Reboxetine for acute treatment of major depression: systematic review and meta-analysis of published and unpublished placebo and selective serotonin reuptake inhibitor controlled trials. BMJ 2010; 341: c4737. Cipriani and colleagues 1 confirm previous findings of a small effect of antidepressants over placebo for depression; however, there are some limitations to their assessment of risk of bias and the classification of included trials. Their method of classification was not in accordance with the Cochrane Handbook, 2 as stated by the authors. 1 According to their assessments of risk of bias and the categories proposed by Cochrane (table), only one (<1%) of 522 included trials fulfilled the criteria for low risk of bias, 383 (73%) had unclear risk of bias, and 138 (26%) had high risk of bias. Several important issues were not considered. First, well known sideeffects of antidepressants mean that adequate blinding in placebocontrolled trials is unlikely. 3 Second, the assessment of selective outcome reporting bias did not consider the three secondary outcomes 4 (eg, dropouts due to adverse events, which were not reported in 91 [17%] of 522 trials). 1 Third, assessment of attrition bias was done on the basis of arbitrary drop out thresholds and did not consider reasons for dropout; 4 all drugs had increased dropout rates due to adverse events compared with placebo. 1 Fourth, the authors rated the studies according to sponsorship but used these ratings in a covariate analysis only, even though vested interests have been shown to intro duce bias and significantly affect trial results. 5 Finally, the authors failed to provide their certainty of evi dence (in accordance with the GRADE framework) for the placebo comparisons. Given the issues we have outlined, certainty in the estimated effects should be rated as very low and be interpreted with caution. We declare no competing interests.

In 28 febrile patients with malignant lymphoma or leukaemia, the hourly temperatures were recorded following an oral dose of 125 mg naproxen (50% of normal single adult analgesic dose). 15 patients had clinical infection, and 13 had fever secondary to their malignant disease. Compared to controls, there was no significant antipyretic effect of 125 mg naproxen in infected patients, whereas this small dose in patients without infection had a significant effect. In the uninfected patients, the antipyretic effect was significantly more marked in fever related to Hodgkin's disease than to non‐Hodgkin lymphoma or leukaemia. This selective antipyretic effect of a prostaglandin‐synthesis inhibitor in turnour‐related fever, especially in Hodgkin's disease, is unexplained but may be useful in the palliative treatment of patients with advanced disease.