Table 2:

Summary of evidence of benefits associated with screening for type 2 diabetes*

No. of studiesQuality assessmentSummary of findingsGRADE quality of evidenceImportance
DeathsEffect
DesignLimitationsInconsistencyIndirectnessImprecisionther considerationsScreening, no. (%)Control, no. (%)Relative (95% CI)Absolute (95% CI)
verall mortality
1 cluster RCT20*
2001–2011Randomized trial§No serious risk of biasNo serious inconsistency**No serious indirectness††No serious imprecision‡‡None§§n = 1532 (9.5)¶¶n = 377 (9.1)***HR 1.06 (0.90 to 1.25)5196 more per million (from 8726 fewer to 21454 more)HighCritical
1 cohort study23*
1990–1992 cohortbservational studyNo serious limitations†††No serious inconsistency**No serious indirectnessNo serious imprecisionNone**n = 1705
116 (6.8)‡‡‡
n = 3231
229 (7.1)§§§
HR 0.79 (0.63 to 1)¶¶¶14 455 fewer per million (from 25 619 fewer to 0 more)LowCritical
2000–2003 cohortbservational studyNo serious limitations†††No serious inconsistency**No serious indirectnessNo serious imprecisionNone**n = 1577
165 (10.5)
n = 1425
126 (8.8)
HR 1.18 (0.93 to 1.51)****15 065 more per million (from 5 927 fewer to 42 039 more)LowCritical
Cardiovascular mortality
1 cluster RCT20*
2001–2011Randomized trial§No serious risk of biasNo serious inconsistency**No serious indirectness††No serious imprecision‡‡None§§n = 482 (3%)¶¶n = 124 (3%)***HR 1.02 (0.75 to 1.38)590 more per million (from 7408 fewer to 11153 more)HighCritical
  • Note: CI = confidence interval, GRADE = Grading of Recommendations Assessment, Development and Evaluation, HR = hazard ratio, RCT = randomized controlled trial.

  • * ur systematic review of benefits associated with screening for type 2 diabetes in adults of any age identified 1 cluster RCT and 1 cohort study, which looked at the effect of screening for type 2 diabetes and related cardiovascular risk factors on overall mortality; the RCT also looked at the effect on cardiovascular mortality. Evidence from modelling studies is available in Appendices 4 and 5.

  • Follow-up was from November 2001 to November 2011 (median 9.6 years, interquartile range [IQR] 8.9–9.9 years; 184 057 person-years).

  • First cohort: follow-up was from 1991 to 1999 (median 10 years; 47 854 person-years of risk). Second cohort: follow-up was from 2000 to 2008 (median 8.1 years; 23 144 person-years of risk).

  • § Population-based cluster RCT. Study reported data from 32 general practices in eastern England randomized to 1 of 3 groups (screening plus intensive treatment for diagnosed diabetes [n = 14]; screening plus routine care for diabetes patients [n = 13]; no-screen control [n = 5]). Study population included 20 184 individuals 40–69 years of age (median 59 years IQR 53–65 years) at high risk of prevalent undiagnosed diabetes on the basis of previously validated risk score (minimum score of 0.17 – reflects top 25% of risk distribution in participating practices).

  • Cochrane Risk of Bias Tool was used to examine this study. The appraisal process was completed by 2 independent reviewers who agreed there was uncertainty regarding allocation concealment and that it was not possible to blind patients and their physicians to their screening status; however, this potential performance bias was unlikely to affect the outcome of interest (mortality). All other domains of bias covered in the Cochrane tool were determined to have a low risk of bias. On the basis of the overall assessment, the evidence was not downgraded for any serious concerns regarding study limitations.

  • ** Single study.

  • †† Study sample characteristics, risk-assessment variables and screening test were similar to the Canadian population and screening context of interest for this review.

  • ‡‡ Large sample and large event rate with narrow confidence interval around estimate of effect.

  • §§ Single study; literature search indicated no other RCTs have been conducted or published for this particular comparison and outcome.

  • ¶¶ Unadjusted prevalence of diabetes in screening practices was 3.0% (standard deviation [SD] 1.0%). Characteristics of eligible participants in screening practices at baseline: mean age 58.2 (SD 7.7) years; 63.9% men (n = 10 260); mean body mass index (BMI) 30.5 (SD 4.6), median diabetes risk score 0.35 (IQR 0.24–0.52); 45.9% prescribed antihypertensive medication (n = 7372); 5.4% prescribed steroids (n = 866).

  • *** Unadjusted prevalence of diabetes in control practices was 3.3% (SD 0.8%). Characteristics of eligible participants in control practices at baseline: mean age 57.9 (SD 7.8) years; 63.9% men (n = 2641); mean BMI 30.6 (SD 4.6), median diabetes risk score 0.34 (IQR 0.24–0.51); 44.8% prescribed antihypertensive medication (n = 1853); 3.7% prescribed steroids (n = 154).

  • ††† The authors reported potential selection bias: “Despite random selection of participants into invitation groups, participants who were offered screening were older at baseline, lived in more deprived areas and included a smaller proportion of men.” However, we did not downgrade this criterion, because in the analysis, the authors adjusted for age, sex and deprivation.

  • ‡‡‡ 52 (45%) of the deaths were recorded as cancer-related, 41 (35%) were due to cardiovascular causes, and 23 (20%) were coded as “other.”

  • §§§ 107 (47%) of the deaths were cancer related, 74 (32%) were due to cardiovascular causes, and 48 (21%) were coded as “other.”

  • ¶¶¶ p = 0.05; adjusted for age, sex and deprivation. For 22 (6%) of those who died (1991–1999), diabetes was included as the underlying cause on the death certificate.

  • **** p = 0.05; adjusted for age, sex and deprivation. For 22 (8%) of those who died (2000–2008), diabetes was included as the underlying cause on the death certificate.