Reporting

• How do we present data in a way that is more helpful for patients/clinicians/regulators/industry to make more informed decisions?– Presentations that are interpretable for non-researchers

• Provide absolute risk AND relative risk summaries• Provide reference or “baseline” risk to aid decision-makers

– Provide risk for alternative therapy or without therapy• Provide summaries for each important outcome (benefits AND harms) • Group information by outcome severity• Visual summaries • Measures of variation• Heterogeneity of effects

    Expected Number in 100 Treated Patients

 

    A B NNT

Responders        

  >30% improvement

XX XX XX

  >50% improvement

XX XX XX

         

Adverse Reactions

       

  Life-threatening XX XX XX

  Severe XX XX XX

  Other XX XX XX

0 1

Confidence Intervals for:

Risk Difference Relative Risk

Proportion of responders0 100

Response Magnitude

Big Winner

Loser

A

B

50

= Boxplot

% Pain Change

0

100

Difference in proportions

NNT

Relative Risk

0

1

0 50

Cumulative Distribution Functions

Plot risk difference, RR, and NNT (w/ confidence bands) as a function of % pain change (on a common horizontal axis)

Scatterplot Approach w/ Distance Metric

• Assume benefit and risk each can be measured on a continuous (e.g., 0-10) scale or can be transformed as such

• Create a scatterplot (benefit vs. risk) based on patient results

• Fit a smooth “tolerability boundary” through 2 or more points– All points on the boundary have equivalent benefit:risk

• Reduce to one-dimensional analyses by defining a distance function– Use the tolerability boundary to standardize the distance– Closer to (risk =0, benefit=10) → Benefit:risk ↑

Patient Level Measures

• Patients rate their overall experience with respect to perceived benefit and risk

• Possibly useful for therapies to treat symptoms (e.g., pain) when “risks” are recognizable to the participant

• Problematic when symptoms do not equate with risk– i.e., silent risk of abnormal labs (e.g., LFTs, bilirubin)

Example: Scatterplot Approach: ACTG A5252

A5252: Study evaluating therapies for neuropathic pain

Benefit: Pain(measured 0-10)

Risk: “how bothersome were the side effects?” (0-10)

Define tolerability boundary based on 3 points:

• Minimum tolerable benefit when risk = 0 (b1)

• Maximum tolerable risk when benefit = 10 (r1)

• Minimum tolerable benefit when risk = r1/2 (b2)

Partition plane into regions of interest and summarize the proportion that fall into these regions

LOSER

WINNER

?

? May be acceptable with a very serious disease with no known cure

? May not be acceptable with a disease that is not life-threatening and other effective and safe treatment options are available

BE

NE

FIT

RISK

Group #Event/#Subj. Prop.

A 1/500 0.2%

B 3/500 0.6%

Measure Point est & 95%CI

Risk Diff.(B-A) 0.4% [-0.4%, 1.2%]

Relative Risk (B/A) 3.0 [0.3, 158]

Group #Event/#Subj. Prop.

A 1/5,000 0.02%

B 3/5,000 0.06%

Measure Point est & 95%CI

Risk Diff.(B-A) 0.04% [-0.04%, 0.12%]

Relative Risk (B/A) 3.0 [0.3, 158]

Absolute vs. Relative Risk

Implications for safety NI trials and rare event trials

Do both as interpretation depends on both.

Excellent discussion:Wei et.al. Food and Drug Law: Regulation and Education Update. January/February 2011