Table 2 Classification of misleading reporting strategies used by authors when presenting prognostic factor studies in oncology, and frequency of each strategy in a sample of 98 prognostic factor studies published in oncology journals with an impact factor of seven or greater
From: Overinterpretation and misreporting of prognostic factor studies in oncology: a systematic review
Misleading reporting strategy | Places within a prognostic factor oncology study report where this strategy can occur | Test to check whether this strategy was used | Number of studies using this strategy | |
|---|---|---|---|---|
Main text | Abstract | |||
Selective reporting | Difference between the numbers of outcomes and prognostic factors pre-specified in Methods section and reported in Results section | The prognostic factor effect for a specific outcome or a specific prognostic factor is missing | 35 | 34 |
Difference between the numbers of subgroup and subpopulation criteria pre-specified in Methods section and reported in Results section | The prognostic factor effect for a specific subgroup or a subpopulation is missing | 5 | Not assessed | |
Inconsistent use of statistics across all prognostic factor–outcome associations, when more than one prognostic factor effects are assessed | Some PF effects are reported within a multivariable model, whereas others are reported within a univariate analysis | 24 | 41 | |
Incomplete reporting of subgroup analysis results for the prognostic factor effect | For any subgroup analysis reported, whether pre-specified or not, an interaction test p value is not reported | 35 (out of 90 studies assessing more than one prognostic factor effect) | 42 (out of 90 studies assessing more than one prognostic factor effect) | |
Incomplete reporting | Incomplete reporting of the main analysis results for the prognostic factor effect | Only adjusted hazard ratios are reported for the prognostic factor effect, not 95% confidence intervals or p values; OR Only a p value is reported for the prognostic factor effect, not adjusted hazard ratios; OR No statistical results are reported | 24 | 41 |
Incomplete reporting of subgroup analysis results for the prognostic factor effect | For any subgroup analysis reported, whether pre-specified or not, an interaction test p value is not reported | 29 (out of 51 studies reporting a subgroup analysis in the main text) | 23 (out of 28 studies reporting a subgroup analysis in the abstract) | |
