i n f o p o e m s
These InfoPOEMs® are selected by JNMA InfoPOEMS Editor Gregory E. Gilbert, MSPH, (Gregory.E.Gilbert@
gmail.com) and Associate InfoPOEMS Editor Amy H. Wahlquist, MS, from www.infopoems.com. InfoPOEMs®
are created by experts who continuously survey medical journals worldwide. They identify and summa-
rize valid and clinically applicable new evidence. For more information or to subscribe to e-mail alerts of
InfoPOEMs®, please visit www.infopoems.com.
STUDY LEVELS OF EVIDENCE (LOE) events, and the reporting of results
From the Centre for Evidence-Based Tight Glycemic was sufficient to calculate individ-
Medicine, Oxford. For the most up-to- Control May Decrease ual event rates. Please note that
date levels of evidence, see www.cebm.
net/levels_of_evidence.asp) Cardiovascular End they excluded 6 large trials because
Points in Patients With cardiovascular events were not part
Therapy/Prevention/Etiology/Harm: Type 2 Diabetes Mellitus of the primary study outcomes or
1a: Systematic reviews of randomized because the original papers did not
controlled trials Clinical Question report individual event rates. Since
1b: Individual randomized controlled Does tight glycemic control there were only 5 studies included
trials
1c: All or none randomized controlled
decrease the rate of cardiovascular (33 000 patients), this is a big con-
trials complications in patients with type cern. Some analysts would have
2a: Systematic reviews of cohort studies 2 diabetes mellitus? contacted authors to obtain missing
2b: Individual cohort study or low- data. Additionally, if cardiovascular
quality randomized controlled
Bottom Line event rates are measured in a com-
2c: “Outcomes” research, ecological
studies If the authors of this systematic parable manner, it shouldn’t matter
review have captured all the rele- if these were primary or secondary
Diagnosis: vant studies, tight glycemic control end points. In the studies the authors
1a: Systematic review of level-1 decreases the rate of cardiovascular chose to include, the overall rate of
diagnostic studies complications in patients with type nonfatal myocardial infarctions was
1b: Independent blind comparison
of an appropriate spectrum
2 diabetes mellitus. There is strong 10 per 1000 patient-years in inten-
of consecutive patients, all of reason to believe, however, that by sively treated patients compared
whom have undergone both the excluding studies in which cardio- with 12.3 in control patients. The
diagnostic test and the reference
standard, or a clinical decision rule
vascular outcomes were secondary, rate of coronary events was 14.3
not validated on a second set of and by relying solely on the pub- and 17.2, respectively, per 1000
patients lished outcomes (and not contact- patient-years. The rate of strokes
1c: Absolute SpPins and SnNouts ing authors for additional data), the (approximately 7 per 1000 patient-
2a: Systematic review of level >2 authors had incomplete data from years) and the total death rate were
2b: Independent blind or objective
comparison, study confined
which to draw their conclusions. comparable regardless of treatment
to a narrow spectrum of study (LOE = 1a-) (approximately 18 deaths per 1000
individuals, or a diagnostic clinical patient-years). The average differ-
rule not validated in a test set
Study Design ence between glycohemoglobin lev-
Prognosis:
Meta-analysis (randomized con- els in patients treated with inten-
1a: Systematic review of inception trolled trials) sive care and with usual care was
cohort studies approximately 1%. In these studies,
1b: Individual inception cohort study Funding blood pressure control and control
with >80% foll