Rules

Statistical significance should not be used to imply unproven clinical significance

Average results should disclose the range of variation

Retrospective statistical mining to find unsupported findings may mislead

Don't use retrospective stats to imply rigor for unsuited study designs

Tables and graphs must not distort study relationships or trends

No falsely presenting reports/opinions as authentic or authoritative

Communications should not overstate study findings

Do not imply statistical rigor (e.g., via p-values) that the analysis lacks

Disclosing limitations does not cure a claim based on an inadequate study

Response should include complete reprints and literature, not just firm summaries

Keep unsolicited-request response material scientific in tone, not promotional

Do not bundle the off-label response with promotional material

Public responses should not be promotional in nature or tone

Do not imply the source data represents broader experience than it does

A deceptive net impression can be misleading even when individual claims are not

Use risk language comprehensible to the same audience as benefit language

Use headlines/signals for risk information whenever benefit information is signaled

Risk information should be preceded by a signal indicating risk information follows

Match specificity of risk statements to benefit statements

Risk wording should not minimize the severity of a risk event

Risk ordering should not falsely suggest a risk applies only to certain populations or conditions

Material study limitations should be disclosed

A risk-information presentation should not open with unrelated content