Aaron B. Caughey, MD, PhD

Although some still believe that GDM is a diagnosis looking for a disease, we have excellent evidence that the treatment of even mild forms of insulin resistance leads to improved maternal and neonatal outcomes.^1,2 (See, for example, the article on GDM by E. Albert Reece, MD, PhD, MBA.) Treatment typically involves dietary modification, exercise, and, if necessary, injectable insulin. The most recent ACOG Practice Bulletin on GDM recommends that further studies be conducted before widespread use of oral hypoglycemic agents is initiated.³

The meta-analysis by Dhulkotia and colleagues compares “any” oral hypoglycemic agent (glyburide or metformin) with insulin and concludes that there is no difference between the two types of treatment. However, this conclusion can’t be drawn from the existing literature, and it is certainly nonsensical to conduct a meta-analysis in which both types of oral agents are combined. Studies that lump different medications with distinct mechanisms of action into the same category are unlikely to produce a coherent conclusion. In this meta-analysis, clinical heterogeneity likely exists because glyburide and metformin have entirely different modes of action.

When viewed in isolation, glyburide looks less promising

Let’s consider the literature on glyburide and metformin separately. All three studies comparing glyburide and insulin demonstrated a higher incidence of neonatal hypoglycemia among women taking glyburide, with one of those studies demonstrating a statistically significant effect. In contrast, the single large randomized, controlled study comparing metformin and insulin showed a non-significant reduction in the rate of neonatal hypoglycemia for women taking metformin. The combined effect of the oral agents in the meta-analysis is a nonsignificant increase in neonatal hypoglycemia, but the odds ratio is 1.59, which certainly has some clinical meaning. A similar clinical heterogeneity of the effect on birth weight is seen when the different oral agents are compared with insulin.

Overall, data on glyburide are confusing. Most studies find glyburide and insulin to be essentially equivalent for glycemic control, but insulin leads to better outcomes (although this effect is not significant, probably owing to insufficient power).⁴ The bias among many researchers exploring the use of glyburide for GDM is obvious. One study observed that glyburide was associated with a greater likelihood of neonatal hypoglycemia and higher birth weight than insulin, but still concluded that glyburide is a reasonable first-line therapy for GDM.⁵

Data on metformin are more promising.⁶ In the large prospective trial of metformin, the great majority of outcomes were essentially the same for both metformin and insulin. However, two findings give pause:

• Approximately half the patients randomized to metformin eventually needed insulin to achieve adequate glucose control
• There was a higher incidence of preterm birth (<37 weeks) in the metformin arm. Although neonatal complications did not occur at a statistically higher rate in the metformin group, the risk of preterm birth should certainly be mentioned to the patient before she is started on the medication.

Ease of use shouldn’t trump safety concerns

Over the past decade, it has become increasingly common for health-care providers to prescribe an oral agent instead of insulin for glycemic control—even though insulin remains the standard of care. I note a prevailing perception that oral agents are easier to use than insulin, thereby improving compliance. Oral agents are easier to prescribe and do not require extensive teaching, which does save time, including nonreimbursable office time. As a result, it makes economic sense to prescribe a pill.

In my experience, however, most pregnant women are adherent with insulin therapy, and insulin injection can be reliably taught and appropriately performed regardless of the patient’s socioeconomic status, education, or language. In terms of efficacy and safety, then, insulin remains our best option.

INSTANT POLL: How would you treat this woman with gestational diabetes?
To read the entire question, enter your response, and see how others have answered, click on the blue title.

Although some still believe that GDM is a diagnosis looking for a disease, we have excellent evidence that the treatment of even mild forms of insulin resistance leads to improved maternal and neonatal outcomes.^1,2 (See, for example, the article on GDM by E. Albert Reece, MD, PhD, MBA.) Treatment typically involves dietary modification, exercise, and, if necessary, injectable insulin. The most recent ACOG Practice Bulletin on GDM recommends that further studies be conducted before widespread use of oral hypoglycemic agents is initiated.³

The meta-analysis by Dhulkotia and colleagues compares “any” oral hypoglycemic agent (glyburide or metformin) with insulin and concludes that there is no difference between the two types of treatment. However, this conclusion can’t be drawn from the existing literature, and it is certainly nonsensical to conduct a meta-analysis in which both types of oral agents are combined. Studies that lump different medications with distinct mechanisms of action into the same category are unlikely to produce a coherent conclusion. In this meta-analysis, clinical heterogeneity likely exists because glyburide and metformin have entirely different modes of action.

When viewed in isolation, glyburide looks less promising

Let’s consider the literature on glyburide and metformin separately. All three studies comparing glyburide and insulin demonstrated a higher incidence of neonatal hypoglycemia among women taking glyburide, with one of those studies demonstrating a statistically significant effect. In contrast, the single large randomized, controlled study comparing metformin and insulin showed a non-significant reduction in the rate of neonatal hypoglycemia for women taking metformin. The combined effect of the oral agents in the meta-analysis is a nonsignificant increase in neonatal hypoglycemia, but the odds ratio is 1.59, which certainly has some clinical meaning. A similar clinical heterogeneity of the effect on birth weight is seen when the different oral agents are compared with insulin.

Overall, data on glyburide are confusing. Most studies find glyburide and insulin to be essentially equivalent for glycemic control, but insulin leads to better outcomes (although this effect is not significant, probably owing to insufficient power).⁴ The bias among many researchers exploring the use of glyburide for GDM is obvious. One study observed that glyburide was associated with a greater likelihood of neonatal hypoglycemia and higher birth weight than insulin, but still concluded that glyburide is a reasonable first-line therapy for GDM.⁵

Data on metformin are more promising.⁶ In the large prospective trial of metformin, the great majority of outcomes were essentially the same for both metformin and insulin. However, two findings give pause:

• Approximately half the patients randomized to metformin eventually needed insulin to achieve adequate glucose control
• There was a higher incidence of preterm birth (<37 weeks) in the metformin arm. Although neonatal complications did not occur at a statistically higher rate in the metformin group, the risk of preterm birth should certainly be mentioned to the patient before she is started on the medication.

Ease of use shouldn’t trump safety concerns

Over the past decade, it has become increasingly common for health-care providers to prescribe an oral agent instead of insulin for glycemic control—even though insulin remains the standard of care. I note a prevailing perception that oral agents are easier to use than insulin, thereby improving compliance. Oral agents are easier to prescribe and do not require extensive teaching, which does save time, including nonreimbursable office time. As a result, it makes economic sense to prescribe a pill.

In my experience, however, most pregnant women are adherent with insulin therapy, and insulin injection can be reliably taught and appropriately performed regardless of the patient’s socioeconomic status, education, or language. In terms of efficacy and safety, then, insulin remains our best option.

INSTANT POLL: How would you treat this woman with gestational diabetes?
To read the entire question, enter your response, and see how others have answered, click on the blue title.

Although some still believe that GDM is a diagnosis looking for a disease, we have excellent evidence that the treatment of even mild forms of insulin resistance leads to improved maternal and neonatal outcomes.^1,2 (See, for example, the article on GDM by E. Albert Reece, MD, PhD, MBA.) Treatment typically involves dietary modification, exercise, and, if necessary, injectable insulin. The most recent ACOG Practice Bulletin on GDM recommends that further studies be conducted before widespread use of oral hypoglycemic agents is initiated.³

The meta-analysis by Dhulkotia and colleagues compares “any” oral hypoglycemic agent (glyburide or metformin) with insulin and concludes that there is no difference between the two types of treatment. However, this conclusion can’t be drawn from the existing literature, and it is certainly nonsensical to conduct a meta-analysis in which both types of oral agents are combined. Studies that lump different medications with distinct mechanisms of action into the same category are unlikely to produce a coherent conclusion. In this meta-analysis, clinical heterogeneity likely exists because glyburide and metformin have entirely different modes of action.

When viewed in isolation, glyburide looks less promising

Let’s consider the literature on glyburide and metformin separately. All three studies comparing glyburide and insulin demonstrated a higher incidence of neonatal hypoglycemia among women taking glyburide, with one of those studies demonstrating a statistically significant effect. In contrast, the single large randomized, controlled study comparing metformin and insulin showed a non-significant reduction in the rate of neonatal hypoglycemia for women taking metformin. The combined effect of the oral agents in the meta-analysis is a nonsignificant increase in neonatal hypoglycemia, but the odds ratio is 1.59, which certainly has some clinical meaning. A similar clinical heterogeneity of the effect on birth weight is seen when the different oral agents are compared with insulin.

Overall, data on glyburide are confusing. Most studies find glyburide and insulin to be essentially equivalent for glycemic control, but insulin leads to better outcomes (although this effect is not significant, probably owing to insufficient power).⁴ The bias among many researchers exploring the use of glyburide for GDM is obvious. One study observed that glyburide was associated with a greater likelihood of neonatal hypoglycemia and higher birth weight than insulin, but still concluded that glyburide is a reasonable first-line therapy for GDM.⁵

Data on metformin are more promising.⁶ In the large prospective trial of metformin, the great majority of outcomes were essentially the same for both metformin and insulin. However, two findings give pause:

• Approximately half the patients randomized to metformin eventually needed insulin to achieve adequate glucose control
• There was a higher incidence of preterm birth (<37 weeks) in the metformin arm. Although neonatal complications did not occur at a statistically higher rate in the metformin group, the risk of preterm birth should certainly be mentioned to the patient before she is started on the medication.

Ease of use shouldn’t trump safety concerns

Over the past decade, it has become increasingly common for health-care providers to prescribe an oral agent instead of insulin for glycemic control—even though insulin remains the standard of care. I note a prevailing perception that oral agents are easier to use than insulin, thereby improving compliance. Oral agents are easier to prescribe and do not require extensive teaching, which does save time, including nonreimbursable office time. As a result, it makes economic sense to prescribe a pill.

In my experience, however, most pregnant women are adherent with insulin therapy, and insulin injection can be reliably taught and appropriately performed regardless of the patient’s socioeconomic status, education, or language. In terms of efficacy and safety, then, insulin remains our best option.

INSTANT POLL: How would you treat this woman with gestational diabetes?
To read the entire question, enter your response, and see how others have answered, click on the blue title.

The rate of cesarean delivery in the United States reached its highest level yet— 31.8%—in 2007.¹ At more than 1.2 million procedures each year, cesarean delivery is the most common major surgery performed, but few studies have explored the techniques involved. Recent investigations have focused on whether it is advisable to close the peritoneum at the time of cesarean section,² how to perform and close the hysterotomy,³ and the timing of prophylactic antibiotics.^4,5 But we need more evidence to optimize outcomes.

Rousseau and colleagues have made a commendable effort to sift through one of the many unanswered questions regarding cesarean technique: What is the best skin closure? Their prospective, randomized, controlled trial compared staple closure with subcuticular closure using 4-0 Monocryl.

The authors did many things we have come to expect from clinical trials, including:

• a priori sample-size calculation to ensure adequate statistical power
• randomization by group of 8
• stratification of randomization by primary and repeat cesarean delivery
• assessment of wound cosmesis in a blinded, masked fashion.

They found a “statistically significant” difference in the pain score at 6 postoperative weeks between staples and sutures, with staples having the lower mean score (0.2 vs 0.5; P=.04). They also demonstrated shorter operative time for staple closure (32 vs 41 minutes; P<.001).

When such a study is published, it is easy to assume that the issue has been settled and to change or not change practice, depending on your existing technique—but that is often unwise. Every study has limitations. Even when statistically significant benefits are demonstrated, as they are in this study, it may not always be clear whether your patients match the patients in the study, or whether your technique matches what has been administered during the investigation.

In this case, a few problems need to be pointed out:

• Although the evaluation of cosmesis was by masked clinicians, assessment of the primary outcome—pain—was conducted by the patients themselves, who were not masked. One can easily see that awareness of a suture retained beneath the skin might bias a patient’s perception of pain and discomfort. It would be relatively easy to mask the type of closure—even from patients—on postoperative day 1, but masking would become much more difficult when the staples needed to be removed.
• The issue of statistical analysis can sometimes be dull, but is occasionally paramount in determining validity of a study. In this case, the primary outcome—the pain scale—was considered a continuous outcome and compared using a Student’s t-test. An important assumption in this test is that the data are normally distributed. However, the authors do not make it clear whether they tested the data for normalcy. Particularly at the 6-week evaluation, when the mean value was between 0 and 1 for both groups, it seems unlikely that the data were normally distributed. As a result, the difference in pain scores—0.2 vs 0.5—could have been driven by a few high values in one group. Statistically, this would have been easy to manage by changing the comparison to a Wilcoxon rank-sum test.

Despite these limitations, it does seem unlikely that the pain at 6 weeks would have been worse in the staple group.

These findings contrast those of another study of the same topic, which found less pain in the subcuticular suture group.⁶ In that unmasked study, subcuticular closure was determined to be more “cosmetically attractive” by the patients and their physicians. Again, one needs to be concerned about bias.

The rate of cesarean delivery in the United States reached its highest level yet— 31.8%—in 2007.¹ At more than 1.2 million procedures each year, cesarean delivery is the most common major surgery performed, but few studies have explored the techniques involved. Recent investigations have focused on whether it is advisable to close the peritoneum at the time of cesarean section,² how to perform and close the hysterotomy,³ and the timing of prophylactic antibiotics.^4,5 But we need more evidence to optimize outcomes.

Rousseau and colleagues have made a commendable effort to sift through one of the many unanswered questions regarding cesarean technique: What is the best skin closure? Their prospective, randomized, controlled trial compared staple closure with subcuticular closure using 4-0 Monocryl.

The authors did many things we have come to expect from clinical trials, including:

• a priori sample-size calculation to ensure adequate statistical power
• randomization by group of 8
• stratification of randomization by primary and repeat cesarean delivery
• assessment of wound cosmesis in a blinded, masked fashion.

They found a “statistically significant” difference in the pain score at 6 postoperative weeks between staples and sutures, with staples having the lower mean score (0.2 vs 0.5; P=.04). They also demonstrated shorter operative time for staple closure (32 vs 41 minutes; P<.001).

When such a study is published, it is easy to assume that the issue has been settled and to change or not change practice, depending on your existing technique—but that is often unwise. Every study has limitations. Even when statistically significant benefits are demonstrated, as they are in this study, it may not always be clear whether your patients match the patients in the study, or whether your technique matches what has been administered during the investigation.

In this case, a few problems need to be pointed out:

• Although the evaluation of cosmesis was by masked clinicians, assessment of the primary outcome—pain—was conducted by the patients themselves, who were not masked. One can easily see that awareness of a suture retained beneath the skin might bias a patient’s perception of pain and discomfort. It would be relatively easy to mask the type of closure—even from patients—on postoperative day 1, but masking would become much more difficult when the staples needed to be removed.
• The issue of statistical analysis can sometimes be dull, but is occasionally paramount in determining validity of a study. In this case, the primary outcome—the pain scale—was considered a continuous outcome and compared using a Student’s t-test. An important assumption in this test is that the data are normally distributed. However, the authors do not make it clear whether they tested the data for normalcy. Particularly at the 6-week evaluation, when the mean value was between 0 and 1 for both groups, it seems unlikely that the data were normally distributed. As a result, the difference in pain scores—0.2 vs 0.5—could have been driven by a few high values in one group. Statistically, this would have been easy to manage by changing the comparison to a Wilcoxon rank-sum test.

Despite these limitations, it does seem unlikely that the pain at 6 weeks would have been worse in the staple group.

These findings contrast those of another study of the same topic, which found less pain in the subcuticular suture group.⁶ In that unmasked study, subcuticular closure was determined to be more “cosmetically attractive” by the patients and their physicians. Again, one needs to be concerned about bias.

The rate of cesarean delivery in the United States reached its highest level yet— 31.8%—in 2007.¹ At more than 1.2 million procedures each year, cesarean delivery is the most common major surgery performed, but few studies have explored the techniques involved. Recent investigations have focused on whether it is advisable to close the peritoneum at the time of cesarean section,² how to perform and close the hysterotomy,³ and the timing of prophylactic antibiotics.^4,5 But we need more evidence to optimize outcomes.

Rousseau and colleagues have made a commendable effort to sift through one of the many unanswered questions regarding cesarean technique: What is the best skin closure? Their prospective, randomized, controlled trial compared staple closure with subcuticular closure using 4-0 Monocryl.

The authors did many things we have come to expect from clinical trials, including:

• a priori sample-size calculation to ensure adequate statistical power
• randomization by group of 8
• stratification of randomization by primary and repeat cesarean delivery
• assessment of wound cosmesis in a blinded, masked fashion.

They found a “statistically significant” difference in the pain score at 6 postoperative weeks between staples and sutures, with staples having the lower mean score (0.2 vs 0.5; P=.04). They also demonstrated shorter operative time for staple closure (32 vs 41 minutes; P<.001).

When such a study is published, it is easy to assume that the issue has been settled and to change or not change practice, depending on your existing technique—but that is often unwise. Every study has limitations. Even when statistically significant benefits are demonstrated, as they are in this study, it may not always be clear whether your patients match the patients in the study, or whether your technique matches what has been administered during the investigation.

In this case, a few problems need to be pointed out:

• Although the evaluation of cosmesis was by masked clinicians, assessment of the primary outcome—pain—was conducted by the patients themselves, who were not masked. One can easily see that awareness of a suture retained beneath the skin might bias a patient’s perception of pain and discomfort. It would be relatively easy to mask the type of closure—even from patients—on postoperative day 1, but masking would become much more difficult when the staples needed to be removed.
• The issue of statistical analysis can sometimes be dull, but is occasionally paramount in determining validity of a study. In this case, the primary outcome—the pain scale—was considered a continuous outcome and compared using a Student’s t-test. An important assumption in this test is that the data are normally distributed. However, the authors do not make it clear whether they tested the data for normalcy. Particularly at the 6-week evaluation, when the mean value was between 0 and 1 for both groups, it seems unlikely that the data were normally distributed. As a result, the difference in pain scores—0.2 vs 0.5—could have been driven by a few high values in one group. Statistically, this would have been easy to manage by changing the comparison to a Wilcoxon rank-sum test.

Despite these limitations, it does seem unlikely that the pain at 6 weeks would have been worse in the staple group.

These findings contrast those of another study of the same topic, which found less pain in the subcuticular suture group.⁶ In that unmasked study, subcuticular closure was determined to be more “cosmetically attractive” by the patients and their physicians. Again, one needs to be concerned about bias.