Graham 2003.
| Methods | ||
| Participants | ||
| Interventions | ||
| Outcomes | ||
| Notes | See Table 1 | |
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Random sequence generation (selection bias) | Unclear risk | “Random assignment [...]” Method of randomisation not described. |
| Allocation concealment (selection bias) | Low risk | “Random assignment was conducted by telephone to a separate data management centre. Patients and physicians were blinded to product allocation.” |
| Blinding (performance bias and detection bias) Subjective outcomes | High risk | Blinding was not done for non‐fragrant 'placebo' group, but was done for the pure essential oil and fragrant placebo groups. Although patients were not told their group allocation, only 9% of the non‐fragrant placebo group believed they had received the essential oil, compared to 25% of the fragrant placebo and 24% of the pure essential oil (P = 0.006). Therefore the non‐fragrant 'placebo' was not a true placebo control. |
| Blinding (performance bias and detection bias) Objective outcomes | Unclear risk | The study did not assess this outcome. |
| Incomplete outcome data (attrition bias) All outcomes | Low risk | Withdraws or dropouts: “The reason for not receiving the full number of possible aromatherapy treatments included research assistant illness and absence, failure to complete the planned radiotherapy course, and withdraw from receiving aromatherapy. Of the 313 patients who were randomly assigned, 285, 286, and 295 completed baseline, HADS anxiety (HADSA), HADS depression (HADSD), and SPHERE questionnaire, respectively providing 91% to 94% questionnaire compliance for analysis of these scores. There was no significant difference in missing data by allocated arms.” |
| Selective reporting (reporting bias) | Unclear risk | There may be a risk of bias from selective outcome reporting, given that the results are presented as the proportion of participants whose score on an ordinal variable exceeded a cut‐off point (> 7); the data could instead have been reported as mean and SD. Normative data for the HAD suggest ≤7 is normal, 8 to 10 is borderline, and ≥11 may indicate clinically relevant anxiety (Millar 1995); this presents a rationale for the cut‐off. |
| Protection from contamination? | Low risk | Interventions administered on an individual basis. |