Comparison of Home Glucose Monitoring with the Oral Glucose Tolerance Test to Detect Gestational Glucose Intolerance Kevin A. Peterson, MD, MPH; Alison M. Peterson, MD; Victor Corbett, MD; Scott Tongen, MD; Martha Guzman, MD; and Roger Mazze, PhD St Paul and St Louis Park, Minnesota

Background. Recent evidence suggests that infant mor­ found between increasing birthweight and increasing bidity is increased among women who have abnormal average fasting SM BG values (P < .0 0 1 ), increasing per­ prenatal glucose screening tests but who do not have centage o f SM BG values above 120 ntg/dL (6.7 gestational diabetes mellitus (GDM). These women fall mmol/L) (P<.01), and increasing average SMBG val­ into a diagnostic gray zone and historically have not ues (P < .0 1 6 ). been treated. Conclusions. Maternal home glucose values at 28 weeks Methods. Forty-eight pregnant women with abnormal correlate with the risk of LGA infant births among oral glucose challenge test results performed self-moni­ women in the diagnostic gray zone. Women with aver­ tored blood glucose (SMBG) testing seven times per age fasting SMBG values >95 mg/dL (5.3 mmol/L) day for 1 week before undergoing a diagnostic 100-g oral glucose tolerance test (OGTT). At delivery, perina­ are at increased risk for giving birth to LGA infants and tal complications and birthweights were recorded. Re­ may be more likely to exhibit complications usually as­ sults of SMBG tests for women with normal OGTTs sociated with GDM. were correlated with infant birthweights. Key words. Screening; diabetes, gestational; gestational Results. Thirteen infants (37%) were found to be large age; infant, large for gestational age; fetal macrosomia. for gestational age (LGA). Significant correlation was (J Fam Pract 1994; 39:558-563)

Gestational diabetes mellitus (G D M ), a term that applies Data Group.3 These criteria are based on an interpreta­ to hyperglycemia first discovered in pregnancy, is present tion o f work performed by O ’Sullivan and Mahan,4 who in 3% to 12% o f pregnancies, making its identification and evaluated women with hyperglycemia in pregnancy for treatment essential to the delivery of optimal obstetric the purpose of identifying those at increased risk for sub­ care.1 GDM is associated with an increased incidence of sequent development o f non-insulin-dependent diabetes fetal macrosomia, neonatal hypoglycemia, hypocalcemia, mellitus (N ID D M ). Subsequently, a positive association polycythemia, and hyperbilirubinemia.2 Diagnostic crite­ was made between an abnormal O G T T and increased ria for GDM using the 100-g oral glucose tolerance test perinatal mortality.5 More recendy, the primary purpose (O G T T ) have been defined by the National Diabetes of diagnosis among patients with GDM has been to re­ duce adverse perinatal outcome6 in conjunction with Submitted, revised, August 15, 1994. evaluation of maternal risk for the subsequent develop­ ment of NIDDM. For that reason, it is appropriate to From the Department o f Family Practice and Comunity Health, University o f Min­ nesota (K.A.P., A.M.P., M.G., R.M.), and United Hospital (V.C., S.T.), St Paul; reexamine the current diagnostic criteria. and the International Diabetes Center (R.M.), St Louis Park, Minnesota. Requests The large-for-gestational-age (LGA) infant and mac- fo r reprints should be addressed to Kevin A. Peterson, MD, Department o f Family Practice and Community Health, University o f Minnesota, Hazel Park Family rosomic infant are among the most common adverse out­ Practice Clinic, 933 White Bear Ave, St Paul, MN 55106. comes of GDM.7 LGA infants rank above the 90th per-

© 1994 Appleton & Lange ISSN 0094-3509

558 The Journal of Family Practice, Vol. 39, No. 6(Dec), 1994 Detecting Gestational Glucose Intolerance Peterson, Peterson, Corbett, et al centile in weight, and macrosomic infants weigh more glucose allows for close assessment o f metabolic control. than 4000 g at birth, regardless o f gestational age. Moth­ The availability of the portable glucose reflectance meter ers of macrosomic infants are at increased risk of pro­ with an internal memory chip has facilitated the collection tracted labor, fetopelvic disproportion, and shoulder dys­ of verifiable self-monitored blood glucose data.27 This tocia.8 They have twice the normal risk of cesarean study prospectively examined the correlation between section, as well as an increased risk of postpartum hemor­ self-monitored blood glucose values and infant morbid­ rhage.9 The infant is at increased risk for brachial plexus ity, measured in terms of occurrence o f LGA infants in palsy, clavicle and humerus fracture, fetal asphyxia, hypo­ women with an abnormal 1-hour GCT and a normal glycemia, hyperbilirubinemia, organomegaly, and low 3-hour OGTT. Apgar scores at the time o f birth.10-11 Current literature also suggests that large infants are at increased risk of late diabetes-related sequelae, including obesity and possible developmental delays.12 Previous attempts to find long­ Methods term measures o f glucose control that could predict ad­ Women were recruited from family medicine and obstet­ verse perinatal outcomes have not been successful. He­ ric practices associated with a large metropolitan hospital moglobin A lc has been demonstrated to be an insensitive center. A consecutive sample o f pregnant women was measure o f abnormal glucose metabolism in pregnancy, identified: those who were between the 24th and 29th identifying only 40% o f women with an abnormal gestational week and had plasma glucose results > 140 OGTT.13-14 Fructosamine and glycosylated serum pro­ mg/dL (7.8 mmol/L) and <200 mg/dL (11.1 mmol/ tein have also been reported as insensitive screening tests L), as determined 1 hour following ingestion o f 50 g o f a in pregnancy for gestational diabetes.15-16 standard liquid glucose solution, regardless o f prandial In G D M , maternal insulin resistance is exaggerated state.1 Following informed consent, subjects were inter­ beyond the levels found in nonpregnant women.17 When viewed, and the following data were obtained: pregravid maternal pancreatic /3-cells fail to compensate, maternal weight, maternal height, family history of diabetes, ob­ hyperglycemia results in increased transplacental diffusion stetrical history, medications, tobacco and drug use, and of glucose to the fetus. This, in turn, stimulates excessive gestational age. Women were excluded from the study it fetal pancreatic /3-cell production of insulin.18~20 Fetal they had a previous history o f GDM or a hemoglobin of hyperinsulinemia fosters abnormal fat deposition. Clinical < 10.0 g/dL (100 g/L) since reflectometer readings be­ studies have shown that the risk o f a large infant is five to low this value were considered unreliable. nine times greater among women with glucose levels Subjects were instructed in blood glucose monitor­ >105 mg/dL (5.8 mmol/L) fasting and 140 mg/dL ing techniques by a nurse specialist in diabetes. A memo­ (7.8 mmol/L) postprandial, as compared with women ry-based reflectance meter capable of storing 440 blood without diabetes and those with treated GDM whose glucose values with the corresponding time and date of glucose level is maintained between 60 and 120 mg/dL the test (Glucometer M, Miles Laboratories Inc, Ames (3.3 and 6.7 mm ol/L).21-22 Early intervention and strict Division, Elkhart, Ind) was provided. Subjects were asked glucose control in women with gestational diabetes has to immediately begin measuring their capillary blood glu­ been demonstrated in several studies to decrease maternal cose seven times per day (fasting, preprandial, 2 hours and fetal complications and reduce the incidence of LGA after the beginning o f each meal, and at bedtime) for 7 days. They remained on an unrestricted diet. At the end of infants.8-21 Recent work has demonstrated that some women 7 days, capillary blood glucose measurements were down­ with an abnormal screening glucose challenge test (G C I) loaded directly from the reflectance meter to a computer, but a normal O G T T result remain at increased risk of and patients underwent a 3-hour oral glucose tolerance adverse neonatal outcome.6-23-24 Langer and associates25 test (O G TT) using a standardized 100-g oral glucose load. Women were instructed to fast for at least 8 hours have demonstrated that women with only one abnormal result on O G T T have a 35% risk of neonatal complica­ before the OGTT. Gestational diabetes was diagnosed in accordance with the criteria of the National Diabetes Data tions. Because these women frequently fall into a diagnos­ Group, which requires that at least two of the following tic gray zone where recommendations for treatment are blood glucose values are met: > 1 0 5 mg/dL (5.8 mmol/ unclear, some investigators have considered lowering L), >190 mg/dL (10.5 mmol/L), >165 mg/dL (9.2 both the screening and the diagnostic criteria.24-26 mmol/L), and > 145 mg/dL (8.0 mmol/L) at 0, 1, 2, With the advent of newer technologies for determi­ nation o f blood glucose levels, other methods for identi­ and 3 hours, respectively.3 Following the OGTT, no further glucose testing was fying glucose intolerance in pregnancy can be evaluated. performed for patients who did not meet National Dia In gestational diabetes, frequent self-monitoring of blood

559 The Journal of Family Practice, Vol. 39, No. 6(Dec), 1994 Detecting Gestational Glucose Intolerance Peterson, Peterson, Corbett, et al

betes Data Group criteria for gestational diabetes. Pa­ Results tients diagnosed with GDM were treated with diet or insulin therapy, according to established protocols to ob­ Of the 101 women with a GCT >140 mg/dL (7.8 tain glycemic control.28 After delivery, the following out­ mmol/L) who were contacted for participation in the come data were obtained: infant birthweight, gestational study, 53 agreed to participate. The most common rea­ age, method of delivery, maternal weight gain, and pres­ sons given for refusal were lack o f time and hesitation to ence of maternal or infant complications during labor or undergo blood glucose testing seven times per day. There delivery. was no difference in insurance status or ethnicity between Glucose metabolism, measured by SMBG testing, the groups. No significant difference was found in the was characterized using Glucofacts (Miles Laboratories average plasma glucose at the 1-hour GCT between women who agreed to be in the study (158±14 mg/dL Inc, Diagnostics Division, Elkhart, Ind, 1989). This [8.8±0.8 mmol/L]) and those who declined (157±14 computer program downloads glucose meters and sum­ mg/dL [8.7±0.8 mmol/L]). Forty-eight women (91%) marizes individual glucose data including ranges and completed the study. O f the five women who dropped averages. Statistical analysis was performed using SAS out, four did not complete daily capillary blood glucose (SAS Institute Inc, Cary, NC). Women with an abnormal measurements, and one refused the 3-hour OGTT. There GGT but a normal OGTT were referred to as the “gray were no statistical differences for the GCT between those zone” group. A product-moment correlation coefficient who remained in the study and those who dropped out. was calculated for the gray zone group between infant Following the week of intensive self-monitoring, five birthweight and 18 study variables. The gray zone subjects were diagnosed with gestational diabetes by group was segmented into four subgroups based on OGTT, according to National Diabetes Data Group cri­ the individual’s average fasting SM BG level: group A, teria. Although no significant difference was found be­ s85 mg/dL (4.7 mmol/L); group B, 86 to 94 mg/dL tween this group and the gray zone group with respect to (4.8 to 5.2 mmol/L); group C, 95 to 104 mg/dL pregravid weight, gravida, or weight gain during preg­ (5.3 to 5.8 mmol/L); and group D, >105 mg/dl (5.8 nancy, the women with gestational diabetes had a signif­ mmol/L). Relative risks o f giving birth to an LGA infant icantly greater BMI (body mass index) (P = .0 3 2 ), higher for groups A through D were calculated using logit esti­ average screening values (PC .003), higher average fasting mators, with group A as the reference to which others values (P < .001), and higher average overall SMBG values were compared. (P<.001) than the gray zone group. O f the women with Iwo-by-two tables were corrected for empty cells GDM, two pregnancies resulted in LGA infants, one of using .5 added to every cell.29 Fisher’s exact test was whom experienced an episode o f neonatal hypoglycemia. employed to detect significance of birthweight difference Six women from the gray zone group had premature for each group as compared with group A. Trend analysis deliveries (less than 37 weeks’ gestation), one o f whom was performed using a Mantel-Haenszel statistic for had twins. Two additional sets o f twins were born. These trend. Univariate and bivariate regression were performed were excluded from analysis because o f inaccuracy in de­ to evaluate possible confounding effects from the follow­ termining LGA. Thirteen infants from the remaining 35 ing factors known to influence birthweight: gestational (37%) born to the eligible participants were LGA. This is age, multiparity, and maternal pregravid weight. Birth­ significantly greater than the 10% that would be expected from the distribution o f birthweights from the Brenner weight was the dependent variable and average fasting curve (P<.005). Additional complications identified at glucose the independent variable in the univariate (re­ the time o f delivery in the gray zone group included one duced) model. 1 he full model included the confounding late intrauterine fetal death, one neonatal hypoglycemia, variable of interest. The change in the beta coefficient for one mild shoulder dystocia, one cephalopelvic dispropor­ average fasting glucose was examined to reflect degree of tion, and two fourth-degree lacerations. confounding. Any change in the beta coefficient o f less The relationship between birthweight and 18 study than 10% was considered insignificant. variables was examined for the gray zone group patients. Birthweight percentiles of infants were obtained by Table 1 contains the correlation coefficient and signifi­ using a standard o f fetal growth in the United States cance for each study variable in estimating birthweight. (Brenner curve)30 to correct for differences in gestational Examination o f the O G T T resulted in no single value or age. LGA infants were identified as those ranking at or combination of values achieving a correlation with birth­ above the 90th percentile. Twins and infants born before weight percentile at the 95% confidence interval (Cl) in the 37th gestational week were excluded from these anal­ the gray zone group. On the other hand, several SMBG yses because o f inaccuracy in determining LGA. values significantly correlated with birthweight.

560 The Journal of Family Practice, Vol. 39, No. 6(Dec), 1994 Detecting Gestational Glucose Intolerance Peterson, Peterson, Corbett, et al

Table 1. Pearson Correlation Coefficients for Gray Zone Group Variables on Birthweight (excluding gestational age <37 weeks)

P Value Variable r Value (2-tailed) Pregravid weight .406 .017

Gravida .356 .039

Weight gain .116 .541

Family history .019 .918

Gestational age .385 .024

Oral glucose tolerance test Fasting .316 .074 1 hour .314 .074 MG/DL 2 hour .318 .071 Figure. Average infant birthweight percentile vs maternal aver­ 3 hour .048 .792 age fasting glucose at 28 weeks for groups A through D. Bars represent standard deviation; ; / MH denotes Mantel Haenszel Self-monitored blood glucose test statistic. Average glucose .409 .016 % above 120 .435 .010 Highest glucose .161 .364 tile in groups A through D rose significantly (PC.OOl), Lowest glucose .262 .134 Average fasting .552 .001 with average birthweight percentile increasing from 58%

Note: Gray zone group is defined as women with an abnormal glucose challenge test to 106% (Figure). The relative risk of producing an LGA but a normal oral glucose tolerance test. infant was approximately 13 times greater in women with a fasting SMBG value > 105 mg/dL (5.8 mmol/L) than that of women with a fasting SMBG value ^ 85 mg/dL The strongest correlate, fasting SMBG, was used to (4.7 mmol/L) (Table 3). Four women in the gray zone determine whether the SM BG association with infant group had one abnormal value on the OG 11.1 hree of birthweight was being confounded by other factors these women were in group B, and one was in group D. known to affect birthweight (pregravid weight, multipar­ Finally, we determined the sensitivity and specificity ity, or gestational age at delivery). A statistical model was of prediction of LGA infants by average fasting SMBG used to determine the change in correlation that occurred values obtained during the study. Only two infants of when the data were adjusted for these factors (Table 2). mothers diagnosed with GDM were LGA; however, since Increased gravida and gestational age did not affect the all mothers with GDM were aggressively treated, all five relationship between birthweight and fasting SMBG val­ of these infants were considered at risk for LGA and were ues. Pregravid weight mildly confounded the relation­ counted as LGA in this calculation. Eight of nine infants ship: the beta coefficient of the full model was 15% less with an average fasting SMBG value > 105 mg/dL (5.8 than that of the reduced model. mmol/L) were large for gestational age (specificity, 95% ; When the gray zone group was segmented according sensitivity, 44%; positive predictive value, 89%). Fifteen of to average SM BG , the average infant birthweight percen- 24 infants with an average fasting SMBG value > 95 mg/dL (5.3 mmol/L) were large for gestational age. Table 2. Contribution of Potential Confounding Factors to Average Fasting Self-monitored Blood Glucose (SMBG) Table 3. Relative Risk of Large-for-Gestational-Age Infants Correlation with Birthweight Among Women with High Screen and Normal Oral Glucose Tolerance Tests Grouped by Fasting Self-monitored Blood Beta Coefficient Standard Deviation Variables Glucose (SMBG) Levels______Average fasting SMBG 22.697 6.067 Fasting 5.976 Relative Average fasting SMBG 20.666 SMBG Birthweight Standard 51.795 Deviation Risk P Value + gravida 92.919 Group (mg/dL) No. Percentile LT3 00 VI 26.4 1 — 5.876 A 6 57.8 20.502 NS Average fasting SMBG 86-94 13 65.7 29.4 3.7 124.327 60.138 B + gestational age c 95-104 12 87 20.6 7.5 .05 10.1 13.6 .005 6.307 D >105 4 105.5 Average fasting SMBG 19.226 2.160 ote + pregravid weight 3.464 N : Twin and premature infants are excluded from analysis.

The Journal of Family Practice, Vol. 39, No. 6(Dec), 1994 Detecting Gestational Glucose Intolerance Peterson, Peterson, Corbett, et al

Using this value as a predictor of LGA infants, the sensi­ These values remained significant when corrected for the tivity increased to 83%, but specificity declined to 59%, correlation also seen between prenatal maternal weight, giving a positive predictive value o f 63%. multiparity, and variation of gestadonal age. It is notable that no significant correlation was found between the fasting value of the OGTT and the birth­ Conclusions weight percentile. It appears that each patient continued her regular lifestyle and meal habits during the period of The identification o f gestational diabetes and the rapid self-monitoring. Since patient behavior is altered during initiation o f therapy to restore euglycemia can effectively or preceding a physician visit, fasting blood glucose mea­ prevent the adverse perinatal and maternal outcomes that surements performed at a single clinic visit might give are directly associated with underlying hyperglycemia.8’31 misleading results. Patients have substantial daily variabil­ Women with an abnormal G C T result and a normal ity in blood glucose. The validity o f SM BG is significantly OGTT result, however, fall into a gray zone that is not increased by having multiple measures during the week. clearly gestational diabetes but for which maternal risks The fasting blood glucose measurements made during may be increased. In this study, a significantly higher- the 3-hour OGTT correlated only moderately at r=.58 than-expected proportion of LGA infants were born to (P<.001) with the average fasting SMBG values. The women with an abnormal GCT and normal OGTT (the average o f seven tests was a better predictor o f LGA and gray zone group). The women who gave birth to large seemed to be more representative of the subject’s overall infants could be distinguished from other women in the carbohydrate metabolism. gray zone by SMBG testing at 28 weeks’ gestation. Al­ Can we realistically expect SM BG testing from most all the LGA infants were born to women who had women with an abnormal GCT? Frequent SMBG testing persistent fasting hyperglycemia >95 mg/dL (5.3 requires training as well as a high degree o f patient moti­ mmol/L) over 7 days undetected by standard diagnostic vation. In general, pregnant women are highly motivated procedures. Three women had average fasting SM BG val­ concerning the welfare of their infants, and training with ues > 1 0 5 mg/dL (5.8 mmol/L) but had four normal new, simplified meters is fairly easy. Additionally, since values on the O G TT. All these women gave birth to fasting SM BG values were most highly correlated with infants who ranked higher than the 90th percentile in LGA, it may not be necessary to monitor as frequently as weight. These women probably represent unrecognized was done in this study. It should be noted, however, that cases o f GDM . SM BG monitoring may not be available to all patients. It has been suggested that the values of the National Pregnancy is accompanied by a lowered hematocrit, Diabetes Data Group criteria for the diagnosis of GDM which may prevent use o f standard capillary monitors in should be lowered.15-17 In this study, O G TT values below some patients. the standard criteria correlated poorly with birthweight. Since this study examined only women who already Lowering the OGTT criteria would be expected to iden­ had an abnormal 1-hour GCT, all women in this study tify more women at risk for giving birth to LGA infants would have a higher likelihood of having gestational dia­ (increased sensitivity), but at the expense of a higher false- betes than would those with normal GCTs. As a result, identification rate (decreased specificity). Even assuming the estimates o f sensitivity, specificity, and predictive that all women in this study who were diagnosed with value would be expected to be overstated if applied to GDM would have given birth to LGA infants if the women without a positive screening test. In addition, it women had not been treated (surely an overestimation), should be noted that several (n= 18) variables were used current criteria would have anticipated only 28% of the in the examination o f patients in this study. Because sev­ infants who were born large for gestational age. When the eral variables were used, the associations found may ap­ criteria of the OGTT were lowered to 95, 180, 155, and pear more impressive than they may prove to be on con­ 140 mg/dL (5.3, 10.0, 8.6, and 7.8 mmol/L), the sen­ firmatory investigation, and clinical implementation sitivity increased to 39%, resulting in the identification o f should await further study. four more women, two of whom gave birth to LGA in­ SMBG testing appears to allow characterization of fants. The poor predictive value o f the O G T T was in the risk o f LGA infants among women in the gray zone. contrast to the better predictive value of self-monitoring. This may be accomplished by a short period o f daily mon­ Significant correlation was found to exist between itoring with a glucose reflectance meter. Women with an infant birthweight and several SM BG values. In addition average fasting capillary glucose >95 mg/dL (5.3 to the average fasting blood glucose, the average overall mmol/L) seem to be at increased risk for LGA infants, blood glucose and the percentage o f values > 1 2 0 mg/dL and women with fasting capillary glucose levels >105 (6.7 mmol/L) were also correlated with infant weight. mg/dL (5.8 mmol/L) are certainly at risk and may rep-

562 The Journal of Family Practice, Vol. 39, No. 6(Dec), 1994 Detecting Gestational Glucose Intolerance Peterson, Peterson, Corbett, et al

resent cases o f GD M missed by the OGTT. Further stud­ HbAlc and sensitivity of various degrees of glucose intolerance. ies are needed to determine the cost of such follow-up, to Ann Intern Med 1979; 9 1 :3 9 0 -6 . determine the extent to which adverse outcomes other 15. Roberts AB, Baker JR, Metcalf P, Mullard C. Fructosamine com­ pared with a glucose load as a screening test for gestational diabetes. than LGA infants are identified by SMBG values, and to Obstet Gynecol 1990; 76:773-5. compare the SM BG level with modified criteria for the 16. Bourgeois FJ, Harbert GM, Paulsoen HP, Thiagarajah S 11. Glyco­ OGTT in determining risk o f adverse neonatal outcome. sylated serum protein level as a screening and diagnostic test for gestational diabetes mellitus. Am J Obstet Gvnecol 1986; 155: 493-6. 17. Buchanan TA. Glucose metabolism during pregnancy: normal References physiology and implications for diabetes mellitus. Isr I Med Sci 1991;27:432-41. 1. ACOG Technical Bulletin No. 92. Management of diabetes melli- 18. Freinkel N. Banting lecture 1980. O f pregnancy and progeny. Di­ tus in pregnancy. Washington, DC: American College of Obstetri­ abetes 1980;29:1023-35. cians and Gynecologists, May 1986. 19. Coustan DR, Imarah J. Prophylactic insulin treatment of gesta­ 2. ADA position statement: gestational diabetes mellitus. Diabetes tional diabetes reduces the incidences of macrosomia, operative Care 1 9 8 6 ;9 :4 3 0 -1 . delivery and birth trauma. Am J Obstet Gynecol 1984; 150:836-42. 3. National Diabetes Data Group. Classification and diagnosis of dia­ 20. Green J, Pawson I, Schumacher L, Perry J, Kretchmer N. Glucose betes mellitus and other categories of glucose intolerance. Diabetes tolerance in pregnancy: ethnic variation and influence of body hab­ 1979;28:1039-57. itus. Am J Obstet Gynecol 1990; 163:86-92. 4. O’Sullivan JB , Mahan CM. Criteria for the oral glucose tolerance 21. hanger O, Mazze R. The relationship between large-for-gcstation test in pregnancy. Diabetes 1964; 13:278-85. al-age infants and glycemic control in women with gestational dia­ 5. O’Sullivan JB , Charles D, Mahan C, Dandrow R. Gestational dia­ betes. Am J Obstet Gynecol 1988; 159:1478-83. betes and perinatal mortality rate. Am J Obstet Gynecol 1973; 116: 22. Karlsson K, Kjellmer I. The outcome of diabetic pregnancies in 901 -4 . relation to the mother’s blood sugar level. Am ) Obstet Gynecol 6. heiken E, Jenkins J, Pomerantz G, Klein L. Abnormal glucose 1972; 112:213-20. screening tests in pregnancy: a risk factor for fetal macrosomia. 23. Tallargio h, Giampietro O, Penno G, Miccoli R, Gregori G, Nave Obstet Gynecol 1987; 69:570-3. lesi R. Relation of glucose tolerance to complication of pregnancy in 7. Canadian Task Force on the Periodic Health Examination. Periodic nondiabetic women. N Engl J Med 1986; 315:989-92. health examination, 1992 update. Screening for gestational diabe­ 24. Magee M, Walden C, Beiedetti T, Knapp R. Influences of diagnos­ tes mellitus. Can Med Assoc J 1992; 147:437-43. tic criteria on the incidence of gestational diabetes and prenatal 8. Neiger R. Fetal macrosomia in the diabetic patient. Clin Obstet morbidity. JAMA 1993; 269:609-15. Gynecol 1992; 35:138-50. 25. hanger 6 , Brustman h, Anyaegbunam A, Mazze R. The signifi­ 9. ACOG Technical Bulletin No. 159. Fetal macrosomia. Washing­ cance of one abnormal glucose tolerance test value on adverse ton, DC: American College of Obstetricians and Gynecologists, outcome in pregnancy. Am J Obstet Gynecol 1987; 157:758-63. September 1991. 26. Carpenter MW, Coustan DR. Criteria for screening tests for gesta­ 10. Pettitt D J, Bennett PH, Knowler WC, Baird R, Aleck K. Gesta­ tional diabetes. Am J Obstet Gynecol 1982; 144:768-73. tional diabetes mellitus and impaired glucose tolerance in the off­ 27. Mazze R, hucido D, hanger O, Hartmann 1C, Rodbard D. Ambu­ spring. Diabetes 1985; 34(suppl 2 ):1 19-22. latory glucose profile: representation of verified self-monitered 11. hanger O. Abnormal fetal growth: rationale for achieving cuglyce- blood glucose data. Diabetes Care 1987; 10:111-7. mia. Presented at Contemporary Issues in Maternal Fetal Medicine: 28. Summary and recommendations of the Third International Work­ High Risk Screening and Diabetes, San Antonio, Texas, April 23, shop-Conference on Gestational Diabetes Mellitus. Diabetes 1991; 40(suppl 2):197—201. 1990. 12. Rizzo T , Freinkel N, Metzger B, Hatcher R, Burns W, Barglow P. 29. Cochran W. Some methods for strengthening the common y2 test. Fuel mediated behavioral teratogenesis: correlations between ma­ Biometrics 1954; 10:417-51. ternal metabolism in diabetic pregnancies and Brazelton tests in the 30. Brenner W, F.delman D, Hendricks C. A standard of fetal growth for the United States of America. Am J Obstet Gynecol 1976; newborn. Diabetes 1988; 37(suppl):86A. 13. Bolli G, Compagnucci P, Cartechini M, Santeusanio F, Cirotto C, 126:555-64. 31. hanger O, Anyaegbunam A, Brustman h, Divon M. Management Scionti L, Brunctti P. HbAlc in subjects with abnormal glucose of women with one abnormal oral glucose tolerance test value tolerance by normal fasting plasma glucose. Diabetes 1979, 29. reduces adverse outcome in pregnancy. Am J Obstet Gynecol 1989; 272-7. 14. Dunn RJ, Cole RA, Soeldner JS, Gleason RE. Reproducibility of 161:593-9.

563 The Journal of Family Practice, Vol. 39, No. 6(Dec), 1994 JFP Journal Club

PREDICTING LARGE-FOR-GESTATIONAL- birthweights. Large for gestational age was defined as AGEINFANTS above the 90th percentile birthweight. Regression analy­ sis was performed to control for the following confound­ T it l e : Comparison of home glucose monitoring with the ing variables: gestational age, multiparity, and maternal oral glucose tolerance test to detect gestational glucose pregravid weight. Similar analysis was not reported for intolerance other potential confounders, including maternal age, A u t h o r s : Peterson KA, Peterson AM, Corbett V, weight gain, history of giving birth to an LGA infant, Tongen S, Guzman M, Mazze R. family history o f diabetes, or previous stillbirth or congen­ J o u r n a l : 7 he Journal of Family Practice ital malformations. D a t e : December 1994; Volume 39:558-63. Results. O f the 48 consenting women with an abnormal G C T who completed the study, five met the diagnostic Background. Gestational diabetes mellitus (GDM ), which criteria for having GDM. Infants born to two of these is glucose intolerance first recognized during pregnancy, women were LGA. One of these infants experienced neo­ occurs in 2% to 3% o f pregnancies.1 GDM has been asso­ natal hypoglycemia. No information was provided re­ ciated with an increased risk of perinatal complications. garding the severity of this event. Thirteen of the 43 Concern exists regarding similar risks for pregnant women not meeting the diagnostic criteria for GDM gave women with an abnormal 1-hour screening glucose chal­ birth to LGA infants. Complications of these deliveries lenge test (GCT) but a normal 3-hour oral glucose toler­ included one intrauterine fetal demise, one neonatal hy­ ance test (OGTT). poglycemia, one mild shoulder dystocia, one cephalopel- Clinical question. Can self-monitored blood glucose lev­ vic disproportion, and two 4th-degree lacerations. The els predict infant morbidity, as measured by the occur­ strongest predictor for increased birthweight among the rence of large-for-gcstational-age (LGA) infants, among women not diagnosed with GDM was the average fasting women with an abnormal GCT but a normal OGTT? self-monitored blood glucose level. This relationship was confounded by pregravid weight but not by parity or Population studied. Pregnant women were recruited from gestational age. The risk for a woman giving birth to an family medicine and obstetric practices associated with a LGA infant increased proportionally with an increase in large metropolitan hospital center. Potential study sub­ her average fasting self-monitored blood glucose. jects were identified after an abnormal GCT was obtained between the 24th and 29th gestational week. Exclusion Recommendations for clinical practice. In this trial of criteria included a history of GDM or a hemoglobin less pregnant women with an abnormal G C T, intensive home than 10.0 g/dL (100 g/L). No information was given blood glucose monitoring was more effective than the regarding age, race, or ethnicity. standard O G T T in predicting which women will deliver LGA infants. The authors have taken advantage o f sophis­ Study design and validity. Consenting women with an ticated technology to record and analyze 49 blood glu­ abnormal GCT measured capillary blood glucose with a cose determinations for each patient. home glucometer seven times daily for 1 week while re­ The difficulties inherent in the intensive glucose maining on an unrestricted diet. They then underwent a monitoring may, however, have affected the results of the standard 3-hour OGTT. Women diagnosed with GDM study. Only one half o f the women who were eligible were treated with either diet or insulin therapy, according agreed to participate in this study. These women may to standard protocols. Following the OGTT, no further have been more willing to undergo the frequent blood glucose testing was performed for padents who did not glucose determinations because they were at an increased meet the standard diagnostic criteria for GDM. risk of either glucose intolerance or giving birth to an Outcomes measured. Statistical correlations were made be­ LGA infant, or both, thus potentially increasing the sta­ tween self-monitored blood glucose levels and infant tistical correlation between these two outcomes. Unlike the current study, previous reports show that any associ­ ation between blood glucose levels and birthweight is lost Drs Slawson and Squillace are from the Department o f Family Medicine, University o f Virginia Health Sciences Center, Charlottesville, Virginia. DrShaughnessyisfrom when adjusted for maternal weight and age.1 the Harrisburg Hospital Family Practice Residency Program, Harrisburg, Pennsyl­ A second difficulty with applying the results of this vania. Requestsfo r reprints should be addressed to David Slawson, MD, Department study to clinical practice is that the expense and discom­ o f Family Medicine, University o f Virginia Health Sciences Center, Box 414, Char­ lottesville, VA 22908. fort associated with the diagnostic procedure— almost 50

© 1994 Appleton & Lange ISSN 0094-3509 592 The Journal of Family Practice, Vol. 39, No. 6(Dec), 1994 JFP Journal Club Slawson, Shaughnessy, and Squillace capillary blood glucose determinations—may not be ac­ this group developed a set of decision rules that were ceptable to many pregnant patients. In addition, the com­ refined and validated through prospective evaluation.1-2 puter support to evaluate the results may not be available to most clinicians. It is unknown whether less extensive Clinical question. Does the instruction of physicians in the monitoring is as useful. implementation of the Ottawa ankle rules have the poten­ A third, broader concern is that no consensus exists tial to change clinical behavior and result in a decrease in regarding the need for universal screening of pregnant the number of patients referred for ankle and foot radio- women for glucose intolerance.1-4 This study suggests graphic series? that 1 week o f intense self-monitored blood glucose levels Population studied. All adult patients presenting to either will identify more pregnant women who will deliver LGA of two emergency departments with acute ankle pain re­ infants. It does not help us determine whether this iden­ sulting from blunt trauma, including twisting, falls, and tification benefits either the mother or the infant. Preg­ direct blows, were enrolled in the study. A community nant women falsely identified with glucose intolerance hospital facility' with no house staff, which served as the may be considered at “ high risk” and subject to extensive control facility', enrolled patients during two 5-month and expensive testing and interventions. Women at risk blocks (February through June) in 1991 and 1992. The for delivering LGA infants with no evidence of glucose intervention hospital was a teaching institution and en­ intolerance may be falsely reassured and not treated. Ev­ rolled patients during the same 5-month block in 1990 idence from randomized controlled trials showing an (before-intervention) and 1992 (after-intervention). Ex­ overall benefit from screening for glucose intolerance in clusion criteria included age less than 18 years, pregnancy, pregnancy is clearly needed. presence of isolated injuries of the skin, referrals from David C. Slawson, MD outside hospitals with radiographs, more than a 10-day Allen F. Shaughnessy, PharmD interval since injury', and reassessment o f an earlier injury.

Study design and validity. The study was a nonrandom- ized controlled trial comparing rates of ankle and foot References radiographic series before and after intervention. Attend­ 1. Stephenson, MJ. Screening for gestational diabetes mellitus: a crit­ ing physicians and house staff at the intervention institu­ ical review. J Fam Pract 1993; 37:277-83. tion were instructed in the use o f the ankle rules through 2. Enkin M, Keirse M J, Chalmers I. A guide to effective care in preg­ lectures, handouts, pocket cards, and two posters nancy and childbirth. 1st ed. Oxford, UK: Oxford University Press, 1989: 41-3. mounted in the ED. Data forms were attached to patient 3. Jarrett RJ. Gestational diabetes: a non-entity? BMJ 1993; 306: charts reminding physicians to evaluate for specific areas 3 7 -8 . of tenderness and the ability' of the injured ankle to bear 4. Hunter DJS. Is gestational diabetes a clinical entity? Can Fam Phy­ sician 1984; 3 0 :1319-21. weight. The decision to refer for radiography was left to the treating physicians. All radiographs were evaluated by a radiologist and final reports were reviewed for all cases. All patients in the after-intervention group were fol­ GUIDELINES FOR MANAGING ANKLE lowed up by telephone to assess progress of healing, im­ INJURIES: The Ottawa Ankle Rules______pact of the injury on lifestyle, and patient satisfaction with their emergency care. Patients who were not studied ra­ T it l e : Implementation o f the Ottawa Ankle Rules diographically were contacted by phone a second time 10 Au t h o r s : Stiell IG, McKnight RD, Greenberg GH, days after their ED visit and asked to return for reevalua­ McDowell I, Nair RC, Wells GA, Johns C, tion if they remained symptomatic. Worthington JR Inclusion of the data form with the patient’s chart J o u r n a l : JAMA during the intervention period may have falsely elevated D a t e : March 16, 1994; Volume 271:827-32. compliance with the rules. Use o f radiography was mon­ itored for an additional 12 months after the intervention Background. Although acute ankle injuries are one of the period when data collection sheets were no longer used. most common chief complaints of patients presenting for The use o f radiographic studies remained unchanged dur­ care in emergency departments (EDs), there are no well- ing this period. The inclusion of house officers as physicians in the validated guidelines to assist physicians in deciding when intervention hospital may have made this study particu­ radiographic evaluation o f the injured extremity is neces­ larly susceptible to the Hawthorne effect. House officers’ sary. As a result, patients are usually referred for a radio­ behavior may have been significantly more likely to graph to rule out treatable fracture. Previous studies by

593 The Journal of Family Practice, Vol. 39, No. 6(Dec), 1994 JFP Journal Club Slawson, Shaughnessy, and Squillace

change because they were being observed, and this behav­ should use caution when applying the rules to patients ioral change might not have continued after the period o f who may be difficult to reliably evaluate, such as those observation. The lack of change in rates of radiography who are intoxicated, have a head injury, have multiple during the 12-month follow-up period suggests that ob­ painful injuries, or have diminished sensation related to servation alone did not account for the decline in radiog­ neurological deficits. Patients discharged with radio- raphy use. graphic evaluation should be instructed to return for fur­ ther evaluation if pain and ability to bear weight have not Outcomes measured. The primary outcome measured for improved in 5 to 7 days. all patients was the proportion referred for ankle or foot The investigators suggest a stepwise evaluation of the radiographic series. Additional outcomes measured were injured ankle, which begins with asking the patient about the direct cost of medical care, time spent in the ED, the circumstance of the injury and the patient’s ability to persistence of symptoms, ability to walk and return to bear weight within the first hour after injury. The exam­ work, and subsequent physician visits and radiographs. iner then proceeds with palpation of the ankle, concen­ Results. A total of 2342 patients were seen during the trating on the posterior edge of both malleoli (from prox­ before- and after-intervention periods. A relative reduc­ imal to tip including the entire distal 6 cm o f the fibula), tion of 28% of patients referred for ankle radiographs was the base of the fifth metatarsal, and the navicular. Con­ seen at the intervention hospital between the two study centrating the examination on the posterior edge of the periods (P<.001). During the same interval, there was a malleoli avoids producing pain secondary only to liga­ 2% relative increase in the use o f the same studies at the mentous tenderness from the anterior edge. Patients with control hospital (P=NS). Twenty-four percent of patients localized tenderness of the posterior malleoli or fibula are at the intervention hospital had no radiographic study referred for ankle radiography. Patients with tenderness of during the postintervention period, compared with 4% of the fifth metatarsal or navicular are referred for a foot patients during the preintervention period. radiographic series. Patients with no localized tenderness Patients discharged from the ED without radio­ are then asked to walk two steps on each lower extremity graphs spent significantly less time in the ED (80 minutes without assistance. If they can accomplish this task, they vs 116 minutes, P<.0()1). Ninety-five percent of the non­ are considered “able to bear weight,” regardless of how radiograph group reported being satisfied with their care, much they limp. Patients who are unable to bear weight compared with 96% of patients who were x-rayed. The both immediately and in the ED but who have no local­ nonradiograph group had fewer follow-up visits (7% vs ized tenderness are referred for both ankle and foot radio- 20%, P<.001) and a significantly lower total cost of care graphic series. Patients with no localized tenderness and ($62 vs $173, P<.001). None of the patients discharged who retain their ability to bear weight are discharged from the ED without radiographs were subsequently without radiography. found to have fractures. Susan P. Squillace, MD Recommendations for clinical practice. This study’s inves­ David C. Slawson, MD tigators have identified a group o f easily applied decision rules that have demonstrated a 30% reduction in the num­ ber of radiographs ordered with no reduction in the abil­ References ity to properly and promptly diagnose clinically significant 1. Stiell IG, Greenberg GH, McKnight RD, Nair RC, McDowell I, Worthington JR. A study to develop clinical decision rules for the fractures of the ankle or midfoot. use of radiography in acute ankle injuries. Ann Emerg Med 1992; The authors offer several caveats regarding appropri­ 2 1 :3 8 4 -9 0 . ate application o f the rules. Because patients less than 18 2. Stiell IG, Greenberg GH, McKnight RD, Nair RC, McDowell I, Reardon M, et al. Decision rules for the use of radiography in acute years old were excluded from the study, the rules cannot ankle injuries. Refinement and prospective validation. JAMA 1993; be generalized to the pediatric population. Physicians 269:1127-32.

594 The Journal of Family Practice, Vol. 39, No. 6(Dec), 1994