Intended for healthcare professionals


Helping patients make sense of the risks of taking GLP-1 agonists

BMJ 2013; 346 doi: (Published 10 June 2013) Cite this as: BMJ 2013;346:f3692
  1. Victor M Montori, professor of medicine
  1. 1Division of Endocrinology, Diabetes, Metabolism, and Nutrition, Mayo Clinic, Rochester, MN 55902, USA
  1. montori.victor{at}

Tough when benefits and harms are unclear

There are three reasons to lower glycaemia in patients with type 2 diabetes: to treat the symptoms of hyperglycaemia; to prevent symptomatic hyperglycaemia; and to reduce the risk of developing complications associated with diabetes. Regulatory agencies approve antihyperglycaemic agents because they prevent and treat hyperglycaemia. The US Food and Drug Administration now requires drug companies to show that their antihyperglycaemic drugs do not increase a patient’s risk of developing cardiovascular disease. Currently, however, no regulatory agency requires evidence of a drug’s efficacy in reducing the risk of developing complications of diabetes.

Does the use of antidiabetic drugs have any benefits in patients with asymptomatic type 2 diabetes? It is possible that lowering glycaemia, regardless of the approach, reduces the risk of microvascular and cardiovascular complications, but this remains uncertain, despite testing in contemporary trials that have enrolled tens of thousands of patients.1 Furthermore, comparative effectiveness studies have shown no antihyperglycaemic drug to be more beneficial in this regard than any other.2 Evidence from long term follow-up of patients in the UK Prospective Diabetes Study suggests that cardiovascular complications might be reduced by preventing hyperglycaemic symptoms.3 Considerable uncertainty exists, however, in applying this evidence to patients who are at risk of cardiovascular complications but who quit smoking and adhere to low dose aspirin, standard or high dose statins, and antihypertensive treatment. The residual modifiable risk might be too small to justify expensive, inconvenient, and risky antihyperglycaemic drugs, let alone inform the selection of a particular agent. Therefore, the clearest indication for the use of antihyperglycaemic drugs in asymptomatic patients with type 2 diabetes remains the prevention of symptomatic hyperglycaemia.

If we are satisfied that this is justification enough for treatment how might we choose antihyperglycaemic agents? In the absence of any clear indication that one agent is better than another, the choice will depend on their unfavourable features, which include costs of treatment, particularly out-of-pocket costs, inconvenient administration, and adverse effects. Although some serious adverse events are rare, they will still play an important role in making a choice, particularly if patients at risk cannot be reliably identified or when the adverse event is severe or lethal.

Pancreatitis and pancreatic cancer are associated with high morbidity and mortality; their risk is increased with obesity and diabetes,4 5 and with conditions associated with these disorders, such as hypertriglyceridaemia or gallbladder disease.

Metformin has been estimated to reduce the risk of pancreatic cancer by 24% (although not statistically significant); sulfonylureas and insulin have been associated with a 70% (significant) and 59% (not significant) increased risk, respectively.6 But these estimates arise from observational studies with important limitations, such as potential for reverse confounding. For example, glycaemic deterioration could be a sign of pancreatic cancer in a patient with diabetes, and the resulting intensification of treatment might produce a spurious association between pancreatic cancer and second line agents and insulin.7 Other factors such as incomplete adjustment for risk factors and the potential association of a risk factor for cancer (such as age) with a higher chance of being prescribed an antidiabetic drug further impair interpretation of these observational analyses. Such problems, which reduce the confidence we have in the estimates of association, also apply to the evolving data that link pancreatitis and pancreatic cancer with glucagon-like peptide-1 (GLP-1) based treatment.

As noted by Cohen and others, there is a compelling biological rationale linking GLP-1 based treatments with pancreatitis and pancreatic cancer.8 9 10 11 12 Faced with emerging data, regulators have added warnings to product information leaflets.

Meta-analyses of observational studies and randomised trials have found no significant association between GLP-1 based treatments and pancreatitis or pancreatic cancer. However, many factors bring the reliability of the evidence into question. Observational studies have lacked sufficient numbers of patients using these agents who were followed for a sufficiently long period and accrued enough instances of adequately ascertained pancreatitis or pancreatic cancer. Careful selection of participants, co-interventions, and comparisons have also affected the results of trials of these agents.13 A recently published population based case-control study based on a large US administrative dataset found a twofold increase in the odds of hospital admission for pancreatitis in patients exposed to GLP-1 based treatments, after accounting for important confounders.14 Taken together, it seems plausible that GLP-1 based treatment can cause pancreatitis and pancreatic cancer, but the causal link is far from established.

So how can clinicians and patients work together to decide on a drug regimen that reflects the available evidence and the patient’s context and values?

Shared decision making tools (such as the Diabetes Medication Choice Decision Aid; can be used to support this process.15 Patients will have to consider carefully their predisposition to diabetes complications and drug side effects, taking into account their personal and family history of diabetes complications and risk factors for pancreatitis and pancreatic cancer in light of their life expectancy. They will also need to think about how well they are likely to tolerate adverse drug effects (such as hypoglycaemia and drug induced weight changes). After careful reflection, most patients and clinicians may opt to avoid GLP-1 based drugs or to avoid using them early in the disease course, alone, or for a prolonged period of time.

Colouring these discussions is the concern that the evidence is corrupt. The public record about drug safety and the opinion of experts, particularly of those with financial ties to drug manufacturers, have often been found to be untrustworthy.16

Patients benefit from genuine advances in therapeutics, many of which are made possible by big pharma. Yet, big pharma has behaved poorly in the past two decades by hiding safety signals, skewing the evidence and debate in favour of its products, and attacking those who raise concerns. The work of investigators, regulators, and advocates who make it their business to look carefully at the safety of drugs may go unrecognised and is often vilified. Yet, such work helps us to moderate our expectations about innovations and to be thoughtful in prescribing, so that the patient’s best interests are to the fore. Patients, clinicians, industry, and regulators must now work together towards increased transparency and accountability. The effects of antidiabetic drugs should be tested independently of the manufacturers, and all data from all studies from all developmental phases should be made available to the regulatory authorities and to the scientific and safety advocacy community. In turn, government and civil society actors must fulfil their duty to protect the health of those who are most vulnerable.


Cite this as: BMJ 2013;346:f3692



View Abstract