DIABETES TECHNOLOGY & THERAPEUTICS Volume 17, Number 10, 2015 ª Mary Ann Liebert, Inc. DOI: 10.1089/dia.2014.0267
ORIGINAL ARTICLE
Screening for Type 2 Diabetes and Dysglycemia in Saudi Arabia: Development and Validation of Risk Scores Ziad A. Memish, MD,1 Jennifer L. Chang, MPH, MSN,2 Mohammad Y. Saeedi, MD,1 Mohammad A. Al Hamid, MD,1 Omer Abid, MD,1 and Mohammed K. Ali, MBChB, MSc, MBA 2
Abstract
Objective: The prevalence of type 2 diabetes in Saudi Arabia is the highest worldwide after excluding small island nations. We developed and validated a noninvasive screening test based on demographic and clinical data for identifying adults with undiagnosed diabetes and dysglycemia in Saudi Arabia. Research Design and Methods: Data from 1,485 nonpregnant Saudi adults ‡20 years of age without a current diagnosis of diabetes were obtained from urban and rural primary healthcare centers in 2009. Clinical and demographic data were obtained through physician-administered interviews. Oral glucose tolerance test data were used to define diabetes (fasting plasma glucose ‡7.0 mmol/L or 2-h post-load glucose ‡11.1 mmol/L) and dysglycemia (fasting plasma glucose ‡5.6 mmol/L or 2-h post-load glucose ‡7.8 mmol/L). Predictive models were developed using data from 1,435 individuals. Multivariable logistic regression and receiver operating characteristic curves were used to develop and evaluate a separate risk score for both diabetes and dysglycemia. Scores were validated on a hold-out sample of 50 individuals. Results: The risk score for undiagnosed diabetes contained age, history of gestational diabetes, smoking, family history of diabetes, and central obesity with a sensitivity of 76.6% and a specificity of 52.1%. The dysglycemia risk score contained age, gestational diabetes, hypertension, and central obesity with a sensitivity of 71.2% and a specificity of 54.0%. All performed equally well, if not better, in the hold-out sample. Conclusions: These risk scores can identify Saudi adults with undiagnosed diabetes or dysglycemia and should be validated in prospective studies. Introduction
D
iabetes was estimated to affect 382 million adults (20–79 years of age) worldwide in 2013.1 More than 90% of diabetes cases are type 2 diabetes mellitus, a progressive disease that leads to organ dysfunction and, in particular, the emergence of retinopathy, neuropathy, and nephropathy.2 Type 2 diabetes also increases the risk of stroke, heart disease, and all-cause mortality by 1.4–4.5 times.3–5 After six small island nations, Saudi Arabia has the highest national diabetes prevalence (20.2%) worldwide, and this percentage is projected to increase to 27.1% by 2035.1 From the literature, the prevalence of diabetes has been found to be approximately 23% in a crosssectional citywide study to 24% in a large nationwide epidemiologic study.6,7 1 2
Early glycemic control reduces the risk for micro- and long-term macrovascular complications and mortality in people with diabetes.8,9 However, type 2 diabetes is an insidious disease, and a little fewer than half of those with diabetes worldwide remain unaware until symptoms or complications develop.1,2 In the United States, 32% of adults with diabetes remain undiagnosed, and this proportion has been noted to be as high as 60.7% in developing countries like China.10,11 Dysglycemia, the inadequate regulation of blood glucose levels, refers to both prediabetes (the precursor phase before diabetes) and type 2 diabetes.2 People with prediabetes have an annual risk of developing diabetes that is five to 10 times higher than that of people with normal glucose levels and tolerance.12 However, robust evidence has shown that lifestyle interventions can slow the progression from prediabetes to diabetes. Research trials show that
Ministry of Health, Riyadh, Kingdom of Saudi Arabia. Emory University, Atlanta, Georgia, USA.
693
694
intensive lifestyle modification can reduce diabetes incidence by half with lasting reductions of 34–45% lower diabetes incidence over a period of 9–23 years of follow-up.13–15 Although the imperative is to identify people with prediabetes or type 2 diabetes as soon as possible to begin interventions, current diagnostic tests for type 2 diabetes—fasting plasma glucose (FPG) or oral glucose tolerance tests—are expensive and inconvenient to be administered populationwide.16,17 Screening for diabetes with risk scores can identify asymptomatic or unaware individuals who are more likely to have the disease, which reduces the number of people who should receive further diagnostic testing.17 Risk scores for diabetes have been shown to be potentially cost-effective, especially by detecting prediabetes and reducing the incidence of type 2 diabetes through early intervention.18,19 A previous effort to develop a type 2 diabetes and dysglycemia risk score in Saudi Arabia used data from a small cross-sectional study (composed of Arab and non-Arabs) from mosques and malls in two cities (Riyadh and Jeddah), reducing the generalizability of these tools.20 We aimed to develop risk scores for undiagnosed type 2 diabetes and dysglycemia in Saudi Arabia using recent nationwide crosssectional data. Materials and Methods Study design
From May to June of 2009, Saudi nationals were recruited from both rural and urban primary healthcare centers (PHCCs) by dividing Saudi Arabia into five regions (North, East, West, South, and Central). Each region contained three to five subregions. Data were collected from one rural and two urban randomly chosen PHCCs from one randomly chosen subregion. From the medical records of nonpregnant participants at least 20 years old without a current diagnosis of diabetes, a target sample of randomly selected 240 participants (120 males and 120 females, with 20 of each sex in each 10-year age strata) from each urban PHCC and 120 participants (60 males and 60 females, with 10 of each sex stratified by age in 10-year increments from 20 to 60+ years) from each rural PHCC was contacted. No incentives were given. In total, 2,671 participants gave written informed consent and were recruited. Data collection
Participants were either invited to their PHCC or visited by a physician who administered a questionnaire regarding sociodemographic, lifestyle, past medical, and family history characteristics. Blood pressure (in mm Hg) and anthropometric measurements (waist circumference [in cm], height [in cm], and weight [in kg]) were also measured. Participants were instructed to fast overnight for at least 8 h before going to their respective PHCC for an oral glucose tolerance test to measure FPG (in mmol/L) followed by ingestion of a 75-g anhydrous solution and measurement of the 2-h post-load glucose (2-h PP [in mmol/L]). Study variables
The primary outcome variables were the presence of type 2 diabetes or dysglycemia. The presence or absence of type 2 diabetes was defined by the American Diabetes Association
MEMISH ET AL.
guidelines. Diabetes was defined as an FPG level of ‡7.0 mmol/L or 2-h PP level of ‡11.1 mmol/L.21 Dysglycemia was defined as either an FPG level of ‡5.6 mmol/L or 2-h PP level of ‡7.8 mmol/L.21 From patients’ reports, current age stratum (20–29, 30–39, 40–49, 50–59, and 60+ years), sex, highest education level (illiterate, literate, primary, secondary, and tertiary education or higher), and marital status (single, married [one wife], married [multiple wives], and divorced) were recorded. Reported lifestyle factors were classified for smoking status (current, former, or never for cigarettes, cigar, pipe, shisha, or guza [flavored tobacco]), average number of days per week engaged in moderate to vigorous activity, and average servings of fruits and/or vegetables consumed per day (1–5 units/day). For the latter, missing values were recoded as having an average intake of zero vegetables and/or fruits per day, as that was not an option on the questionnaire. Medical history included questions related to the absence or presence of hypertension, gestational diabetes, previous abnormal blood glucose concentrations, and a family history of diabetes (parents and/or siblings). ‘‘Don’t know’’ was coded as missing data. The presence of hypertension was defined by self-report or by an average of one to three current blood pressure measurements as a systolic pressure of 140 mm Hg or higher or a diastolic of 90 mm Hg or higher.22 Body mass index (BMI) was calculated as weight (in kg)/(height [in m])2 and categorized as underweight or normal (200 kg), BMI (if ‡70 kg/m2), waist (if £39 cm), average diastolic blood pressure (if
ORIGINAL ARTICLE
Screening for Type 2 Diabetes and Dysglycemia in Saudi Arabia: Development and Validation of Risk Scores Ziad A. Memish, MD,1 Jennifer L. Chang, MPH, MSN,2 Mohammad Y. Saeedi, MD,1 Mohammad A. Al Hamid, MD,1 Omer Abid, MD,1 and Mohammed K. Ali, MBChB, MSc, MBA 2
Abstract
Objective: The prevalence of type 2 diabetes in Saudi Arabia is the highest worldwide after excluding small island nations. We developed and validated a noninvasive screening test based on demographic and clinical data for identifying adults with undiagnosed diabetes and dysglycemia in Saudi Arabia. Research Design and Methods: Data from 1,485 nonpregnant Saudi adults ‡20 years of age without a current diagnosis of diabetes were obtained from urban and rural primary healthcare centers in 2009. Clinical and demographic data were obtained through physician-administered interviews. Oral glucose tolerance test data were used to define diabetes (fasting plasma glucose ‡7.0 mmol/L or 2-h post-load glucose ‡11.1 mmol/L) and dysglycemia (fasting plasma glucose ‡5.6 mmol/L or 2-h post-load glucose ‡7.8 mmol/L). Predictive models were developed using data from 1,435 individuals. Multivariable logistic regression and receiver operating characteristic curves were used to develop and evaluate a separate risk score for both diabetes and dysglycemia. Scores were validated on a hold-out sample of 50 individuals. Results: The risk score for undiagnosed diabetes contained age, history of gestational diabetes, smoking, family history of diabetes, and central obesity with a sensitivity of 76.6% and a specificity of 52.1%. The dysglycemia risk score contained age, gestational diabetes, hypertension, and central obesity with a sensitivity of 71.2% and a specificity of 54.0%. All performed equally well, if not better, in the hold-out sample. Conclusions: These risk scores can identify Saudi adults with undiagnosed diabetes or dysglycemia and should be validated in prospective studies. Introduction
D
iabetes was estimated to affect 382 million adults (20–79 years of age) worldwide in 2013.1 More than 90% of diabetes cases are type 2 diabetes mellitus, a progressive disease that leads to organ dysfunction and, in particular, the emergence of retinopathy, neuropathy, and nephropathy.2 Type 2 diabetes also increases the risk of stroke, heart disease, and all-cause mortality by 1.4–4.5 times.3–5 After six small island nations, Saudi Arabia has the highest national diabetes prevalence (20.2%) worldwide, and this percentage is projected to increase to 27.1% by 2035.1 From the literature, the prevalence of diabetes has been found to be approximately 23% in a crosssectional citywide study to 24% in a large nationwide epidemiologic study.6,7 1 2
Early glycemic control reduces the risk for micro- and long-term macrovascular complications and mortality in people with diabetes.8,9 However, type 2 diabetes is an insidious disease, and a little fewer than half of those with diabetes worldwide remain unaware until symptoms or complications develop.1,2 In the United States, 32% of adults with diabetes remain undiagnosed, and this proportion has been noted to be as high as 60.7% in developing countries like China.10,11 Dysglycemia, the inadequate regulation of blood glucose levels, refers to both prediabetes (the precursor phase before diabetes) and type 2 diabetes.2 People with prediabetes have an annual risk of developing diabetes that is five to 10 times higher than that of people with normal glucose levels and tolerance.12 However, robust evidence has shown that lifestyle interventions can slow the progression from prediabetes to diabetes. Research trials show that
Ministry of Health, Riyadh, Kingdom of Saudi Arabia. Emory University, Atlanta, Georgia, USA.
693
694
intensive lifestyle modification can reduce diabetes incidence by half with lasting reductions of 34–45% lower diabetes incidence over a period of 9–23 years of follow-up.13–15 Although the imperative is to identify people with prediabetes or type 2 diabetes as soon as possible to begin interventions, current diagnostic tests for type 2 diabetes—fasting plasma glucose (FPG) or oral glucose tolerance tests—are expensive and inconvenient to be administered populationwide.16,17 Screening for diabetes with risk scores can identify asymptomatic or unaware individuals who are more likely to have the disease, which reduces the number of people who should receive further diagnostic testing.17 Risk scores for diabetes have been shown to be potentially cost-effective, especially by detecting prediabetes and reducing the incidence of type 2 diabetes through early intervention.18,19 A previous effort to develop a type 2 diabetes and dysglycemia risk score in Saudi Arabia used data from a small cross-sectional study (composed of Arab and non-Arabs) from mosques and malls in two cities (Riyadh and Jeddah), reducing the generalizability of these tools.20 We aimed to develop risk scores for undiagnosed type 2 diabetes and dysglycemia in Saudi Arabia using recent nationwide crosssectional data. Materials and Methods Study design
From May to June of 2009, Saudi nationals were recruited from both rural and urban primary healthcare centers (PHCCs) by dividing Saudi Arabia into five regions (North, East, West, South, and Central). Each region contained three to five subregions. Data were collected from one rural and two urban randomly chosen PHCCs from one randomly chosen subregion. From the medical records of nonpregnant participants at least 20 years old without a current diagnosis of diabetes, a target sample of randomly selected 240 participants (120 males and 120 females, with 20 of each sex in each 10-year age strata) from each urban PHCC and 120 participants (60 males and 60 females, with 10 of each sex stratified by age in 10-year increments from 20 to 60+ years) from each rural PHCC was contacted. No incentives were given. In total, 2,671 participants gave written informed consent and were recruited. Data collection
Participants were either invited to their PHCC or visited by a physician who administered a questionnaire regarding sociodemographic, lifestyle, past medical, and family history characteristics. Blood pressure (in mm Hg) and anthropometric measurements (waist circumference [in cm], height [in cm], and weight [in kg]) were also measured. Participants were instructed to fast overnight for at least 8 h before going to their respective PHCC for an oral glucose tolerance test to measure FPG (in mmol/L) followed by ingestion of a 75-g anhydrous solution and measurement of the 2-h post-load glucose (2-h PP [in mmol/L]). Study variables
The primary outcome variables were the presence of type 2 diabetes or dysglycemia. The presence or absence of type 2 diabetes was defined by the American Diabetes Association
MEMISH ET AL.
guidelines. Diabetes was defined as an FPG level of ‡7.0 mmol/L or 2-h PP level of ‡11.1 mmol/L.21 Dysglycemia was defined as either an FPG level of ‡5.6 mmol/L or 2-h PP level of ‡7.8 mmol/L.21 From patients’ reports, current age stratum (20–29, 30–39, 40–49, 50–59, and 60+ years), sex, highest education level (illiterate, literate, primary, secondary, and tertiary education or higher), and marital status (single, married [one wife], married [multiple wives], and divorced) were recorded. Reported lifestyle factors were classified for smoking status (current, former, or never for cigarettes, cigar, pipe, shisha, or guza [flavored tobacco]), average number of days per week engaged in moderate to vigorous activity, and average servings of fruits and/or vegetables consumed per day (1–5 units/day). For the latter, missing values were recoded as having an average intake of zero vegetables and/or fruits per day, as that was not an option on the questionnaire. Medical history included questions related to the absence or presence of hypertension, gestational diabetes, previous abnormal blood glucose concentrations, and a family history of diabetes (parents and/or siblings). ‘‘Don’t know’’ was coded as missing data. The presence of hypertension was defined by self-report or by an average of one to three current blood pressure measurements as a systolic pressure of 140 mm Hg or higher or a diastolic of 90 mm Hg or higher.22 Body mass index (BMI) was calculated as weight (in kg)/(height [in m])2 and categorized as underweight or normal (200 kg), BMI (if ‡70 kg/m2), waist (if £39 cm), average diastolic blood pressure (if
