Evidence Tables and Graphs

A comment about evidence-based medicine

I, too, believe in evidence-based medicine - that is, when there is evidence, which seems to apply to a rather small minority of the patients that I see. It's easy to treat a patient who meets all the inclusion and exclusion criteria for a huge study which had clear, significant results. It's much harder to treat all of the other patients. Consider the statement by David Sackett (J National Cancer Institute 1997;89:1003): "When a patient comes to me with symptomatic incurable cancer, my pledge to them is restricted to doing my best . . . when this "Level 1" evidence is lacking I resort to nonexperimental evidence gleaned with study architectures at higher risk of error; my sick patient cannot wait for better evidence."

What to do when the evidence is incomplete and the experts cannot agree

(Adapted from editorial by Harold Sox, Screening mammography in women younger than 50 years of age, Annals of Internal Medicine, 1995, 122:550.)
  1. Review and summarize available studies.
  2. Biological thinking may help. Is it physiologically plausible?
  3. Be sure that current thinking is based on valid evidence. Trust differences in subgroup results only when the intervention works unambiguously in one and fails utterly in another.
  4. Costs do matter.
  5. Primum No Nocere. "Many believe that this principle has particular force when applied to healthy persons. . . when we are in doubt we should take special care to avoid actions that might cause harm."
  6. Talk to the patient
  7. Plan for the usual, adapt for the unusual. Algorithms are applied to usual patients and modified for unusual patients. Patient care decisions should be made on an individual basis.

This web site aims to provide you with evidence where it exists, and with principals of physiology which will help you extend the evidence for those sick patients who cannot wait.

From the poem Doctors
by Anne Sexton

But all along the doctors remember:
First do no harm.
They would kiss if it would heal.
It would not heal.

EVIDENCE TABLES

Here are a collection of evidence tables. Some are rough drafts that will, hopefully, be continually updated. They are Excel text files, and also in GIF format.

GRAPHS

Reproducibility of total hip bone density in 200 women, showing difference "caused" by walking around a room
Reproducibility in 200 women and 100 men, showing cummulative frequency of difference in total hip BMD seen after repositioning.
ALENDRONATE: Summary of results of the Fracture Intervention Trial. The difference in reduction of clinical fractures was not significant in women who did not have a vertebral fracture at the baseline visit. Notice how the incidence of new vertebral fractures depends on presence of baseline fracture.
RALOXIFENE: Preliminary results of MORE study, after 2 years, from data presented at the ASBMR in 12/98.
CALCIUM: controlled trials that reported bone density in each group. The results of the controls are plotted vs the results of the treated subjects, so that the points that fall ABOVE the diagonal line represent studies in which the treatment group was BETTER than the control group. The size of the point is proportional to the nubmer of subjects in the study.
ESTROGEN: controlled trials that reported bone density in each group. The results of the controls are plotted vs the results of the treated subjects, so that the points that fall ABOVE the diagonal line represent studies in which the treatment group was BETTER than the control group. The size of the point is proportional to the nubmer of subjects in the study.
EXERCISE: controlled trials that reported bone density in each group. The results of the controls are plotted vs the results of the treated subjects, so that the points that fall ABOVE the diagonal line represent studies in which the treatment group was BETTER than the control group. The size of the point is proportional to the nubmer of subjects in the study.
CALCITRIOL: controlled trials that reported bone density in each group. The results of the controls are plotted vs the results of the treated subjects, so that the points that fall ABOVE the diagonal line represent studies in which the treatment group was BETTER than the control group. The size of the point is proportional to the nubmer of subjects in the study.
Home page