Always important—statistical justification for pooling heterogeneous studies?

Abstract

To the Editor: We read with interest the meta-analysis by Marmar et al. (1Marmar J.L. Agarwal A. Prabakaran S. Agarwal R. Short R.A. Benoff S. et al.Reassessing the value of varicocelectomy as a treatment for male subfertility with a new meta-analysis.Fertil Steril. 2007; 88: 639-648Abstract Full Text Full Text PDF PubMed Scopus (207) Google Scholar), which favors surgical varicocelectomy in infertile men with palpable lesions and at least one abnormal semen parameter. These results are in direct contrast with earlier studies by Evers and Collins (2Evers J.L. Collins J.A. Surgery or embolisation for varicocele in subfertile men.Cochrane Database Syst Rev. 2004; (CD000479)PubMed Google Scholar) and expected to have a major impact in clinical practice, because meta-analyses are considered to be level I evidence. Although the new study has many advantages, such as inclusion of patients with clinical varicocele and abnormal semen parameters only, it raises some concerns about specific methodologic issues:1)Many randomized controlled trials (RCTs) were excluded from the meta-analysis, because they had included some patients with subclinical varicocele or patients treated with embolization. Nevertheless, RCTs by Breznik et al. (3Breznik R. Vlaisavljevic V. Borko E. Treatment of varicocele and male fertility.Arch Androl. 1993; 30: 157-160Crossref PubMed Scopus (71) Google Scholar) and Nieschlag et al. (4Nieschlag E. Hertle L. Fischedick A. Abshagen K. Behre H.M. Update on treatment of varicocele: counselling as effective as occlusion of the vena spermatica.Hum Reprod. 1998; 13: 2147-2150Crossref PubMed Scopus (183) Google Scholar) studied many patients who fulfil the inclusion criteria and could be incorporated in the present meta-analysis after communication with the authors of the original studies.2)Owing to strict exclusion criteria and obvious lack of RCTs, observational studies were included in the meta-analysis, despite being considered to be of inferior quality, according to the principles of Evidence-Based Medicine. Although the authors clearly state that “the outcome measure was the effect of surgical varicocelectomy on natural or spontaneous pregnancy outcome during follow-up of up to 24 months,” some observational studies (Okuyama et al. [5Okuyama A. Fujisue H. Matsui T. Doi Y. Takeyama M. Nakamura M. et al.Surgical repair of varicocele: effective treatment for subfertile men in a controlled stusy.Eur Urol. 1998; 14: 298-300Google Scholar] and Onozawa et al. [6Onozawa M. Endo F. Suetomi T. Takeshima H. Akaza H. Clinical study of varicocele: statistical analysis and results of long term follow-up.Int J Urol. 2002; 9: 455-461Crossref PubMed Scopus (34) Google Scholar]) report pregnancy rates during longer follow-up periods (mean 33.9 and 76.1 months, respectively). We believe that inclusion of suitable data from RCTs as well as exclusion of data from specific observational studies could have significantly altered the results of this meta-analysis and therefore led to different clinical decisions.