neuropathy, including vitamin B12 (see blog which notes the overall increased incidence of B12 deficiency in diabetics) as well as infections (HIV, hepatitis B, Lyme), thyroid disease, paraproteinemia, alcohol or medications, heavy metal poisoning/work-related exposures, etc.
    • Symptom management: consider pregabalin or duloxetine as the initial​ approach. Though gabapentin may also be used (they do comment that pregabalin has a more linear, dose-proportional absorption and more rapid onset of action). They also note that tricylcic antidepressants (TCAs) are effective but beware of adverse effects. There seems to be some efficacy for the selective norepinephrine/serotonin reuptake inhibitor venlafaxine (dose 150-225 mg/d), mechanistically similar to duloxetine. Opioids should be avoided, given the risks of addiction, as either first- or second-line agents. However, tapentadolextended release, which has analgesic effects both through the m-receptor and noraderenaline reuptake inhibition, is FDA-approved, though there are systematic reviews/ meta-analyses which challenge its effectiveness. And some patients do seem to respond to adding low doses of these opioids in combo with the above agents).
  • Autonomic Neuropathies
    • Cardiovascular autonomic neuropathy (CAN)
      • May be present prior to a formal diagnosis of diabetes, ​is found in up to 60% of patients after 15 years, and is an independent risk factor for cardiovascular mortality (2+ fold increased risk, controlling for other risk factors), arrhythmia, silent ischemia, any major cardiovascular event, and myocardial dysfunction.
      • May be asymptomatic early and detected only by decreased heart rate variability with deep breathing, esp with EKG monitoring (see which discusses ways to measure CAN), but can include symptoms of light-headedness, weakness, palpitations, fainting/syncope. Exam may show resting tachycardia, or orthostatic hypotension without compensatory increase in pulse.
      • Symptom management:
        • Optimize glucose control (to prevent or delay CAN, though this is most evident in type 1 diabetes, but some benefit in studies with type 2), reinforcing lifestyle interventions both in prediabetics and diabetics prior to developing CAN. For those with orthostatic hypotension, can use both nonpharmacologic treatments (exercise, assuring adequate fluid intake/volume repletion) and meds (fludrocortisone, midodrine)
      • Gastrointestinal neuropathies can be anywhere in GI tract, from esophageal dysmotility to gastroparesis to lower GI symptomsof diarrhea, constipation, incontinence
        • Gastroparesis
          • ​May be present in 1% of type 2 diabetics, from a community-based study (higher in type 1 diabetes). Can affect glucose variability and unexplained hypoglycemia because of changes in food absorption. [my experience suggests that gastroparesis may well be more common than this]
          • ​Consider checking for symptoms in those with other microvascular complications  (“C” recommendation), exclude other causes (g. opiates, GLP-1 agonists (grade “C” recommendation”) [and, I would add, considering decreasing dose of metformin, esp since 500mg once a day or even 250mg seems to add substantial clinical benefit], and can do gastric emptying studies to document (grade “B” recommendation”) [I would also add that gastroparesis is usually evident by history, and that it is probably useful just to try the nonpharmacologic and even pharmacologic therapies empirically]
          • Treatment includes eating multiple small meals/d, decreasing dietary fat (which also causes gastroparesis), decreasing other drugs than those mentioned above which can make it worse (g. anticholinergics, pramlintide and ? DPP-4 inhibitors), and one can prescribe metoclopramide, the only FDA-approved agent, though it is associated with extrapyramidal symptoms, acute dystonic reactions, akathisia, tardive dyskinesia, acute dystonic reactions) and is recommended to use for only 5 days (which is problematic for such a chronic condition, and I have patients on this agent for much longer, with frequent assessments by me for the above adverse reactions)
        • Urogenital neuropathies includes bladder and sexual dysfunction, the latter including erectile dysfunction (3x more common in diabetics, may involve combo of autonomic neuropathy, vascular disease, and I would add psych issues, such as depression, stress, etc.) and/or retrograde ejaculation in men and sexual dysfunction in women (decreased sexual desire, increased pain with intercourse, decreased sexual arousal, inadequate lubrication).
          • Bladder dysfunction should be assessed in those with recurrent urinary tract infections, pyelonephritis, incontinence, palpable bladder
          • Recommendations: consider screening men with other forms of neuropathy for ED (grade C) and women with other forms of neuropathy for lower urinary tract symptoms and sexual dysfunction (grade E)
        • ​Sudomotor dysfunction includes dry skin, anhidrosis, or heat intolerance, and occasionally gustatory sweating (food consumption, and occasionally just the smell of food, leading to sweating of head and neck area)
        • Other neuropathies:
          • Mononeuropathies: especially of median, ulnar, radial and common peroneal nerves. cranial neuropathies are rare but include cranial nerves III, IV, VI, VII and usually resolve spontaneously over months
          • Diabetic radiculoplexus neuropathy (also called diabetic amyotrophy): unilateral thigh pain, weight loss, followed by motor weakness. self-limited (though I have a type 1 diabetic patient with this in one of his shoulders)


  • They do promote pregabalinand duloxetine as their primary go-to’s. I personally do not use them till much later in the pyramid of meds, partly because they are relatively new agents (and the older ones have stood the test of time), partly because there are mechanistically similar drugs available (gabapentin and venlafaxine), partly because these are non-generic and quite expensive, and partly (e., a lot) because they require prior approvals from many insurers.​
  • In the vast majority of cases, I have prescribed tricyclic antidepressants with great success. Although amitriptyline is the one used the most overall, it has the most adverse effects. I prescribe either desipramine or nortriptyline, which work as well and with many fewer adverse effects (desipramine has the fewest, but nortriptyline is helpful to take at night if the patient has trouble sleeping). The usually effective doses are desipramine 25-50mg (occasionally 75), or nortriptyline 10-50mg. Not sure why, but the 2012 ADA guidelines (see Diabetes Care 35:2451–2458, 2012) found that there was no significant difference between amitriptyline, duloxetine and pregabalin, though the current guidelines seems to have booted TCAs off the top tier (they are generic with long history of use and knowledge of long-term adverse effects, both plusses, which does raise the question to me of adverse drug-company induced bias….)
  • Gabapentin is used a lot for neuropathy, though the studies have been mixed (and the drug company has been taken to task for withholding large, unpublished negative studies).  And in my limited experience, is associated with many adverse effects and requires a very slow titration up.

So, a pretty useful compilation of diabetic neuropathies, along with reasonable approaches (though there are no medications which actually treat the neuropathies, only ameliorate the symptoms). My own approach is that anyone with any mono or polyneuropathy should be checked for diabetes (for example, they limit the cranial nerve neuropathies to the facial and extraocular movement nerves, though I have seen a couple of diabetic patients with anosmia.) Also, they do not comment that it is not so uncommon in diabetics to have radiculopathies typically on the trunk which simulating zoster clinically and respond to the above meds.

See for a meta-analysis of the meds used for DSPN, finding that SNRIs, capsaicin, tricyclics and anticonvulsants work for short-term pain control (seemed that SNRIs and TCAs were best). Opiates were last by a fair margin.

Primary Care Corner with Geoffrey Modest MD: Fructose and NASH

7 Mar, 17 | by EBM

By Dr. Geoffrey Modest

A recent study found that fructose consumption and serum uric acid were independently associated with non-alcoholic steatohepatitis (NASH) in obese kids with non-alcoholic fatty liver disease (NAFLD), (see


  • 271 obese children (by BMI) with NAFLD were studied
  • NASH was diagnosed by biopsy, with a NAFLD score of at least 5, and by the fatty liver inhibition of progression (FLIP) algorithm (another algorithm for the diagnosis of NASH)
  • Fructose intake was determined by a food frequency questionnaire
  • Hyperuricemia was defined as a serum uric acid level >5.9 mg/dl


  • NASH occurred in 37.6% of the children
  • Mean age 11.5, 38% female, BMI 27, waist circumference 87cm, AST 48/ALT 62, uric acid 5.8, LDL 100/HDL 45, BP 112/68, TNF-a Of note, there were significant differences between those with NASH and those without, but only for: waist circumference, AST/ALT, total cholesterol (but not LDL, HDL alone), triglycerides, fructose consumption, and TNF- a
  • Hyperuricemia was found in 47% of the kids with NASH, vs 29.7% without NASH (p=0.003)
  • Adjusting for multiple measured confounders:
    • Uric acid level was associated with NASH, OR 2.49 (1.87-2.83), p=0.004
    • Fructose consumption was associated with NASH, OR 1.61 (1.25-2.85), p=0.001
    • These associations with NASH were independent of each other
    • Though, fructose consumption was still independently associated with hyperuricemia, OR 2.02 (1.66-2.78), p=0.01
    • These data on NASH were confirmed by using the FLIP algorithm


  • As noted in prior blogs, there seems to be a pretty consistent relationship between fructose consumption and uric acid levels, as was shown in this study. And there are data suggesting that dietary fructose can be part of the pathogenesis of NAFLD (induction of de novo lipogenesis, inflammation, insulin resistance). Studies in adults have found that hyperuricemia is associated with insulin resistance, type II diabetes and metabolic syndrome.
  • There are many dietary sources of fructose. The major ones for most people are table sugar (sucrose, a disaccharide of glucose and fructose) and high fructose corn syrup (in a surprising number of foods, as a very cheap and potent sweetener). Perhaps the “low-hanging fruit” here is sodas, consumed by very large numbers of people ( , finding that on average 26.3% of US adults consume at least one sugar-sweetened beverage daily, up to 41.4% in Mississippi, the highest of states, and that soda by itself was consumed by 24.5% of those 18-34 yo, and 47.4% in Mississippi). And a NHANES study (Welsh JA. JAMA 2010; 303(15): 1490) found that on average, 15.8% of calories came from added sugars, and that >25% of the patients got >25% of their total energy from added sugar. My experience is that it is easier to help people stop sodas and juices, substituting water, than other dietary interventions.
  • I am not sure why we focus so exclusively on BMI, since the data are pretty consistent over the decades that abdominal obesity is really the bad actor, more metabolically active and associated with inflammatory markers, diabetes/insulin resistance/metabolic syndrome and is independently associated with cardiovascular risk (BMI is not: its association is mediated by its association with other risk factors, such as blood pressure, lipids). Waist circumference is a much better, though not perfect, marker of visceral obesity than BMI. That being said, there is a pretty strong relationship (but not always) between BMI and waist circumference, especially in those with BMI >35. The most reasonable recommendations I have seen is to measure the waist circumference regularly, especially if the BMI is between 25-35. Although practically we should reinforce lifestyle changes in all patients with high BMI, independent of waist circumference, I think the patient should understand that for those with a high waist circumference, their cardiometabolic risk is even higher. See for the fuller argument

So, this study does add some important information: it confirms that both uric acid levels and fructose are associated with NASH in kids that they are associated with each other, but that they are also independent predictors (and there are several studies which show that decreasing fructose consumption, as in sodas, is associated with decreased uric acid levels. See blogs below). So, bottom line is that fructose consumption is bad and should be decreased, even if the uric acid level is just fine.

See for prior blog of fructose consumption in kids and cardiometabloic and weight improvements is one of 3 articles on NAFLD, highlighting an important role of fructose  more on the microbiome and hepatic changes with fructose

Primary Care Corner with Geoffrey Modest MD: vitamin B12 and diabetic autonomic neuropathy

23 Feb, 17 | by EBM

By Dr. Geoffrey Modest

A recent Danish study found that vitamin B12 deficiency was associated with diabetic cardiovascular autonomic neuropathy, CAN (see Hansen CS. J Diabetes Complic. 2017; 31(1); 202)


  • 469 type II diabetic patients were screened for CAN by several measures, as well as for peripheral neuropathy.
  • Mean age 59, 60% male, diabetes duration 10 years, 5% excessive alcohol consumption, 12% smokers, BMI 32, blood pressure 132/82, 6% on vitamin B12 supplementation (though 16% in those in the highest B12 quartile), 80% on lipid-lowering drugs, 75% on metformin, 4% on PPIs alone and 10% on the combination with metformin, 40% on insulin
  • CAN was measured after a 5-minute supine resting period:
    • Heart rate variability (HRV)
    • 3 tests assessing cardiovascular autonomic reflexes:
      • Lying-to-standing test
      • Deep breathing test (E/I ratio), a measure of heart rate variation during deep breathing [which is affected by an abnormality in the parasympathetic nervous system]
      • Valsalva
    • Peripheral neuropathy was measured electronically by vibration sensation


  • B12 level varied from the lowest quartile mean of 190 to the highest quartile of 486 pmol/l
  • Serum levels of B12 were significantly lower in those on metformin or proton pump inhibitors, p <0.001.
  • Higher level of B12 were significantly associated with a lower odds ratio of CAN, p=0.04
  • A 25 pmol/l higher level of vitamin B12, adjusted for age, sex, diabetes duration, and alcohol consumption, was associated with:
    • 6% lower level of CAN diagnosis, odds ratio 0.94 (0.88-1.00, p= 0.034)
    • An increase of E/I ratio of 0.21% (p= 0.038)
    • A decrease in resting heart rate of 0.25 bpm (p= 0 .025)
  • No association between B12 levels and decreased vibration/peripheral neuropathy


  • Cardiovascular autonomic neuropathy is very common in patients with type II diabetes, ranging in prevalence from 20 to 65% and increasing with length of diabetes. CAN is also an independent predictor of cardiovascular mortality and morbidity. But CAN may well be overlooked clinically until a patient is symptomatic, typically late in its course.
  • Vitamin B12 deficiency is also quite common in diabetics, with estimates from 2-33%, potentially mediated in part by the use of metformin through a not-so-well understood mechanism. This relationship is both metformin dose-dependent and treatment duration dependent, and may be measurable in as little as 4 months after the onset of use. In addition, the frequent use of proton pump inhibitors may decrease vitamin B 12 levels. Another potential and common mechanism for B12 deficiency in older patients is the age-associated decrease in several digestive enzymes, leading to the inability to liberate B12 from foods thereby decreasing its absorption (studies have found b12 deficiency in 10-25% of elderly, typically asymptomatic).
  • This was an observational study, therefore it is difficult to attribute causation. In addition, there is no compelling evidence that correcting B12 deficiency decreases the likelihood of CAN [one Indian population-based study of healthy elderly showed that B12 supplementation in those deficient led to normalization of decreased heart rate variability (see Sucharita S. Autonoom Neurosci 2012; 168 (1-2); 66)].
  • Also, the effect of B12 deficiency on CAN in the study was not particularly large. Part of this is that there were very few patients (0.6% of their population) who they defined as having vitamin B12 deficiency (that being below 125 pmol/l in this study, though many consider the cutpoint to be <148 pmol/l, which translates to <200 pg/ml), so the lowest quartile had lots of patients who were probably not actually B12 deficient. And the likely reason for the low B12 deficiency rates was that the standard clinical practice in that area was to check B12 levels in patients every other year. They did not test for methylmalonic acid or homocysteine, which might have been relevant in those with borderline B12 deficiency (35% had B12 levels between 125 and 250 pmol/l, though others consider borderline to be between 148 and 221 pmol/l​, or 200-300 pg/ml). Also, the fact that the effect was particularly evident for the E/I ratio with deep breathing suggests that a parasympathetic abnormality may predominate, and parasympathetic denervation is in fact typically the first abnormality in CAN, leading to increased sympathetic tone.
  • Of note, several different studies, but not all, show that those with peripheral neuropathy associated with B12 deficiency do improve with B12 supplementation, though the degree of improvement tracks inversely with both the extent and duration of disease.

So, my take on this is that given the clear importance of vitamin B12 for several aspects of health (neurologic, psychiatric, hematologic), and that some of these manifestations may be pretty subtle/very hard to detect early on, I personally think it makes sense to check vitamin B12 levels in the elderly as well as those on metformin and PPIs. And now, perhaps more so in diabetics overall, perhaps when they reach the ripe old age of 50 or so.

Primary Care Corner with Geoffrey Modest MD: Metformin in those with CKD, CHF, CLD

6 Feb, 17 | by EBM

By Dr. Geoffrey Modest

A systematic review from the VA synthesized data on use of metformin in patients with chronic kidney disease (CKD), congestive heart failure (CHF), or chronic liver disease (CLD) with hepatic impairment (see doi:10.7326/M16-1901). The goal was to assess all-cause mortality, major adverse cardiac events (MACEs), and other outcomes in patients with these underlying diseases (patients with these diseases having been the ones in whom the FDA initially had warned against using metformin).


  • 17 observational studies that included patients with CKD, CHF, or CLD with hepatic impairment were analyzed. These studies compared patients on diabetes regimens that included Metformin vs those that did not.


  • CKD
    • 6 studies were included, with sample sizes ranging from 1246 to 11,481 patients, median age ranging from 65 to 76. Only one study reported median daily metformin dose (1100 to 1900 mg in the different subgroups)
    • All-cause mortality
      • 22% lower for patients on metformin, HR 0.78 (0.63-0.96)
      • 2 studies reported CKD severity subcategories:
        • eGFR of 30 to <45 had neither benefit nor harm
        • eGFR of 30 to 60 had clear benefit of around 38%
        • eGFR <30 (one study) had neither benefit nor harm
      • MACEs:
        • 2 studies were evaluated, finding no difference in outcomes with metformin in the subgroups of patients with eGFR <45
        • And much, much more hypoglycemia in those on non-metformin-based regimens (specifically, glyburide or insulin)
      • CHF
        • 11 observational studies were included, with sample sizes from 346 to 13,930 patients, median age 55 to 77 years old. No studies included median metformin dose
        • All-cause mortality:
          • 22% lower for patients on metformin, HR 0.78 (0. 71-0.87)
          • 2 studies reported CHF severity subcategories:
            • One study looked at LVEF, finding that both an LVEF of 30-39% and LVEF<30% had a nonsignificant 13% decreased mortality; another study looked at patients with LVEF < 40%, finding a nonsignificant 21% decrease
          • MACEs:
            • The relative chance of readmission for CHF during follow-up was 13% lower for patients on metformin: HR 0.87 (0.78-0.97)
            • The relative risk for cardiovascular mortality was 23% lower in those on metformin (their figure shows that the three studies that looked at this found statistically significant improvement with metformin, but their overall summary said it was nonsignificant?????)
          • CLD
            • 3 observational studies included, with sample sizes ranging from 82 to 250 patients, median age 60-61. No studies reported median metformin dose.
            • All cause mortality:
              • The one study with low risk of bias (n=250) found significantly longer survival: HR 0.43 (0.24-0.78), regardless of cirrhosis severity. Post hoc analysis found a positive association between metformin and survival only in those with nonalcoholic steatohepatitis, though the number of patients in the other subgroups was small.
            • The other studies in those with moderate-to-high risk of bias showed a trend to lower all-cause mortality with metformin


  • Metformin is accepted as the first line drug for diabetes in the US and other countries that I know of. It is such a good and appropriate drug, given both its positive effects on diabetes (including its being weight-neutral or leading to weight loss) as well as data suggesting decreased cardiovascular disease and all-cause mortality. As a result, many clinicians have been using it despite FDA precautions/contraindications, with estimates that 20-30% of patients have been prescribed metformin who have had these precautions/contraindications. The FDA itself has been progressively relaxing these restrictions. In 2006 they removed CHF as a contraindication (though acute or unstable CHF remains a precaution). In addition in 2016, the FDA changed the warning for CKD to be based on eGFR instead of creatinine, making approximately 1 million additional patients with moderate CKD eligible to receive metformin. See blogs noted below for other studies promoting the use of metformin.
  • Most of the above conclusions were based on studies which had low strength-of-evidence, moderate risk-of-bias. However there was consistency in their finding that metformin therapy was associated with reduced all-cause mortality among patients with moderate CKD, CHF, or CLD; fewer CHF admissions among those with moderate CKD or CHF; and a much lower hypoglycemia rate among those with moderate CKD
  • There are other concerns about a meta-analysis such as above, including the fact that they don’t have data on doses of metformin for most studies, what other medications were being used in addition to metformin (the studies did not have rigorous details about which patients were on which other hypoglycemic medications), whether there was “confounding by indication” (where people might have been selected to be on or off metformin based on unknown individual clinical considerations), or even more than baseline data on metformin use in most studies (i.e. patients may have started on metformin but somewhere during the study had stopped it; or alternatively patients may have started off metformin but then put on it during the course of the study)

But, bottom line, these studies reinforce not just the safety of metformin in what had previously been considered concerning underlying comorbidities, but strongly suggest a significant benefit of metformin-based regimens. I.e., there really is an imperative to use metformin as the first-line therapy. We know through our clinical practice that metformin’s major adverse reactions are GI. This is clearly less the case in those on lower doses or if metformin is taken with meals. The somewhat skimpy data suggest that much of the benefit of metformin is conferred by much less than full doses (one oft-repeated quote is that about 80% of the benefit of metformin is by giving 1000mg vs 2000mg). My personal experience is that many people get much better glucose control on just 500mg once a day (which is my starting dose, though I leave it there if there is good control, as happens pretty frequently), and I even have a person on 250mg (1/2 tablet) because of GI intolerance, who seems to get benefit…

Relevant past blogs: is a systematic review of studies in patients with chronic kidney disease, finding for example no cases (as in, zero) of lactic acidosis in 70,490 patient-years on metformin​  which gives the updated FDA changes for metformin prescribing in those with CKD, with reference to a study of 813 patients with creatinine >6 who did NOT have increased mortality on low dose metformin (<= 500 mg/d), as well as the study finding that metfomrin induces changes in the microbiome, which leads to decreased insulin resistance.

Primary Care Corner with Geoffrey Modest MD: Microbiome 2

24 Jan, 17 | by EBM

By Dr. Geoffrey Modest

This is the second of two blogs on the microbiome, inspired by a recent review that highlighted several other health-related data besides the non-caloric artificial sweeteners (see Lynch SV. N Engl J Med 2016;375:2369).


  • ​The microbiome is huge, with 9.9 million microbial genes represented, as found from studying 1200 people in the US, China, and Europe. And it has >1000 species of microbes
  • Although the microbiome was previously felt to develop after birth, bacteria are found in the placentas of healthy mothers, in the amniotic fluid of preterm infants, and in meconium. And, the mode of infant delivery does influence postnatal microbial exposure: intravaginal delivery does seem to confer an infant microbiome taxonomically similar to the maternal gut and vaginally microbiota. Also the infant microbiome does become more similar to the adult one with the cessation of breast-feeding, and over the years bacterial diversity and functional capacity expand. The microbiome becomes less diverse in elderly, which could reflect coexisting conditions and age-related declines in immunocompetence.
  • Things that affect the microbiome include sex, age, diet, exposure to antimicrobial agents, changes in stool consistency, PPIs and other meds, travel, malnutrition, exercise (the effect of exercise on the microbiome is pretty clear in mice, not so clear in humans, since it is hard to sort out the effect of exercise vs different diets in those who exercise more). Also, host genetic features, host immune response, xenobiotics (including antibiotics), other drugs, infections, diurnal rhythms (see below), and environmental microbial exposures.
  • Clostridium difficile infections
    • This is probably the most advanced and practicable microbiome application. See for many studies and analyses. However about 90% of patients affected with severe, recurrent antibiotic-resistant C. difficile infections respond to fecal microbial transplants
  • Effects on immunity:
    • There are data that the infant microbiota at one month of age is significantly related to allergy in two-year-old children and to asthma in four-year-old children. Several of the products of the higher risk microbiota are associated with subclinical inflammation, which precedes childhood disease. Also other studies have found that children born by cesarean section, who do have differences in their microbiota, are more likely to develop type I diabetes, celiac disease, asthma, hospitalizations for gastroenteritis, and allergic rhinitis.
  • Obesity/metabolic syndrome/insulin resistance/diabetes
    • There are several studies finding that there are significant differences in the microbiome between obese and lean human subjects, with a decrease in Bacteroidetes and an increase in Firmicutes species in obese individuals. Studies have shown that taking microbiome samples from pairs of identical human twins, one lean and one obese, and placing them into genetically identical baby mice, have found that the mice with the microbiota from the obese twin develops more weight gain and more body fat, along with a less diverse microbiome, than those from the lean twin. Also, interestingly, women in their third trimester of pregnancy have an altered microbiome, which, when transplanted into mice, leads to more obesity, and that pro-obesity microbiome is more efficient in extracting energy from food [one common clinical issue with overweight/obese patients is that they may often eat much less than others but still do not lose weight, which has been shown in several studies, and attributed to their being more efficient in metabolizing foods. But perhaps this is mediated through the microbiome???]
    • Some proteins elaborated by E. coli stimulate glucagon-like peptide-1 (GLP-1) secretion, which could augment glycemic control in diabetics, where this hormone is less active than in nondiabetics. In addition, E. coli can elaborate peptide YY (produced in the ileum in response to feeding), which can activate anoxeretic pathways in the brain, mediating satiety.
  • Atherosclerosis/cerebral artery occlusion
    • There are pretty convincing studies that eating red meat leads to changes in the gut microbiota, which leads to increase production of trimethylamine-N-oxide (TMAO), which is a very strong risk factor for human atherosclerotic disease. And feeding meat to vegetarians does not increase TMAO until there are these microbiota changes from recurrent red meat diets. See blogs listed below for more details. Also, experimental data on mice show that cerebral arterial occlusion leads to 60% less damage in those with microbiota which are sensitive to antibiotics; mice given probiotics have less impairment after spinal cord injury.
  • Cancer
    • In mice, specific gut bacteria (most clearly shown for Bifidobacterium) enhance the efficacy of cancer immunotherapy, delaying melanoma growth. Human data has shown that certain microbiota species (B. Thetaiotaomicron or B. fragilis) can improve the effects of anti-tumor therapy targeting cytotoxic T-lymphocytes-associated antigen 4.
  • Autism
    • There are even some suggestive data that the microbiome may play a role in autism spectrum disorders. MIA mice, a maternal immune activation mouse model, exhibits autistic-like behavior, gut microbiome dysbiosis, increased gut mucosal permeability, and an increase in 4-ethylphenylsulfate (4EPS, a metabolite of gut bacteria). Injection of 4EPS into healthy, normal mice results in anxiety. And, feeding the MIA strain of mice a strain of Bacteroides fragilis normalized these adverse gut changes and decreased behavioral abnormalities, associated with decreasing circulating 4EPS levels. There are other neuropsych issues potentially related to the microbiome: gut bacteria can produce several neurotransmitters (eg norepinephrine, serotonin, dopamine, GABA, acetylcholine), and can change emotional behavior of mice (which seems to be related to central GABA receptor expression).
  • Other diseases with suggestive data of a linkage to microbiome dysbiosis include inflammatory bowel disease, kwashiorkor, juvenile rheumatoid arthritis, and multiple sclerosis. Also, in mice, stress leads to altered microbiota (less Bacteroides and more Clostridia), and in humans undergoing bariatric surgery, there are huge differences in the microbiome by either the Roux-en-Y gastric bypass or vertical banded gastroplasty, and this microbiome transplanted into germ-free mice leads to reduced fat deposition, suggesting that these microbiome changes themselves might play a direct role in decreasing adiposity (see Tremaroli V. Cell Metabolism2015; 22:228)​. And perhaps the changes in the microbiome, through the gut-brain relationship is part of the reason for the documented improvement in memory noted after bariatric surgery.
  • Diurnal rhythms (see Thaiss CA. Cell. 2014; 159: 514): the gut microbiota has diurnal variations that reflect feeding rhythms; humans with jet lag have dysbiosis; this jet lag leads to microbiome changes promoting glucose intolerance and obesity and are transferable to germ-free mice.


  • We should approach these studies on the microbiome with caution: some of the most impressive studies were done in animals in highly controlled conditions, and predictions in humans based on the studies is always fraught. For example, in general the use of probiotics in human adults has not shown as dramatic a response as found in rodents. (Although an interesting study of human neonatal probiotic supplementation in the first month of life was associated with a 60% reduction in the risk of pancreatic islet cell autoimmunity, a precursor to type 1 diabetes, before school-age). In addition, a stool sample may not be an adequate proxy for the microbial content of the entire GI tract. And, most of these studies have focused primarily on bacterial species in the microbiota, not taking into account the many other types of microorganisms found or their complex interactions.
  • One concern I have in general is our tendency towards reductionism. The microbiome appears to be a quite complex organ, composed of many different varieties of organisms which undoubtedly interact with each other in complex ways, and which are influenced by many known and undoubtedly unknown external cues (diet, antibiotic use, etc., etc.). So, for example, simply attempting to manipulate that microbiome through the introduction of one species or another of probiotics (i.e., our usual medical fix) may not deal with the complexity of this situation.
  • There have been a slew of other blogs on the microbiome over the years. See . One particularly interesting finding in one of the blogs was that one of metformin’s major action might be in its effects on the microbiome (see, which also reviews some of the TMAO data.
  • So, although I am pretty convinced of the importance of a healthy microbiome, it does seem to me that the major initiative should be around lifestyle changes overall: a healthy diet (and specifically one which is predominantly vegetarian), adequate exercise, perhaps adequate sleep (would be great to have more data on the effect of sleep patterns overall on the microbiome and if changing those patterns changes the microbiome), and minimizing exposure to unnecessary antibiotics (both in humans and in animals that make it into our food chain).

Primary Care Corner with Geoffrey Modest MD: Artificial Sweeteners Microbiome1

23 Jan, 17 | by EBM

By Dr. Geoffrey Modest

As mentioned in prior blogs, I think that the microbiome represents a very important mediator between the external environment and health/disease. A few recent articles supplement and strengthen this understanding. The first in a series of two is a study reinforcing the potentially deleterious effects of non-caloric sweeteners on the microbiome and health outcomes. The second (to be sent tomorrow) is a broader description of our understanding of the microbiome overall and its potential relationship to health.

​Non-caloric artificial sweeteners (NAS) were developed from the biological perspective that these potent sweeteners (more than 100 times sweeter than sucrose) are non-caloric and  are excreted unchanged; they should therefore be an important sugar alternative to help people lose weight and control glucose intolerance. Although a study done in the 1980s, prior to DNA sequencing capabilities, did show that saccharin could alter the rat microbiome, it is only relatively recently that we understand the fuller effects of NAS on both the microbiota as well as clinical outcomes. Many of the clearest studies were done on animals, since it is easier to control the environment completely and isolate the effects attributable to NAS. A recent study looked further into the relationship between NAS, the microbiome, and the clinical effects (see Suez J. Gut Microbes 2015; 6(2), 149). This is an update of a prior article in Nature (see prior blog:​ )


  • The human weight control studies here are a bit mixed. However it should be noted that most of the comparisons were between individuals consuming NAS to those consuming caloric sweeteners, with very few comparing NAS consumption to avoiding all sweeteners.
  • Several studies have shown NAS leads to weight gain in rats (including saccharin, sucralose, aspartame and Stevia), and are associated with increased adiposity
  • NAS can also induce hyperinsulinemia, impaired insulin tolerance, impaired glucose homeostasis, and worsened atherosclerosis in genetically susceptible mice
  • It should be noted that there are some genetically-altered mice where there are some discordant defects: some with decreasing glucose and insulin levels but increasing adiposity, and in some cases hyperinsulinemia

Details of the current study:

  • Mice drinking water supplemented with high doses of commercial saccharin, sucralose, or aspartame, after 11 weeks had marked glucose intolerance, as compared to controls drinking water, sucrose, or glucose.
  • Further studies of saccharin showed that mice on different baseline diets (e.g. high-fat or other) and at different doses of saccharin had increased glucose intolerance
  • The glucose intolerance induced by NAS was ameliorated by prior dosing with antibiotics (ciprofloxacin and metronidazole, in an attempt to sterilize the gut)
  • There were specific changes in the microbiome associated with NAS, including enrichment of Bacteroides and some Clostridiales and decreases in Lactobacilli and some other members of Clostridiales, several of the microbiota changes previously associated with type II diabetes in humans
  • Fecal microbiomes from mice consuming either water or commercial saccharin were then transplanted into germ-free mice, finding that those germ-free mice receiving the saccharin-associated microbiome developed glucose intolerance
  • In 381 nondiabetic humans, NAS consumption was associated with increases in BMI, blood pressure, hemoglobin A1c, and fasting glucose levels. Also there were changes in microbial taxa in the microbiome: more Actinobacteria, Enterobacteriales, and certain Clostridiales.
  • A preliminary small-scale human study found that supplementing the regular diet with higher doses of saccharin led to elevated glycemic responses in four of the seven volunteers, those 4 had microbiome alterations. And when these microbiomes were transplanted into germ-free mice, these mice also developed the same abnormal glycemic responses. Of note, in two of these 4 volunteers, their microbiome changes reverted to normal within 2 to 8 weeks.


  • NAS is consumed by approximately 32% of adult Americans.
  • The microbiome can be rapidly altered by diet, as noted in diets rich in fat (for example, see
  • There are a remarkable number of largely unregulated food additives in the current food supply, many added for purely commercial ends, such as preservatives to extend the shelf life of some foods. I believe this NAS data challenges the concept that even those ingredients that are not absorbed and internalized could conceivably adversely affect the human microbiome. The main point here is not that all additives or chemicals are necessarily bad, but that we should be very circumspect about assuming that they are probably benign based on our often incomplete models (i.e. It did make intuitive sense at the time that a non-absorbed sweetener would lead to less obesity and diabetes; but as our understanding and models have expanded/become more complex, our “intuitive” sense has changed). But, I think all of this reinforces what Michael Pollan (author or many books, including The Omnivore’s Dilemma) has suggested: it really does make sense to eat natural foods, especially ones which our bodies have evolutionarily adapted to, and avoid foods with ingredients that your grandmother would not know.

In my practice, I have focused on trying to get patients to decrease their consumption of sodas, and with some reasonable success. I think this is often the low-hanging fruit (though less healthy than other fruits), and at least most of my patients say they have dramatically decreased or eliminated sodas by either substituting water (best) or water slightly flavored by fruit juice. For regular sodas, the attempt is to decrease the consumption of high-fructose corn syrup (a bad actor with multiple bad effects, including increasing uric acid levels), was well as “diet” sodas (commenting on the fact that they really are not benign, non-sugar alternatives, as above). I think my patients have been able to change this soda habit by our regularly and repeatedly targeting this issue (with motivational interviewing) over the past several years, especially with my patients who are overweight, glucose intolerant/diabetic or hyperuricemic.

EBM blog homepage

Evidence-Based Medicine blog

Analysis and discussion of developments in Evidence-Based Medicine Visit site

Creative Comms logo

Latest from Evidence-Based Medicine

Latest from EBM