Table 1.
Characteristics of pairs of Cochrane reviews (C) and industry supported paper based reviews (I) of the same drugs
|
References
|
Disease
|
Interventions
|
No of included trials (both reviews/C only/I only)
|
No of included patients (C/I)
|
Significant difference favouring drug of interest
|
Comments
|
|
|---|---|---|---|---|---|---|---|
| Benefits (C/I) | Harms (C/I) | ||||||
| w1, w4 | Acute coronary syndrome | Enoxaparin v unfractionated heparin | 2/0/0 | 7081/7081 | Yes/yes | No/no | C: control drug is cheaper; new trials with longer follow-up are needed. I: health economic study shows that enoxaparin lowers total costs of care. Our comment: both reviews found that control drug causes significantly fewer minor bleeds |
| w5, w6 | Schizophrenia or similar psychoses | Amisulpride v typical antipsychotics | 14/0/0 | 1702/1700 | Yes/yes | Yes/yes | Both reviews noted that funnel plot suggested publication bias in favour of amisulpride. C: the result must be considered with provisos; amisulpride is expensive. I: dismissed finding, noting that according to manufacturer no further studies had been done; no conclusion. Our comment: bias should not be dismissed; reasons for sample size bias other than selective publication exist |
| w7, w8 | Rheumatoid arthritis | Celecoxib v other NSAIDs | 4/1/1 | 4465/4191 | No/no | Yes/yes | C: 12 month results from CLASS study suggest that short term benefit of celecoxib on gastrointestinal ulcers may not persist; this is important as rheumatoid arthritis is a chronic disease and patients are likely to be taking celecoxib for extended periods; increased cost. I: no such reservations |
| w9, w10 | Schizophrenia | Risperidone v haloperidol | 6/4/0 | 2326/1047 | Yes/yes | No/yes | C: more weight gain with risperidone, which is costly; included schizophrenia-like psychoses as no evidence that these should be treated differently from schizophrenia; funnel plots showed greater benefit for risperidone in smaller studies. I: all results favoured risperidone; no comment on possible sample size bias. Our comment: inconsistency in inclusion of patients in relation to dose of risperidone in I review; excluded trials had rather negative results |
| w11, w12 | Asthma | Inhaled fluticasone v budesonide and beclometasone | 14/16/0 | 7775/3564 | Yes/yes | No/yes | C: plasma cortisol is an unreliable measure of harm. I: plasma cortisol favours experimental drug. Our comment: no information on where studies in I review came from and only those published after 1995 included; C review included 10 such studies that were missing in I review; complete discrepancy between two reviews in studies with cortisol values, and for several studies in I review C authors did not know that cortisol had been measured; authors of I review had access to individual patient data; all requests by authors of C review to get access to further data from authors of trial reports were unsuccessful |
| w13, w14 | Asthma | Salmeterol v theophylline | 3/2/6 | 777/1330 | No/yes | Yes/yes | C: trend towards better effect of salmeterol, but inconsistent reporting of data precluded meta-analysis; cost analysis needed. I: significant differences in favour of salmeterol for all efficacy outcomes. Our comment: I review included only studies carried out by company; three missing studies in C review were unpublished, two were abstracts, and one was untraceable as reference in I review was wrong |
| w15, w16 | Depression | Paroxetine v TCAs | 15/17/24 | 5910/3758 | NA/no | Yes/yes | C: SSRIs up to 30 times more costly than TCAs. I: no search strategy; company's worldwide clinical database used. Our comment: 12 trials in C review indisputably fulfilled inclusion criteria for I review but were missing; 20 trials included only in I review were “data on file;” C review had missed three small trials, two of which were published only as abstracts |
| w17, w18 | Depression | Venlafaxine v other antidepressants | 0/4/5 | 708/450 | No/yes | NA/no | C: costs of SSRIs and limited benefit do not justify routine first line use. I: economic analyses have shown that venlafaxine is cost effective. Our comment: no significant benefit in favour of venlafaxine in C review; I review was obscure; unclear selection criteria (“some selected SSRIs” and “certain TCAs”); impossible to verify which trials were included as references were to trial protocols; inappropriate meta-analysis of study arms |
NA=not available; NSAID=non-steroidal anti-inflammatory drug; SSRI=selective serotonin reuptake inhibitor; TCA=tricyclic antidepressant.