For inclusion, the studies were required to meet the following criteria 202122: a) Design: studies conducted in 10 or more patients that contained relevant primary clinical data; b) Population: patients with symptomatic BPH; c) Intervention: TUNA; d) Comparator: any other medical or surgical procedure; e) Outcomes: the studies were required to include quantitative information relating to at least one of the following primary interest variables: symptom score, quality-of-life score, maximum urinary flow rate (Qmax), post-void residual volume (PVR), pressure-flow studies, need for new therapeutic interventions, adverse effects. The use of health care resources was included as secondary variable. Two independent reviewers carried out study selection and data extraction. A third reviewer checked the resulting extractions and the team resolved any discrepancies.

Methodological quality was evaluated for each selected paper using previously validated guidelines 19.

Intercooled Stata 8 computer software (StataCorp LP Texas USA 1984–2005) was used to obtain an overall measure of the effect of TUNA on the outcomes of interest. Quantitative combination of results was made only when studies showed analogy in terms of type of intervention and clinical homogeneity. We used as a definition of clinical homogeneity that trials have fixed and clearly defined inclusion criteria and fixed and clearly defined outcomes 23. The meta-analysis was conducted using a random-effects model. Dichotomous outcomes were analyzed using Odds Ratio (OR) or expressed as proportions with their corresponding 95% confidence interval. Continuous variables were analyzed as the difference in means (with 95% confidence interval) between pre- and post-treatment values at the respective assessment dates, with the difference in variances being estimated using standard formulae 24. In cases where the original study solely furnished the mean of a continuous variable, standard deviation was attributed by weighting the standard deviations of the selfsame variable in other studies 24. Difference in means was calculated as Weighed Mean Difference (WMD) but when different methods of measurement were used among the studies- such as in symptom scores- we used Standardized Mean Difference (SMD). Statistical heterogeneity was analyzed using the X² statistic with a P value of < .05. Sensitivity analyses were performed to assess the influence of methodological issues, such as study design, on the overall effect estimation. Absolute Risk Reduction and the Number Needed to Treat (NNT) were estimated where appropriate. The NNT was calculated using the Internet-accessible Visual Rx program http://www.nntonline.net. Where it proved impossible for meta-analytic techniques to be implemented owing to the characteristics of the studies or their manner of expressing the results, an individual analysis was performed. Results were calculated on the basis of the respective studies' initial populations, provided that this was possible and a P value of < .05 was deemed significant.

Table 1 contains a summary of the 26 studies included. Although a number of efforts were made to prevent any possible duplication of patients, this cannot be entirely ruled in some studies 3948. In general, patients were diagnosed as having symptomatic BPH with a symptom duration of over 3 months. Whereas some studies solely included patients with acute refractory urinary retention 263031, other authors 2947 specifically considered this a criterion of exclusion. In contrast, while median lobe hypertrophy was generally deemed to be an exclusion criterion for TUNA, 2 studies 4045 typically included this group of patients. Thirteen studies 26282930333536384143464849 furnished data on baseline prostatic size, with the mean value being 45.5 ± 8.8 (range:38–66). Only 3 studies 424449 had follow-up periods exceeding 2 years. Only 4 studies 26333945 expressly stated that industry had been the source, totally or partially, of funding.

In terms of grades of scientific evidence, these were level-IV studies. Even so, study quality, based on the principle of methodologic biases, was acceptable. As shown in Table 1, most were prospective studies, used clear and, in general, homogeneous inclusion and exclusion criteria, plus outcome measures, objective as well as subjective, that were validated and likewise homogeneous. Consequently, selection, performance and detection risks would thus appear to be relatively small. Nevertheless, none contained an express statement to the effect that assessment of results had been performed blindly or independently. Furthermore, despite the fact that some studies reported a scant number of losses to follow-up, in others this number was considerable or unspecified.

All studies compared TUNA with TURP whereas 4 compared TUNA with other minimally invasive techniques such as: Transurethral Microwave Thermotherapy (TUMT), Water-induced Thermotherapy (WIT), Interstitial Laser Coagulation (ILC), Transurethral Electrovaporization (TUVP), Visual Laser Ablation (VLAP) and Transrectal High Intensity Focused Ultrasound (HIFU).

In terms of design, 4 were non-randomized and 5 were randomized studies. Of these, three 505154 were papers based on a single clinical trial. Nevertheless, it was decided that all should be included since they furnished information on different variables and follow-up periods. Bearing this in mind, the number of patients included in the randomized studies was 336 (167 treated with TUNA, 169 with TURP). Their general and methodologic characteristics are shown in Table 4. In line with these characteristics and biases, these studies provide level-II evidence.

In addition, we identified 4 non-randomized studies (level-III evidence) 55565758. Of the two papers by Schaltz which correspond to a single study, the more complete 57 was used for analysis of efficacy. We also decided to include the 1997 paper 55, since it furnished safety data that were not included in the later paper. The characteristics of these studies are set out in Table 4. Allocation to each treatment arm was made in line with the wishes of the patient 56, sequentially 5557 or in accordance with anesthetic risk, prostatic volume and the patient's desire to maintain normal ejaculation 58.

The results of these comparisons are shown in Tables 6 and 7. Arai's study 56 showed that, in the assessment at 3 months, TUNA-related improvements were significantly superior to those of TUMT in terms of symptom index, quality-of-life score, Qmax and PVR; in addition, there was a trend, albeit non-significant, towards a lower number of erectile dysfunctions.

Likewise, TUNA seemed to have a significantly better effect than WIT on the objective variables analyzed in the study by Minardi 58. The results vis-à-vis ILC proved more difficult to analyze, as each of the techniques seemed somewhat more effective than the other, depending on the point in time at which they were evaluated. However, adverse effects tended to be more frequent in the group treated with ILC, particularly in the case of irritative symptoms, postoperative hematuria, and reduction in seminal volume (Table 7).

Analysis of the studies that compared TUNA with TVP 555758 indicated that TVP achieved improvements significantly superior to TUNA in both subjective and objective variables, but that this was accompanied by a higher incidence of adverse effects.

As against VLAP, Schatzl's results based on 30 patients indicated that, while both techniques led to similar improvements in the symptom index, VLAP exercised a significantly superior effect on the objective variables (Qmax and PVR). Furthermore, there seemed to be fewer adverse effects with this technique. Notable differences between TUNA and HIFU were not in evidence in the analysis of the 35 subjects in whom these techniques were compared.

It has to be recognized that publication bias is possible and that, by not including unpublished studies, the beneficial effects of the TUNA system may have been overestimated. Nevertheless, we feel that any such bias would necessarily have been minimized by the scope of and systematic strategy used in the search of the literature and are confident that majority of the research conducted in this field was successfully identified. So, in accordance with some recommendations 1922 we have made a broad search of literature using at least three database sources supplemented by other search strategies. In addition, we have included both fully published studies and grey literature 60. However, in line with prior reports we decided not to include unpublished data from industry given both the difficulties encountered in obtaining this information and the recognition that the use of these data may not necessarily reduce the bias in meta-analysis 6061.

We also endeavored to reduce such biases by having the studies separately evaluated by two independent reviewers and using defined criteria for the purposes of study inclusion and analysis but, in accordance with some recent literature 6062 we have not used funnel plots to examine the possibility of publication bias given the limitations and potential misleading results of these graphs.

The methodologic quality of the studies, and particularly the high percentage of non-controlled case-series could constitute another limitation of this study. This situation is particularly common in surgery 63, where less than 10% of the evidence sustaining surgical interventions comes from randomized studies 64 and where systematic reviews play an essential role in evaluating new technologies 65.

Although we are aware of the potential difficulties and limitations posed by such evidence, we completely agree with recent publications highlighting the usefulness of these studies for evaluating the efficacy and safety of therapeutic devices and procedures 763646566676869. The evaluation of such studies in systematic reviews provides information on the possible benefits and risks of the procedures and on the limitations of the evidence supporting their use. Such studies moreover help identify those techniques that should be subjected to more rigorous evaluation, and identify areas of uncertainty which otherwise would not be addressed and which would allow some procedures to be accepted and others rejected in clinical practice, on the basis of individual appeal alone 65666768. It is therefore not surprising that a large percentage of systematic reviews conducted by organizations of repute 6368 include studies of case series, or that over 50% of all meta-analyses in the literature correspond to non-randomized studies 69. Furthermore, the results of recent studies suggest that non-randomized and randomized trials may yield similar estimations of the effects of a given treatment 70 and that meta-analyses of non-randomized studies may produce results similar to those of meta-analyses of randomized clinical trials 21.

Here we have to say that we have tried to reduce the possible impact of the studies' design using a conservative random-effects model to combine the results of the individual reports in the meta-analysis 71, and that sensitivity analysis by excluding poorer quality studies in comparisons with TURP did not change our results. However, one of the main problems of the existing studies on TUNA is the absence of independent evaluation of the results obtained. To this we must also add that the number of losses was not reported in a fair number of studies. Consequently, it has to be recognized that detection and attrition biases may have influenced to some degree the results of the review 1972.

A further problem could reside, in our opinion, in the fact that possible bias in the subjective variables might lead to the beneficial effects of TUNA being overestimated. In clinical drug trials, it has been observed that this effect is blocked by administering a course of placebo-based therapy to patients prior to the start of the study 73. Yet, the studies included make no mention of such a measure having been taken, so that the influence of such a placebo effect on the results cannot be ruled out, an effect that is known to cause an improvement in symptoms and other functional parameters, such as Qmax, as well as clinically important adverse effects 73.

Apart from the need for the controlled clinical research and sound methodologic quality that would resolve the above-mentioned uncertainties, the following elements should be borne in mind before TUNA is sought to be introduced into routine clinical practice. Firstly, with regard to its indication, current evidence show that the potential advantages of TUNA in terms of morbidity and anesthetic requirements should be balanced with the fact that it does not reach the same level of efficacy and long-lasting success as TURP. Secondly, due to the fact that it has a greater impact on symptoms and less-pronounced impact on obstructive parameters, TUNA would appear to be especially indicated in the early stages of BPH. Nevertheless, the optimal point in the course of BPH progression for performing the technique has not yet been identified. Thirdly, account must be taken of the high re-treatment rate when the results of the procedure are presented in terms of use of healthcare resources.