Diabetes drugs compared

The Daily Mirror has reported that drugs used to control blood sugar in type 2 diabetes could increase the risk of heart failure and even death. It said that when taken alone, one group of drugs, called the sulphonylureas, increased risk of death by 61% and of heart failure by 30% compared to another drug called metformin.

This study looked at data on 92,000 patients with type 2 diabetes, comparing the outcomes in those prescribed a range of different oral medications. It found there were higher rates of death in people taking sulphonylurea alone than in those combining it with metformin, but it is possible that this may be partly explained by differences between these groups that affect their risk of death. For example, people taking sulphonylureas were older than those taking metformin. While the study took many of these types of factors into account, this may still not have fully removed their influence.

Importantly, this research compared outcomes of different drugs, but did not compare them to taking no treatment, which is likely to be more dangerous. People should not stop taking diabetes medication as a result of this research, as uncontrolled blood sugar can have serious consequences. People with concerns about these medications can get further advice from their GP or diabetes care team.

Where did the story come from?

This research was conducted by Ioanna Tzoulaki and colleagues from Imperial College London and other research centres in the UK. No specific funding was obtained for this study, as it used anonymous data routinely collected by the general practice research database. The study was published in the peer-reviewed British Medical Journal .

The Daily Telegraph and Daily Mirror have both reported on this study. The Telegraph 's piece is more comprehensive and points out that current treatment guidelines already suggest that metformin is used in preference to sulphonylureas. Both the newspapers include a caution that diabetics should not stop taking their drugs based on this study.

What kind of research was this?

This was a retrospective cohort study looking at the effect different oral medications for type 2 diabetes have on the risk of heart attack, heart failure and death. The study used a large amount of data collected by general practices and stored in the general practice research database (GPRD).

During their development and licensing period new drugs are subject to randomised controlled trials looking at their safety and effectiveness. However, these trials have constraints to their size and length of follow-up, meaning drugs continue to be monitored once they come into general use. These monitoring studies have greater potential to pick up very rare harms or harms that only become evident after a long period of exposure.

This study was people taking drugs medically prescribed to them on the basis of their doctor’s judgement about what is most appropriate, rather than randomly assigned to them by researchers: as such, groups taking the different drugs may not be balanced for other factors that could affect the outcome. For example, the group of people taking one drug might be older than another group, and therefore might be more likely to die. In this study the researchers took steps to minimise the influence of these differences, but they note that the steps they took may not have been sufficient to eliminate potential differences entirely.

Another limitation is that the GPRD was not set up specifically to collect data for this study, meaning some relevant data may be missing or may not have been collected.

What did the research involve?

The researchers looked at anonymous medical records of people who received care for diabetes between 1990 and 2005. These data were on a total of 91,521 people aged between 35 and 90 years old. The researchers looked at which oral diabetes drugs these patients were taking. There are many different oral drugs available for treating type 2 diabetes, including:

  • First generation sulphonylureas
  • Second generation sulphonylureas (newer members of this drug class)
  • Thiazolidinediones (rosiglitazone and pioglitazone)
  • Metformin

People who were not taking any diabetes drugs were not included in the study, and any time periods where people were taking insulin were excluded from the analysis.

The researchers identified which of these individuals had experienced a heart attack or heart failure during the period studied, plus those who died from any cause.

They then compared the risk of these events occurring in people taking different diabetes drugs. In all cases they started out by comparing each drug or drug group against metformin, as it is the first oral anti-diabetic drug that should be considered in people with type 2 diabetes according to guidelines from the International Diabetes Federation. They also compared the effects of the different types of thiazolidinedione drugs (rosiglitazone and pioglitazone). Only a small number of people were taking pioglitazone alone, so they were pooled into a group with those people taking pioglitazone alongside another drug.

The researchers took into account a large number of factors that could affect results, including age at diagnosis, gender, how long a person had diabetes, previous complications from diabetes, previous heart disease, other medication being taken, body mass index, cholesterol concentration, blood pressure, smoking and certain blood test results including one that shows how well blood sugar has been controlled (called HbA1c).

The analyses took these factors into account in three stages, with each stage adjusting for an additional set of factors. Where data on these factors were missing for any person, the person was not included in the analyses. In the fully adjusted analyses 28,812 people had missing data for at least one factor and were excluded on this basis.

What were the basic results?

The study had extensive results: those reported below mainly concentrate on the fully adjusted analyses.

The average age of the 91,521 people assessed was 65 years, and they were followed up for an average of 7.1 years. Metformin was the most commonly prescribed drug (74.5% of people), followed by second generation sulphonylureas taken alone (63.5% of people). During the study period 3,588 people had a first heart attack, 6,900 had first heart failure and 18,548 died.

In their analyses the researchers took into account a large number of factors that could be affecting their results, but they noted that there may have been further factors, other than the diabetes drug used, that were having an effect. In particular, they note the possibility that the different drugs could have been prescribed for people with differing characteristics, which would lead to confounding.


People taking a single sulphonylurea drug alone were significantly more likely to die during the study period than people taking metformin. After adjusting for all factors that could affect results, they found that risk of death was increased 37% in people taking first generation sulphonylureas and by 24% in people taking second generation sulphonylureas. The fully adjusted analyses showed that people taking these newer, second generation sulphonylureas were also at 18% greater risk of heart failure than people taking metformin.


People taking thiazolidinediones (pioglitazone or rosiglitazone) were not at increased risk of heart attack compared with those taking metformin. Those taking pioglitazone alone or in combination with other drugs were at a 39% lower risk of death in the study period than people taking metformin in the fully adjusted analyses.

In the fully adjusted analyses, people taking rosiglitazone alone or in combination with other drugs did not have a significantly different risk of death to those taking metformin. People taking rosiglitazone were at an increased risk of death compared with those taking pioglitazone, but this increase in risk was not statistically significant in the fully adjusted analyses.

How did the researchers interpret the results?

The researchers conclude that:

  • Their findings “suggest a relatively unfavourable risk profile of sulphonylureas compared with metformin for all outcomes examined”
  • Pioglitazone was associated with reduced risk of death in the study period compared with metformin
  • Pioglitazone “had a favourable risk profile compared with rosiglitazone”
  • Their results on pioglitazone need to be confirmed in other studies, but they may have implications for doctors choosing which type of thiazolidinedione to prescribe


People whose type 2 diabetes cannot be controlled by diet and exercise need to be prescribed medications to regulate their blood sugar. If blood sugar levels are not controlled this can lead to a number of different complications, including eye and kidney problems, problems with the nerves in the extremities, and heart disease.

This thorough study compared the risk levels associated with different medications for type 2 diabetes, and did not compare the different drugs versus no treatment. For this reason people should not stop taking their medication based on this study, as taking no medication is likely to be more dangerous. If people taking sulphonylureas are concerned, they should discuss this with their doctors or diabetes care team, who will be able to advise them.

It should also be noted that the people taking the drugs were not randomly assigned to which drugs they received and therefore people taking the different drugs may not have been balanced for other factors that could affect the outcome. For example, the people taking sulphonylureas were on average in their 70s, while those taking the other drugs were in their 60s on average. While the study did take many factors, including age, into account, the authors acknowledge that this may still not have fully removed the effect.

Further points to note include:

  • Another limitation is that as the GPRD was not set up specifically to collect data for this study, some relevant data may be missing or may not have been collected
  • The authors note that their fully adjusted analyses excluded more people than their other analyses because of missing data about the potential confounders. Therefore, these analyses would have had less power to detect differences than the analysis models that adjusted for fewer factors
  • Determining whether a person was exposed to a particular drug was based on what they were prescribed. The study could not determine whether people were actually taking the medication

Current NICE guidelines on treating type 2 diabetes suggest that metformin is the first choice for people who need drug treatment. The sulphonylureas are suggested as possible alternatives in people who are not overweight, who do not tolerate metformin or who have specific reasons why they cannot take metformin, or who need to have their blood sugar lowered quickly because they have particularly high glucose levels.

NHS Attribution