ACCORD Mined for Clues About Cardiovascular Deaths

When a landmark trial was stopped early due to a higher death rate in patients randomized to intensive glycemic control, the key question was the medical equivalent of the murder-mystery line, "Who done it?" What caused the higher death rate with intensive control compared with standard glycemic targets?

The randomized controlled trial, Action to Control Cardiovascular Risk in Diabetes (ACCORD), involving 10,251 patients with diabetes who were at high risk for cardiovascular disease, showed a 22% higher all-cause mortality rate among the 5,128 patients whose treatment arm aimed for a hemoglobin A_1c level of less than 6% compared with patients whose target HbA_1c was in the range of 7%-7.9% (N. Engl. J. Med. 2008;358:2545-59).

Subsequent analyses showed that a rapid drop in glucose wasn’t the cause. Neither was a low HbA_1c level in and of itself. "It’s not a problem to get a lower A_1c if you’re trying to," explained Dr. Richard M. Bergenstal of the ACCORD study team.

Neither could weight gain or the use of thiazolidinediones or other medications explain the higher mortality risk with intensive glycemic control.

Episodes of severe hypoglycemia increased the risk for death in both the intensive- and standard-therapy groups, but the general consensus is that severe hypoglycemia did not explain the difference in death rates between those groups, according to Dr. Bergenstal, executive director of the International Diabetes Center at Park Nicollet Health Services, Minneapolis.

Mild hypoglycemia was also investigated as a possible factor in the higher death rates. To get a profile of the potential risk associated with mild hypoglycemia, the ACCORD researchers analyzed more than 9.4 million data points from self-monitored blood glucose tests recorded over an average 2-year period by 5,347 patients, approximately half of the ACCORD cohort, he said at the annual meeting of the American Diabetes Association in San Diego.

What they found was surprising. It wasn’t a high or low HbA_1c per se that was most associated with the increased mortality risk in the intensive control group, but unintended high or low blood sugars.

"The more you diverge from what you’re trying to achieve in glycemic targets, the higher the risk of mortality," Dr. Bergenstal said.

In Dr. Bergenstal’s opinion, this implies that clinicians should set a goal not only for HbA_1c levels but for blood glucose levels, and they should monitor and evaluate patients’ blood glucose profiles.

"If you’re not achieving the target you set, be careful, and think about whether you want to intensify further if you’re diverging from that," he suggested.

Of the patients with self-monitored blood glucose data, the 2,691 patients in the intensive control group tested their blood sugar levels a mean of 2.7 times per day, compared with 2 times per day for the 2,656 patients with data in the standard control group.

Those who tested more frequently tended to have lower HbA_1c levels. In the intensive control group, those who tested blood glucose once per day had a mean HbA_1c of 6.9% and those who tested five or more times per day had a mean HbA_1c of 6.5%. In the standard control group, those who tested blood glucose once per day had a mean HbA_1c of 7.8% and those who tested five or more times per day had a mean HbA_1c of 7.3%.

When the investigators analyzed the blood glucose monitoring data by 2-hour intervals, a profile emerged of rising blood sugar levels during the day and a significant drop in glucose levels overnight, in both the intensive and standard control groups.

"The more steep both of those trajectories are, the worse they did," Dr. Bergenstal said.

This indicates that medication approaches are needed to help smooth out this type of curve, he said.

A modal day profile of patients with one or more severe hypoglycemic reactions produced the same pattern but with more instability – a sharper curve of blood glucose values going up during the day, and a sharper curve down overnight. Among patients who died, the modal day profile featured even steeper increases in blood glucose values during the day and sharper drops overnight.

"I can’t say we’ve had the database open long enough to go back and say, ‘At 8 p.m. on March 22, what happened to this person?’ But we will, trust me. We will look even closer at these numbers over time," he said. "I think they’re already giving a sense that it’s probably not a good thing to be going up and down with such velocity."

Difficulties in Controlling Blood Glucose Values

Patients in the intensive-control group recorded three times as many low blood glucose levels as did patients in the standard-control group. That might lead one to think that excess hypoglycemia was the cause of excess mortality in the intensive group, but that wasn’t the case. In the intensive control group, patients who died had more high blood glucose readings than did patients who survived, but not more low blood glucose readings.

The point is that patients who died diverged from their targets, Dr. Bergenstal said. Investigators expected to see low glucose levels in the intensive control group. Patients who died had a lot of blood glucose readings of 200 mg/dL or 300 mg/dL.

Patients in the standard-control group recorded twice as many high blood sugar readings as did patients in the intensive-control group. Patients in the standard group who died were more likely to have high or low blood glucose values compared with patients who survived.

"Again, they diverged" from the goal. "We were trying to keep these people about in the middle," he said.

The findings imply that clinicians should relax the intensity of therapy, he added. The HbA_1c and self-monitored blood glucose values diverged significantly from targets, despite reasonably intensive efforts at control.

High HbA_1c levels were associated with both dangerously high glucose values and dangerously low glucose values. High levels were dangerous because of hyperglycemia but also because people who had some of the worst hypoglycemia had high HbA_1c levels, he said.

Severe hypoglycemia increased risk for death in both patient groups.

And mild to moderate hypoglycemia? "We want to minimize it, particularly in those patients for whom we relax our A_1c and self-monitored blood glucose targets, such as the frail elderly," Dr. Bergenstal advised. "Be very careful about even mild to moderate hypoglycemia in those individuals."

Characteristics of patients with self-monitored blood glucose data were similar to those without the data, so this sample was fairly representative of all patients in ACCORD, he said.

The investigators are working to distill the findings into a practical message about how to quantify stability or variability in self-monitored blood glucose levels.

Dr. Bergenstal holds stock in Merck. He has received research support from, or been an adviser or consultant to, many companies including Abbot Diabetes Care, Amylin Pharmaceuticals, Bayer Health Care, Biodel, Calibra Medical, Eli Lilly, Hygieia, Intuity Medical, Lifescan, MannKind Corp., Medtronic, Novo Nordisk, Pfizer, ResMed, Roche Diagnostics, Roche Pharmaceuticals, Sanofi-Aventis, Takeda Pharmaceutical Company, Valeritas, United Health Group/i3 Statprobe, Dexcom, and Intarcia Therapeutics.

The ACCORD trial was sponsored by Abbot Laboratories, Amylin Pharmaceuticals, AstraZeneca Pharmaceuticals, Bayer HealthCare, Closer Healthcare, GlaxoSmithKline Pharmaceuticals, King Pharmaceuticals, Merck, Novartis Pharmaceuticals, Novo Nordisk, Omron Healthcare, Sanofi U.S., Takeda Pharmaceuticals, the National Institute of Diabetes and Digestive and Kidney Diseases, the National Eye Institute, the National Institute on Aging, and the Centers for Disease Control and Prevention.

When a landmark trial was stopped early due to a higher death rate in patients randomized to intensive glycemic control, the key question was the medical equivalent of the murder-mystery line, "Who done it?" What caused the higher death rate with intensive control compared with standard glycemic targets?

The randomized controlled trial, Action to Control Cardiovascular Risk in Diabetes (ACCORD), involving 10,251 patients with diabetes who were at high risk for cardiovascular disease, showed a 22% higher all-cause mortality rate among the 5,128 patients whose treatment arm aimed for a hemoglobin A_1c level of less than 6% compared with patients whose target HbA_1c was in the range of 7%-7.9% (N. Engl. J. Med. 2008;358:2545-59).

Subsequent analyses showed that a rapid drop in glucose wasn’t the cause. Neither was a low HbA_1c level in and of itself. "It’s not a problem to get a lower A_1c if you’re trying to," explained Dr. Richard M. Bergenstal of the ACCORD study team.

Neither could weight gain or the use of thiazolidinediones or other medications explain the higher mortality risk with intensive glycemic control.

Episodes of severe hypoglycemia increased the risk for death in both the intensive- and standard-therapy groups, but the general consensus is that severe hypoglycemia did not explain the difference in death rates between those groups, according to Dr. Bergenstal, executive director of the International Diabetes Center at Park Nicollet Health Services, Minneapolis.

Mild hypoglycemia was also investigated as a possible factor in the higher death rates. To get a profile of the potential risk associated with mild hypoglycemia, the ACCORD researchers analyzed more than 9.4 million data points from self-monitored blood glucose tests recorded over an average 2-year period by 5,347 patients, approximately half of the ACCORD cohort, he said at the annual meeting of the American Diabetes Association in San Diego.

What they found was surprising. It wasn’t a high or low HbA_1c per se that was most associated with the increased mortality risk in the intensive control group, but unintended high or low blood sugars.

"The more you diverge from what you’re trying to achieve in glycemic targets, the higher the risk of mortality," Dr. Bergenstal said.

In Dr. Bergenstal’s opinion, this implies that clinicians should set a goal not only for HbA_1c levels but for blood glucose levels, and they should monitor and evaluate patients’ blood glucose profiles.

"If you’re not achieving the target you set, be careful, and think about whether you want to intensify further if you’re diverging from that," he suggested.

Of the patients with self-monitored blood glucose data, the 2,691 patients in the intensive control group tested their blood sugar levels a mean of 2.7 times per day, compared with 2 times per day for the 2,656 patients with data in the standard control group.

Those who tested more frequently tended to have lower HbA_1c levels. In the intensive control group, those who tested blood glucose once per day had a mean HbA_1c of 6.9% and those who tested five or more times per day had a mean HbA_1c of 6.5%. In the standard control group, those who tested blood glucose once per day had a mean HbA_1c of 7.8% and those who tested five or more times per day had a mean HbA_1c of 7.3%.

When the investigators analyzed the blood glucose monitoring data by 2-hour intervals, a profile emerged of rising blood sugar levels during the day and a significant drop in glucose levels overnight, in both the intensive and standard control groups.

"The more steep both of those trajectories are, the worse they did," Dr. Bergenstal said.

This indicates that medication approaches are needed to help smooth out this type of curve, he said.

A modal day profile of patients with one or more severe hypoglycemic reactions produced the same pattern but with more instability – a sharper curve of blood glucose values going up during the day, and a sharper curve down overnight. Among patients who died, the modal day profile featured even steeper increases in blood glucose values during the day and sharper drops overnight.

"I can’t say we’ve had the database open long enough to go back and say, ‘At 8 p.m. on March 22, what happened to this person?’ But we will, trust me. We will look even closer at these numbers over time," he said. "I think they’re already giving a sense that it’s probably not a good thing to be going up and down with such velocity."

Difficulties in Controlling Blood Glucose Values

Patients in the intensive-control group recorded three times as many low blood glucose levels as did patients in the standard-control group. That might lead one to think that excess hypoglycemia was the cause of excess mortality in the intensive group, but that wasn’t the case. In the intensive control group, patients who died had more high blood glucose readings than did patients who survived, but not more low blood glucose readings.

The point is that patients who died diverged from their targets, Dr. Bergenstal said. Investigators expected to see low glucose levels in the intensive control group. Patients who died had a lot of blood glucose readings of 200 mg/dL or 300 mg/dL.

Patients in the standard-control group recorded twice as many high blood sugar readings as did patients in the intensive-control group. Patients in the standard group who died were more likely to have high or low blood glucose values compared with patients who survived.

"Again, they diverged" from the goal. "We were trying to keep these people about in the middle," he said.

The findings imply that clinicians should relax the intensity of therapy, he added. The HbA_1c and self-monitored blood glucose values diverged significantly from targets, despite reasonably intensive efforts at control.

High HbA_1c levels were associated with both dangerously high glucose values and dangerously low glucose values. High levels were dangerous because of hyperglycemia but also because people who had some of the worst hypoglycemia had high HbA_1c levels, he said.

Severe hypoglycemia increased risk for death in both patient groups.

And mild to moderate hypoglycemia? "We want to minimize it, particularly in those patients for whom we relax our A_1c and self-monitored blood glucose targets, such as the frail elderly," Dr. Bergenstal advised. "Be very careful about even mild to moderate hypoglycemia in those individuals."

Characteristics of patients with self-monitored blood glucose data were similar to those without the data, so this sample was fairly representative of all patients in ACCORD, he said.

The investigators are working to distill the findings into a practical message about how to quantify stability or variability in self-monitored blood glucose levels.

Dr. Bergenstal holds stock in Merck. He has received research support from, or been an adviser or consultant to, many companies including Abbot Diabetes Care, Amylin Pharmaceuticals, Bayer Health Care, Biodel, Calibra Medical, Eli Lilly, Hygieia, Intuity Medical, Lifescan, MannKind Corp., Medtronic, Novo Nordisk, Pfizer, ResMed, Roche Diagnostics, Roche Pharmaceuticals, Sanofi-Aventis, Takeda Pharmaceutical Company, Valeritas, United Health Group/i3 Statprobe, Dexcom, and Intarcia Therapeutics.

The ACCORD trial was sponsored by Abbot Laboratories, Amylin Pharmaceuticals, AstraZeneca Pharmaceuticals, Bayer HealthCare, Closer Healthcare, GlaxoSmithKline Pharmaceuticals, King Pharmaceuticals, Merck, Novartis Pharmaceuticals, Novo Nordisk, Omron Healthcare, Sanofi U.S., Takeda Pharmaceuticals, the National Institute of Diabetes and Digestive and Kidney Diseases, the National Eye Institute, the National Institute on Aging, and the Centers for Disease Control and Prevention.

When a landmark trial was stopped early due to a higher death rate in patients randomized to intensive glycemic control, the key question was the medical equivalent of the murder-mystery line, "Who done it?" What caused the higher death rate with intensive control compared with standard glycemic targets?

The randomized controlled trial, Action to Control Cardiovascular Risk in Diabetes (ACCORD), involving 10,251 patients with diabetes who were at high risk for cardiovascular disease, showed a 22% higher all-cause mortality rate among the 5,128 patients whose treatment arm aimed for a hemoglobin A_1c level of less than 6% compared with patients whose target HbA_1c was in the range of 7%-7.9% (N. Engl. J. Med. 2008;358:2545-59).

Subsequent analyses showed that a rapid drop in glucose wasn’t the cause. Neither was a low HbA_1c level in and of itself. "It’s not a problem to get a lower A_1c if you’re trying to," explained Dr. Richard M. Bergenstal of the ACCORD study team.

Neither could weight gain or the use of thiazolidinediones or other medications explain the higher mortality risk with intensive glycemic control.

Episodes of severe hypoglycemia increased the risk for death in both the intensive- and standard-therapy groups, but the general consensus is that severe hypoglycemia did not explain the difference in death rates between those groups, according to Dr. Bergenstal, executive director of the International Diabetes Center at Park Nicollet Health Services, Minneapolis.

Mild hypoglycemia was also investigated as a possible factor in the higher death rates. To get a profile of the potential risk associated with mild hypoglycemia, the ACCORD researchers analyzed more than 9.4 million data points from self-monitored blood glucose tests recorded over an average 2-year period by 5,347 patients, approximately half of the ACCORD cohort, he said at the annual meeting of the American Diabetes Association in San Diego.

What they found was surprising. It wasn’t a high or low HbA_1c per se that was most associated with the increased mortality risk in the intensive control group, but unintended high or low blood sugars.

"The more you diverge from what you’re trying to achieve in glycemic targets, the higher the risk of mortality," Dr. Bergenstal said.

In Dr. Bergenstal’s opinion, this implies that clinicians should set a goal not only for HbA_1c levels but for blood glucose levels, and they should monitor and evaluate patients’ blood glucose profiles.

"If you’re not achieving the target you set, be careful, and think about whether you want to intensify further if you’re diverging from that," he suggested.

Of the patients with self-monitored blood glucose data, the 2,691 patients in the intensive control group tested their blood sugar levels a mean of 2.7 times per day, compared with 2 times per day for the 2,656 patients with data in the standard control group.

Those who tested more frequently tended to have lower HbA_1c levels. In the intensive control group, those who tested blood glucose once per day had a mean HbA_1c of 6.9% and those who tested five or more times per day had a mean HbA_1c of 6.5%. In the standard control group, those who tested blood glucose once per day had a mean HbA_1c of 7.8% and those who tested five or more times per day had a mean HbA_1c of 7.3%.

When the investigators analyzed the blood glucose monitoring data by 2-hour intervals, a profile emerged of rising blood sugar levels during the day and a significant drop in glucose levels overnight, in both the intensive and standard control groups.

"The more steep both of those trajectories are, the worse they did," Dr. Bergenstal said.

This indicates that medication approaches are needed to help smooth out this type of curve, he said.

A modal day profile of patients with one or more severe hypoglycemic reactions produced the same pattern but with more instability – a sharper curve of blood glucose values going up during the day, and a sharper curve down overnight. Among patients who died, the modal day profile featured even steeper increases in blood glucose values during the day and sharper drops overnight.

"I can’t say we’ve had the database open long enough to go back and say, ‘At 8 p.m. on March 22, what happened to this person?’ But we will, trust me. We will look even closer at these numbers over time," he said. "I think they’re already giving a sense that it’s probably not a good thing to be going up and down with such velocity."

Difficulties in Controlling Blood Glucose Values

Patients in the intensive-control group recorded three times as many low blood glucose levels as did patients in the standard-control group. That might lead one to think that excess hypoglycemia was the cause of excess mortality in the intensive group, but that wasn’t the case. In the intensive control group, patients who died had more high blood glucose readings than did patients who survived, but not more low blood glucose readings.

The point is that patients who died diverged from their targets, Dr. Bergenstal said. Investigators expected to see low glucose levels in the intensive control group. Patients who died had a lot of blood glucose readings of 200 mg/dL or 300 mg/dL.

Patients in the standard-control group recorded twice as many high blood sugar readings as did patients in the intensive-control group. Patients in the standard group who died were more likely to have high or low blood glucose values compared with patients who survived.

"Again, they diverged" from the goal. "We were trying to keep these people about in the middle," he said.

The findings imply that clinicians should relax the intensity of therapy, he added. The HbA_1c and self-monitored blood glucose values diverged significantly from targets, despite reasonably intensive efforts at control.

High HbA_1c levels were associated with both dangerously high glucose values and dangerously low glucose values. High levels were dangerous because of hyperglycemia but also because people who had some of the worst hypoglycemia had high HbA_1c levels, he said.

Severe hypoglycemia increased risk for death in both patient groups.

And mild to moderate hypoglycemia? "We want to minimize it, particularly in those patients for whom we relax our A_1c and self-monitored blood glucose targets, such as the frail elderly," Dr. Bergenstal advised. "Be very careful about even mild to moderate hypoglycemia in those individuals."

Characteristics of patients with self-monitored blood glucose data were similar to those without the data, so this sample was fairly representative of all patients in ACCORD, he said.

The investigators are working to distill the findings into a practical message about how to quantify stability or variability in self-monitored blood glucose levels.

Dr. Bergenstal holds stock in Merck. He has received research support from, or been an adviser or consultant to, many companies including Abbot Diabetes Care, Amylin Pharmaceuticals, Bayer Health Care, Biodel, Calibra Medical, Eli Lilly, Hygieia, Intuity Medical, Lifescan, MannKind Corp., Medtronic, Novo Nordisk, Pfizer, ResMed, Roche Diagnostics, Roche Pharmaceuticals, Sanofi-Aventis, Takeda Pharmaceutical Company, Valeritas, United Health Group/i3 Statprobe, Dexcom, and Intarcia Therapeutics.

The ACCORD trial was sponsored by Abbot Laboratories, Amylin Pharmaceuticals, AstraZeneca Pharmaceuticals, Bayer HealthCare, Closer Healthcare, GlaxoSmithKline Pharmaceuticals, King Pharmaceuticals, Merck, Novartis Pharmaceuticals, Novo Nordisk, Omron Healthcare, Sanofi U.S., Takeda Pharmaceuticals, the National Institute of Diabetes and Digestive and Kidney Diseases, the National Eye Institute, the National Institute on Aging, and the Centers for Disease Control and Prevention.