| 1. Recomendations can be made only if there is direct scientific evidence of improved health outcomes (see no. 2), not because a panel member believes there is benefit nor because it is accepted practice in CML care. When such evidence is lacking, the results of the analysis should state: “There is insufficient evidence to make a recommendation.” |
| 2. The Analysis will not result in recommendation for one intervention over another unless there is evidence from a controlled study (internal controls) or from dramatic findings in an uncontrolled study that patients treated with that intervention experience better outcomes (eg, higher survival) than those treated by the alternative. The outcomes that matter most are those that patients experience (eg, lengthened survival), not intermediate outcomes for which the linkage to health outcomes is less certain (eg, cytogenetic remission). |
| 3. When extrapolations of evidence are made from one patient population to another to infer effectiveness, the Analysis should make this explicit and discuss the implications. |
| 4. Claims of proofshould be accompanied by full disclosure of the limitations of the evidence. |
| 5. Claims about benefit should clarify the magnitude of benefit, preferably in absolute terms rather than as relative benefit. These claims should be accompanied by a description of potential harms, preferably by estimating the probability of these harms. Confidence intervals should be used to clarify the range of uncertainty about the estimates. |
| 6. When there are complex tradeoffs between benefits and harms such that patients might have different views about the best choice depending on personal preferences, the Analysis should not make categorical recommendations but should instead advise shared decision-making based on the values patients assign to potential outcomes. |
| 1. Recomendations can be made only if there is direct scientific evidence of improved health outcomes (see no. 2), not because a panel member believes there is benefit nor because it is accepted practice in CML care. When such evidence is lacking, the results of the analysis should state: “There is insufficient evidence to make a recommendation.” |
| 2. The Analysis will not result in recommendation for one intervention over another unless there is evidence from a controlled study (internal controls) or from dramatic findings in an uncontrolled study that patients treated with that intervention experience better outcomes (eg, higher survival) than those treated by the alternative. The outcomes that matter most are those that patients experience (eg, lengthened survival), not intermediate outcomes for which the linkage to health outcomes is less certain (eg, cytogenetic remission). |
| 3. When extrapolations of evidence are made from one patient population to another to infer effectiveness, the Analysis should make this explicit and discuss the implications. |
| 4. Claims of proofshould be accompanied by full disclosure of the limitations of the evidence. |
| 5. Claims about benefit should clarify the magnitude of benefit, preferably in absolute terms rather than as relative benefit. These claims should be accompanied by a description of potential harms, preferably by estimating the probability of these harms. Confidence intervals should be used to clarify the range of uncertainty about the estimates. |
| 6. When there are complex tradeoffs between benefits and harms such that patients might have different views about the best choice depending on personal preferences, the Analysis should not make categorical recommendations but should instead advise shared decision-making based on the values patients assign to potential outcomes. |