Screening for Type 2 Diabetes Mellitus: A Systematic Review for the U.S. Preventive Services Task Force

Shelley Selph; Tracy Dana; Ian Blazina; Christina Bougatsos; Hetal Patel; Roger Chou

doi:10.7326/M14-2221

Reviews |2 June 2015

Screening for Type 2 Diabetes Mellitus: A Systematic Review for the U.S. Preventive Services Task Force Free

Shelley Selph, MD, MPH; Tracy Dana, MLS; Ian Blazina, MPH; Christina Bougatsos, MPH; Hetal Patel, MD; Roger Chou, MD

Article, Author, and Disclosure Information

This article was published online first at www.annals.org on 14 April 2015.

From Pacific Northwest Evidence-based Practice Center and Oregon Health & Science University, Portland, Oregon.
Acknowledgment: The authors thank Agency for Healthcare Research and Quality Medical Officer Quyen Ngo-Metzger, MD, MPH.
Grant Support: By the Agency for Healthcare Research and Quality (contract HHSA 290-2007-10057-I, Task Order 13).
Disclosures: Disclosures can be viewed at www.acponline.org/authors/icmje/ConflictOfInterestForms.do?msNum=M14-2221.
Editors' Disclosures: Christine Laine, MD, MPH, Editor in Chief, reports that she has no financial relationships or interests to disclose. Darren B. Taichman, MD, PhD, Executive Deputy Editor, reports that he has no financial relationships or interests to disclose. Cynthia D. Mulrow, MD, MSc, Senior Deputy Editor, reports that she has no relationships or interests to disclose. Deborah Cotton, MD, MPH, Deputy Editor, reports that she has no financial relationships or interest to disclose. Jaya K. Rao, MD, MHS, Deputy Editor, reports that she has stock holdings/options in Eli Lilly and Pfizer. Sankey V. Williams, MD, Deputy Editor, reports that he has no financial relationships or interests to disclose. Catharine B. Stack, PhD, MS, Deputy Editor for Statistics, reports that she has stock holdings in Pfizer.
Requests for Single Reprints: Shelley Selph, MD, MPH, Oregon Health & Science University, 3181 Southwest Sam Jackson Park Road, Mail Code BICC, Portland, OR 97239; e-mail, selphs@ohsu.edu.
Current Author Addresses: Drs. Selph, Patel, and Chou; Ms. Dana; Mr. Blazina; and Ms. Bougatsos: Oregon Health & Science University, 3181 Southwest Sam Jackson Park Road, Mail Code BICC, Portland, OR 97239.
Author Contributions: Conception and design: S. Selph, T. Dana, R. Chou.
Analysis and interpretation of the data: S. Selph, T. Dana, I. Blazina, H. Patel, R. Chou.
Drafting of the article: S. Selph, T. Dana, I. Blazina, H. Patel, R. Chou.
Critical revision of the article for important intellectual content: S. Selph, T. Dana, I. Blazina, R. Chou.
Final approval of the article: S. Selph, T. Dana, I. Blazina, R. Chou.
Statistical expertise: R. Chou.
Obtaining of funding: R. Chou.
Administrative, technical, or logistic support: T. Dana, I. Blazina, C. Bougatsos.
Collection and assembly of data: S. Selph, T. Dana, I. Blazina, C. Bougatsos, H. Patel, R. Chou.

Abstract

Background:

Screening for type 2 diabetes mellitus could lead to earlier identification and treatment of asymptomatic diabetes, impaired fasting glucose (IFG), or impaired glucose tolerance (IGT), potentially resulting in improved outcomes.

Purpose:

To update the 2008 U.S. Preventive Services Task Force review on diabetes screening in adults.

Data Sources:

Cochrane databases and MEDLINE (2007 through October 2014) and relevant studies from previous Task Force reviews.

Study Selection:

Randomized, controlled trials; controlled, observational studies; and systematic reviews.

Data Extraction:

Data were abstracted by 1 investigator and checked by a second; 2 investigators independently assessed study quality.

Data Synthesis:

In 2 trials, screening for diabetes was associated with no 10-year mortality benefit versus no screening (hazard ratio, 1.06 [95% CI, 0.90 to 1.25]). Sixteen trials consistently found that treatment of IFG or IGT was associated with delayed progression to diabetes. Most trials of treatment of IFG or IGT found no effects on all-cause or cardiovascular mortality, although lifestyle modification was associated with decreased risk for both outcomes after 23 years in 1 trial. For screen-detected diabetes, 1 trial found no effect of an intensive multifactorial intervention on risk for all-cause or cardiovascular mortality versus standard control. In diabetes that was not specifically screen-detected, 9 systematic reviews found that intensive glucose control did not reduce risk for all-cause or cardiovascular mortality and results for intensive blood pressure control were inconsistent.

Limitation:

The review was restricted to English-language articles, and few studies were conducted in screen-detected populations.

Conclusion:

Screening for diabetes did not improve mortality rates after 10 years of follow-up. More evidence is needed to determine the effectiveness of treatments for screen-detected diabetes. Treatment of IFG or IGT was associated with delayed progression to diabetes.

Primary Funding Source:

Agency for Healthcare Research and Quality.

In the United States, approximately 21 million persons received diabetes diagnoses in 2010, and an estimated 8 million cases were undiagnosed; roughly 90% to 95% of them have type 2 diabetes mellitus (1, 2). Prevalence of diabetes among U.S. adults has increased, from approximately 5% in 1995 to 8% in 2010 (3). Diabetes is the leading cause of kidney failure, nontraumatic lower-limb amputations, and blindness; a major cause of heart disease and stroke; and the seventh-leading cause of death in the United States (1).

Risk factors for diabetes include obesity, physical inactivity, smoking, and older age (1). Diabetes is more common among certain ethnic and racial minorities (1, 3). Type 2 diabetes is caused by insulin resistance and relative insulin deficiency, resulting in the inability to maintain normoglycemia. Diabetes typically develops slowly (4, 5), although microvascular disease, such as retinopathy and neuropathy, may be present at the time of diagnosis due to vascular damage during the subclinical phase (4, 6).

Screening asymptomatic persons (those without signs or symptoms of hyperglycemia and no clinical sequelae) may lead to earlier identification and earlier or more-intensive treatments, potentially improving health outcomes (2). Strategies for screening include routine screening or targeted screening based on the presence of risk factors, such as obesity or hypertension. In 2008, the U.S. Preventive Services Task Force (USPSTF) recommended diabetes screening in asymptomatic adults with sustained blood pressure (BP) (treated or untreated) greater than 135/80 mm Hg (B recommendation). Although direct evidence on benefits and harms of screening was not available, the recommendation was based on the ability of screening to identify persons with diabetes and evidence that more-intensive BP treatment was associated with reduced risk for cardiovascular events, including cardiovascular mortality, in patients with diabetes and hypertension. The USPSTF found insufficient evidence to assess the balance of benefits and harms of screening in adults without elevated BP (I statement). It also found that lifestyle and drug interventions for impaired fasting glucose (IFG) or impaired glucose tolerance (IGT), defined as a hemoglobin A_1c level of 5.7% to 6.4% or a fasting blood glucose level between 5.55 and 6.94 mmol/L (100 and 125 mg/dL) (2), were associated with reduced risk for progression to diabetes (7–14). Other groups also recommend screening persons with risk factors (15–20).

This article updates previous USPSTF reviews (21–23) on diabetes screening in nonpregnant adults.

Methods

Scope of the Review

We developed a review protocol and analytic framework (Appendix Figure 1) that included the following key questions:

Appendix Figure 1.

Analytic framework.

DM = diabetes mellitus; IFG = impaired fasting glucose; IGT = impaired glucose tolerance; KQ = key question; MI = myocardial infarction.

1. Is there direct evidence that screening for type 2 diabetes, IFG, or IGT among asymptomatic adults improves health outcomes?

2. What are the harms of screening for type 2 diabetes, IFG, or IGT?

3. Do interventions for screen-detected or early diabetes, IFG, or IGT provide an incremental benefit in health outcomes compared with no interventions or initiating interventions after clinical diagnosis?

4. What are the harms of interventions for screen-detected or early diabetes, IFG, or IGT?

5. Is there evidence that more-intensive glucose, BP, or lipid control interventions improve health outcomes in adults with type 2 diabetes, IFG, or IGT compared with traditional control? Is there evidence that aspirin use improves health outcomes in these populations compared with nonuse?

6. What are the harms of more-intensive interventions compared with traditional control in adults with type 2 diabetes, IFG, or IGT?

7. Do interventions for IFG or IGT delay or prevent the progression to type 2 diabetes?

The full report (24), on which this article is based, provides detailed methods and data for the review, including search strategies, evidence tables, and quality ratings of individual studies (available at www.uspreventiveservicestaskforce.org). The full report includes an additional key question on whether the effects of screening or interventions for screen-detected or early diabetes, IFG, or IGT vary by subgroup; effects of treatments on microvascular outcomes; and evidence on effects of more- versus less-intensive lipid control and aspirin use (24).

Data Sources and Searches

A research librarian searched the Cochrane Central Register of Controlled Trials and the Cochrane Database of Systematic Reviews and MEDLINE (2007 to October 2014). We supplemented electronic searches by reviewing previous USPSTF reports and reference lists of relevant articles.

Study Selection

At least 2 reviewers independently evaluated each study to determine inclusion eligibility using predefined inclusion and exclusion criteria (Appendix Figure 2). Because of the limited evidence on treatment of screen-detected diabetes (key question 5), we also included studies of treatment of early diabetes (defined as a pharmacologically untreated hemoglobin A_1c level <8.5% or diabetes diagnosis in the past year) that was not specifically screen-detected. Appendix Figure 3 summarizes the selection of literature.

Appendix Figure 2.

Inclusion and exclusion criteria per KQ.

BP = blood pressure; DM = diabetes mellitus; IFG = impaired fasting glucose; IGT = impaired glucose tolerance; KQ = key question; MI = myocardial infarction.

Appendix Figure 3.

Summary of evidence search and selection.

KQ = key question.

* Cochrane Central Register of Controlled Trials and Cochrane Database of Systematic Reviews.

† Other sources include previous reports, reference lists of relevant articles, and systematic reviews.

‡ An additional 27 publications are included in the full report (23). Some studies have several publications and some are included for more than 1 KQ.

Image: 5ff3_Appendix_Figure 3_Summary_of_evidence_search_and_selection

Data Abstraction and Quality Rating

One investigator abstracted details about the study design, patient population, setting, screening method, interventions, analysis, follow-up, and results. A second investigator reviewed data abstraction for accuracy. Two investigators independently applied criteria developed by the USPSTF (25) to rate the quality of each study as good, fair, or poor. Discrepancies were resolved through a consensus process.

Data Synthesis and Analysis

We conducted meta-analyses to calculate risk ratios (RRs) on effects of interventions with the DerSimonian–Laird random-effects model using Stata, version 12 (StataCorp). Statistical heterogeneity was assessed using the I² statistic (26). When statistical heterogeneity was present, we performed sensitivity analyses using the profile likelihood method because the DerSimonian–Laird model results in overly narrow 95% CIs (27). Two studies (28–30) that used a 2 × 2 factorial design reported no interaction between treatments and were analyzed as a 2-group parallel group trial for the comparison of interest. When studies evaluated several lifestyle strategies, we combined the lifestyle groups. We included all studies in meta-analyses, regardless of event rates. For rare events (incidence <1%), we calculated the Peto odds ratio (31). We stratified results by drug class or lifestyle intervention and performed additional sensitivity analyses based on study quality and presence of outlier trials. We assessed the aggregate internal validity (quality) of the body of evidence for each key question (good, fair, or poor) using methods developed by the USPSTF, based on the quality of studies, precision of estimates, consistency of results, and directness of evidence (25).

Role of the Funding Source

This research was funded by the Agency for Healthcare Research and Quality (AHRQ) under a contract to support the work of the USPSTF. Investigators worked with USPSTF members and AHRQ staff to develop and refine the scope, analytic framework, and key questions; resolve issues arising during the project; and finalize the report. The AHRQ had no role in study selection, quality assessment, synthesis, or development of conclusions. The AHRQ provided project oversight; reviewed the draft report; and distributed the draft for peer review, including to representatives of professional societies and federal agencies. It also performed a final review of the manuscript to ensure that the analysis met methodological standards. The investigators are solely responsible for the content and the decision to submit the manuscript for publication.

Results

Benefits of Screening

Two randomized, controlled trials (ADDITION [Anglo-Danish-Dutch Study of Intensive Treatment in People With Screen Detected Diabetes in Primary Care]–Cambridge [Cambridge, United Kingdom] trial [n = 19 226] [32], rated good-quality, and a trial conducted in Ely, United Kingdom [n = 4936] [33], rated fair-quality) evaluated effects of diabetes screening versus no screening on mortality (Appendix Table 1). The ongoing ADDITION trial includes sites in Cambridge, the Netherlands, and Denmark on intensive versus standard treatment of screen-detected diabetes; however, only the Cambridge site had a no-screening component (34). Mean age ranged from 51 to 58 years, 36% to 54% of participants were women, and follow-up was 10 years in both studies (32, 33). In ADDITION-Cambridge, persons at high risk for diabetes, based on known risk factors, were randomly assigned in clusters by clinic site to screening or no screening (32). The Ely study randomly enrolled participants (not selected based on high risk for diabetes) to screening or no screening from a single practice site (33). Seventy-eight percent of participants (11 737 of 15 089) invited to screening had screening in the ADDITION trial (32); 68% of participants in the Ely study were screened (33). Methodological shortcomings in the Ely study included unclear randomization and allocation concealment methods, with baseline differences between groups.

Appendix Table 1. Effect of Screening for Diabetes on Health Outcomes

Screening was not superior to no screening in reducing risk for all-cause mortality in either the ADDITION (hazard ratio [HR], 1.06 [95% CI, 0.90 to 1.25]) (32) or the Ely (unadjusted HR, 0.96 [CI, 0.77 to 1.20]; adjusted HR, 0.79 [CI, 0.63 to 1.00]) (33) trial, with point estimates close to 1. The ADDITION trial also found that screening was not associated with reduced risk for cardiovascular mortality (HR, 1.02 [CI, 0.75 to 1.38]), cancer-related mortality (HR, 1.08 [CI, 0.90 to 1.30]), or diabetes-related mortality (HR, 1.26 [CI, 0.75 to 2.10]) (32). Neither study reported nonmortality health outcomes.

Harms of Screening

A fair-quality pilot study of 116 persons invited for screening in the ADDITION trial found that a new diagnosis of diabetes was associated with increased short-term anxiety 6 weeks after screening, compared with no new diagnosis, based on short-form Spielberger State-Trait Anxiety Inventory scores (46.7 vs. 37.0; P = 0.031) (35). Studies lasting longer than the ADDITION pilot study (≥1 year) found no negative psychological effects associated with invitation to screening or notification of positive diabetes status (36, 37). We identified no studies estimating the rate of false-positive results, psychological effects, or other harms associated with a diagnosis of IFG or IGT.

Benefits of Treating Screen-Detected or Early Diabetes, IFG, or IGT

A randomized trial conducted in Da Qing, China, of overweight (mean body mass index [BMI], 25.8 kg/m²) persons with IGT found that, compared with usual care, a 6-year lifestyle intervention was associated with reduced risk for all-cause (HR, 0.71 [CI, 0.51 to 0.99]) and cardiovascular (HR, 0.59 [CI, 0.36 to 0.96]) mortality after 23 years of follow-up (38). The trial was rated fair-quality because of unclear randomization and allocation concealment methods. This study had previously reported no difference in these outcomes after 20-year follow-up (39). Other trials of lifestyle interventions in persons with IFG or IGT and elevated BMI (40, 41) or newly diagnosed diabetes (42–44) with shorter follow-up also reported no beneficial effects on all-cause or cardiovascular mortality (Appendix Table 2).

Appendix Table 2. Health Outcomes in Studies of Interventions for Screen-Detected/Early DM, IFG, or IGT

Image: 5tt3_Appendix_Table_2_Health_Outcomes_in_Studies_of_Interventions_for_Screen-Detected_Early_DM_I

Appendix Table 2. Health Outcomes in Studies of Interventions for Screen-Detected/Early DM, IFG, or IGT

Trials of pharmacologic interventions (alone [28–30, 45–49] or in combination with lifestyle modification [50] vs. placebo or usual care) for early diabetes, IFG, or IGT found few differences in health outcomes, including all-cause and cardiovascular mortality (Appendix Table 2). Mean age ranged from 45 to 64 years, and studies enrolled persons who were overweight (BMI >25.0 kg/m²) or obese (BMI >30.0 kg/m²). Five studies were rated good-quality and 3 were rated fair-quality; common methodological shortcomings in the fair-quality studies included unclear randomization and allocation concealment methods. Although individual studies were generally underpowered to detect these outcomes and few events were reported in most studies, pooled estimates were close to 1. Based on 8 studies (10, 28, 45–48, 51, 52) of glucose-lowering agents, including 3 (10, 51, 52) from the previous USPSTF review (22), the pooled odds ratio for all-cause mortality was 1.01 (CI, 0.87 to 1.18; I² = 28%) (Appendix Figure 4). For cardiovascular mortality, the pooled odds ratio was 1.06 (CI, 0.84 to 1.35; I² = 7%) based on 5 studies (28, 48, 52–54) of glucose-lowering agents, including 3 (52–54) from the previous USPSTF review (22) (Appendix Figure 5).

Appendix Figure 4.

Meta-analysis of the effect of pharmacologic interventions on all-cause mortality.

M-H = Mantel–Haenszel fixed-effects model; OR = odds ratio.

* Included in the 2008 report (22).

Image: 5ff4_Appendix_Figure 4_Meta-analysis_of_the_effect_of_pharmacologic_interventions_on_all-cause_m

Appendix Figure 5.

Meta-analysis of the effect of pharmacologic interventions on cardiovascular mortality.

M-H = Mantel–Haenszel fixed-effects model; OR = odds ratio.

* Included in the 2008 report (22).

Image: 5ff5_Appendix_Figure 5_Meta-analysis_of_the_effect_of_pharmacologic_interventions_on_cardiovascu

Harms of Treating Screen-Detected or Early Diabetes, IFG, or IGT

Of 4 good-quality and 5 fair-quality trials that reported harms associated with interventions (28–30, 40, 43–49), 1 study was conducted in persons with screen-detected or early diabetes and the others enrolled persons with IFG or IGT. No study was specifically designed to assess harms. There were few differences between medications or lifestyle modification versus placebo or usual care in risk for harms (Appendix Table 2). One trial found that, compared with placebo, acarbose was associated with greater risk for withdrawal because of adverse events (47). Rosiglitazone was associated with increased congestive heart failure in 1 trial, although the estimate was imprecise (HR, 7.04 [CI, 1.60 to 31]) (30). One study found that nateglinide was associated with increased risk for hypoglycemia versus placebo (RR, 1.73 [CI, 1.57 to 1.92]), and valsartan was associated with increased risk for hypotension-related adverse events (RR, 1.16 [CI, 1.11 to 1.23]) (28, 29).

Benefits of More-Intensive Treatment Versus Standard Treatment

The treatment phase of the ADDITION-Europe trial evaluated effects of more-intensive multifactorial treatment of screen-detected diabetes (55–57). It was rated fair-quality because of unclear methods of randomization and allocation concealment. The mean hemoglobin A_1c level was 6.5%, approximately one fourth of participants were smokers, mean BMI was 31.5 kg/m², and 6% to 7% of participants had a previous myocardial infarction (MI). Participants were randomly assigned to a multifactorial intervention that included use of intensive glucose-, BP-, and lipid-lowering targets (hemoglobin A_1c level <7.0%, BP <135/85 mm Hg, and total cholesterol level ≤4.5 to 5.0 mmol/L [≤173.7 to 193.1 mg/dL]) plus a lifestyle education component (n = 1678) versus treatment to standard targets according to local guidelines (n = 1379). Participants were followed for 5 years or until their first cardiovascular event (cardiovascular mortality, nonfatal MI or stroke, revascularization, or [nontraumatic] amputation) (55).

After adjustment for country, intensive treatment was not associated with reduced risk for a first cardiovascular event (HR, 0.83 [CI, 0.65 to 1.05]) (55), all-cause (HR, 0.83 [CI, 0.65 to 1.05]) or cardiovascular (HR, 0.88 [CI, 0.51 to 1.51]) mortality, stroke (HR, 0.98 [CI, 0.57 to 1.71]), MI (HR, 0.70 [CI, 0.41 to 1.21]), or revascularization (HR, 0.79 [CI, 0.52 to 1.18]), although most estimates favored intensive therapy. Mortality and cardiovascular event rates were lower than anticipated, with little difference between groups in final hemoglobin A_1c and total cholesterol levels and BP (55). There was also no difference in self-reported measures of general and diabetes-specific quality of life (57).

In persons with diabetes that was not specifically screen-detected, 9 good-quality systematic reviews found consistent evidence that intensive glucose-lowering treatment to a target hemoglobin A_1c level less than 6.0% to 7.5% was not associated with decreased risk for all-cause or cardiovascular mortality compared with less-intensive therapy (Appendix Table 3) (58–66). One of the largest and most recent reviews (60) analyzed evidence from 14 trials (n = 28 614), including several large, good-quality trials (67–69) published since the previous USPSTF report. Intensive glucose-lowering therapy was consistently associated with reduced risk for nonfatal MI in 6 reviews (RR range, 0.83 to 0.87) (58, 60, 61, 63, 64, 66).

Appendix Table 3. Good-Quality Systematic Reviews of Intensive Versus Standard Glucose Control in People With DM Reporting Health Outcomes and Harms

Image: 5tt4_Appendix_Table_3_Good-Quality_Systematic_Reviews_of_Intensive_Versus_Standard_Glucose_Contr

Appendix Table 3. Good-Quality Systematic Reviews of Intensive Versus Standard Glucose Control in People With DM Reporting Health Outcomes and Harms

Intensive BP-lowering therapy was associated with reduced risk for all-cause mortality (RR, 0.90 [CI, 0.82 to 0.98]; I² = 0%) and stroke (RR, 0.83 [CI, 0.73 to 0.95]; I² = 27%) in 1 good-quality systematic review (70), but individual trials defined intensive BP control differently and some trials showed inconsistent effects (Appendix Table 4). One recent large trial (n = 4732) (71) found no difference between a systolic BP target of 140 mm Hg and 120 mm Hg in risk for all-cause (RR, 1.11 [CI, 0.89 to 1.38]) or cardiovascular (RR, 1.04 [CI, 0.73 to 1.48]) mortality, whereas another trial (n = 11 140) (72, 73) found that, compared with placebo, the addition of an angiotensin-converting enzyme inhibitor plus a diuretic was associated with decreased risk for all-cause (RR, 0.87 [CI, 0.76 to 0.98]) and cardiovascular (RR, 0.33 [CI, 0.15 to 0.74]) mortality. Results from older studies (22) were also mixed and were characterized by variability in antihypertensive treatments and baseline, target, and achieved BP levels (74–79).

Appendix Table 4. Trials of Variably Defined Intensive Versus Standard BP Control in People With DM

Image: 5tt5_Appendix_Table_4_Trials_of_Variably_Defined_Intensive_Versus_Standard_BP_Control_in_People

Appendix Table 4. Trials of Variably Defined Intensive Versus Standard BP Control in People With DM

Harms of More-Intensive Treatment Versus Standard Treatment

The ADDITION-Netherlands study found no difference between intensive multifactorial treatment versus standard treatment in risk for severe hypoglycemia after 1 year of follow-up, but the event rate was low and the estimate was imprecise (0.4% vs. 0.0%; RR, 2.86 [CI, 0.12 to 70]) (80).

In persons with diabetes not specifically screen-detected, intensive glucose control was associated with increased risk for severe hypoglycemia and serious nonhypoglycemia adverse events requiring medical intervention (Appendix Table 3) (59, 60, 63, 65). Harms of other interventions, including intensive BP-lowering and intensive multifactorial interventions, were mixed (71, 72, 81, 82).

Benefits of Treatment in IFG or IGT on the Delay or Prevention of Progression to Diabetes

We identified 14 randomized, controlled trials (28, 29, 38–40, 45–47, 49, 83–89), 1 quasi-randomized trial (48), and 1 cohort study (90) on the effects of interventions for IFG or IGT on risk for progression to diabetes (Appendix Table 5) (28, 29, 38–40, 45–49, 83–90). Three trials were rated good-quality (28, 29, 46, 49), and the remainder were fair-quality. Methodological shortcomings in the fair-quality studies included unclear randomization and allocation concealment methods, unblinded design, and lack of intention-to-treat analysis. The studies assessed lifestyle interventions (6 studies) (38, 40, 84, 86–88), pharmacologic interventions (8 studies in 9 publications) (28, 29, 45–49, 89, 90), and multifactorial interventions (2 studies) (83, 85). Treatment duration ranged from 6 months to 6 years, with follow-up extending up to 23 years. Mean age ranged from 45 to 65 years. In all but 1 study (86), participants were overweight or obese. Mean total cholesterol levels ranged from 4.3 to 5.9 mmol/L (166 to 228 mg/dL) (Appendix Table 5).

Appendix Table 5. Studies of Interventions to Prevent or Delay Progression to DM

Lifestyle Interventions

Lifestyle interventions were associated with decreased risk for progression to diabetes, based on 6 studies (38, 40, 84, 86–88), including 4 (7–10) that were in the previous USPSTF review (22) (pooled RR, 0.55 [CI, 0.43 to 0.70]; I² = 77%; profile likelihood estimate, 0.57 [CI, 0.43 to 0.70]) (Appendix Figure 6). After exclusion of the Da Qing trial, an outlier study with very long (23-year) follow-up (38), we found similar results (pooled RR, 0.53 [CI, 0.44 to 0.63]; I² = 25%).

Appendix Figure 6.

Meta-analysis of the effect of lifestyle interventions on incidence of progression to DM.

DM = diabetes mellitus; D+L = DerSimonian–Laird random-effects model; PL = profile likelihood model.

* Included in the 2008 report (22).

Image: 5ff6_Appendix_Figure 6_Meta-analysis_of_the_effect_of_lifestyle_interventions_on_incidence_of_pr

Pharmacologic Interventions

Eight studies published since the previous USPSTF review assessed the effect of pharmacologic interventions (28, 45–49, 89, 90). Thiazolinediones were associated with decreased risk for progression to diabetes (3 studies; pooled RR, 0.50 [CI, 0.28 to 0.92]; I² = 92%) (Appendix Figure 7) (45, 48, 52). Statistical heterogeneity was substantial, and the estimate was no longer statistically significant using the profile likelihood method (RR, 0.51 [CI, 0.23 to 1.06]). Excluding the Indian Diabetes Prevential Programme-2 trial (48), which was conducted in India among mostly male participants, eliminated much of the heterogeneity (RR, 0.42 [CI, 0.37 to 0.47]; I² = 36%). A similar effect was found in 4 studies of α-glucosidase inhibitors (RR, 0.64 [CI, 0.45 to 0.90]; I² = 67%; profile likelihood method, 0.65 [CI, 0.44 to 0.91]) (Appendix Figure 8) (46, 47, 51, 91). Other studies found that valsartan (29) and a combination of low-dose metformin and rosiglitazone (49), but not nateglinide (28) or glimepiride (89), was associated with reduced risk for progression to diabetes.

Appendix Figure 7.

Meta-analysis of the effect of thiazolidinediones on incidence of progression to DM.

DM = diabetes mellitus; D+L = DerSimonian–Laird random-effects model; PL = profile likelihood model.

* Included in the 2008 report (22).

Image: 5ff7_Appendix_Figure 7_Meta-analysis_of_the_effect_of_thiazolidinediones_on_incidence_of_progres

Appendix Figure 8.

Meta-analysis of the effect of α-glucosidase inhibitors on incidence of progression to DM.

DM = diabetes mellitus; D+L = DerSimonian–Laird random-effects model; PL = profile likelihood model.

* Included in the 2008 report (22).

† Included in the 2003 report (21).

Image: 5ff8_Appendix_Figure 8_Meta-analysis_of_the_effect_of_-glucosidase_inhibitors_on_incidence_of_pr

Multifactorial Interventions

Two trials examined the multifactorial interventions consisting of intensive glucose, BP, and lipid control, in addition to lifestyle counseling and aspirin (83, 85). The ADDITION-Denmark trial (n = 1510) found that the multifactorial intervention was associated with a decreased risk for progression to diabetes that was nearly statistically significant (RR, 0.89 [CI, 0.78 to 1.02]) (85). Effects were greater in the subgroup that also received motivational interviewing (RR, 0.83 [CI, 0.68 to 1.00]) than in the subgroup that did not (RR, 0.95 [CI, 0.80 to 1.14]). A smaller Chinese study (n = 181) reported a lower incidence of progression to diabetes in the intervention group than in the control group, but the estimate was imprecise (0.0% vs. 5.8%; RR, 0.08 [CI, 0.00 to 1.42]) (83).

Discussion

The Table summarizes the evidence reviewed for this update. In 2 trials, 1 of which focused on persons at greater risk for diabetes, screening was not associated with decreased risk for mortality versus no screening after 10 years of follow-up (32, 33). Point estimates from both trials were close to 1 and did not indicate a trend toward benefit in the good-quality trial, although the CIs encompass potentially meaningful effects (for example, 10% and 37% reduction in risk for all-cause mortality). Possible explanations for the lack of a mortality effect include limited screening uptake, increased mortality among nonattendees invited to screening (potentially attenuating estimates based on intention-to-treat analyses), increased diabetes screening across groups outside of the study protocol, improved management of cardiovascular disease risk factors and diabetes contributing to decreased mortality, and inadequate length of follow-up to adequately assess mortality. In addition, screening trials did not report nonmortality clinical outcomes, which may require less lengthy follow-up to detect clinically relevant effects. Evidence on harms associated with screening is sparse, although limited evidence showed no clear long-term negative effects on psychological measures (35–37).

Table. Summary of Evidence

Lifestyle and pharmacologic interventions both seem to be effective in delaying or preventing progression from IFG or IGT to diabetes in persons with high BMI (7–10, 39, 40, 45–47, 51, 52, 84, 86, 88, 89, 91). Effects of interventions on long-term clinical outcomes are less clear. The study with the longest follow-up (23 years) found that lifestyle modification for 6 years for early diabetes, IFG, or IGT was associated with a mortality benefit (38). Studies with shorter duration of follow-up found no beneficial effects of treatment on mortality, although evidence for improvement in microvascular outcomes was limited, as discussed in more detail in the full report (24).

Pharmacologic treatment of screen-detected or early diabetes, IFG, or IGT was associated with increased risk for withdrawal because of adverse events versus placebo in 1 study (47), with no clear increased risk for serious adverse events. In general, trials were not designed or powered to specifically assess the risk for serious but uncommon or rare adverse events, although studies not restricted to persons with screen-detected or early diabetes did not show a clear increase in risk for such events, such as lactic acidosis with metformin (92).

Since the previous USPSTF review, there is now evidence from a large, good-quality trial that an intensive multifactorial intervention for screen-detected diabetes aimed at decreasing glucose and lipid levels and BP was not associated with a statistically significant reduction in risk for all-cause or cardiovascular mortality or morbidity versus standard treatment, although estimates favored intensive treatment (56). For diabetes not specifically identified by screening, systematic reviews consistently found no association between intensive versus less-intensive glucose-lowering therapy and reduced risk for all-cause or cardiovascular mortality (58–66). Intensive glucose-lowering therapy was associated with reduced risk for nonfatal MI but increased risk for severe hypoglycemia. Other outcomes, such as retinopathy and neuropathy (discussed in the full report [24]), were found less frequently in these reviews, and pooled risk estimates were inconsistent, precluding reliable conclusions.

The 2008 USPSTF review (22) found that effects of intensive BP control were greater in persons with diabetes versus those without it, based on subgroup analyses from trials that were generally less successful at achieving lower BP than recent studies (71, 72). Since then, there is more evidence on the benefits of more effective, intensive BP control versus standard therapy, specifically in persons with diabetes. Although a good-quality systematic review found that intensive BP control in persons with diabetes was associated with reduced risk for all-cause mortality versus less-intensive BP control (70), results from individual studies, including those from the recent, large, well-conducted trials (71, 72), were inconsistent.

Our review has limitations. We only included English-language articles, although a recent review found that this limitation did not introduce bias into systematic review findings (93). We identified only 2 screening studies, and only 1 treatment study was conducted in a screen-detected population. We included evidence on intensive treatment from studies of persons with early diabetes that was not specifically screen-detected because studies in screen-detected populations were lacking, which could limit applicability to screening settings.

We identified many important research gaps. Screening studies in U.S. populations, in which the prevalence of undiagnosed diabetes (and IFG or IGT) is likely to be greater than the 3% identified in the ADDITION-Cambridge and Ely studies, would be more applicable for informing U.S. screening decisions. As detailed in the full report, there is also little evidence on the effect of screening on ethnic and racial minorities, in whom the prevalence of diabetes is greater than in persons of white, European ancestry (24). Longer-term follow-up of the treatment phase of the ADDITION trial is needed to determine whether beneficial trends become statistically significant as more events occur (56). Studies of the effect of interventions for early diabetes, IFG, or IGT, particularly studies of lifestyle interventions with long-term (>20 years) follow-up, are needed to confirm the findings of the Da Qing study (38).

In conclusion, screening for diabetes did not improve mortality rates after 10 years of follow-up in 2 trials (32, 33) but was found to decrease mortality rates in a lifestyle intervention study with 23 years of follow-up (38). More evidence is needed to determine the effectiveness of treatments for screen-detected diabetes. Treatment of IFG or IGT was associated with delayed progression to diabetes.