Since therapeutic efficiency cannot be defined in general terms but only in respect of a specific medical situation, the EVITA algorithm can only be used to evaluate a new compound for a given indication. EVITA attributes positive or negative rating points to the drug in question and its comparator in clinical trials by assessing both therapeutic benefit and risk profile, i.e. rate and strength of adverse effects attributed to the drug.

EVITA is not intended to generate an absolute score value solely but a benefit/risk score annotated with specific information on therapeutic aim, disease category and trial setting. The therapeutic aim, distinguishing treatment from prevention, has an impact on the EVITA score (see below) (Table 1 + 2). Disease severity, however, graduated in four categories from reversible and moderate discomfort to life-threatening symptoms, has no effect on the EVITA score even though it is essential when considering the benefit of a drug as opposed to its risk profile. However, such appraisal appears to be rather subjective, if not arbitrary. Thus, the severity of the disease is categorized and documented but is not a variable in the calculation (Table 1).

EVITA assessment initially defines the conditions which must be met for the randomized controlled trials (RCTs) to be included in the evaluation process. The decision to include or exclude an RCT depends on the one hand on the presence or absence of an established therapy and on the other hand on the type of outcome variables employed in these studies. As a minimum requirement for an RCT to be taken into consideration for EVITA evaluation, a Jadad Score of at least 3 has to be met 3 . The respective decision pathway is outlined in the form of a flowchart in Figure 1. As a result of the decision-making tree, different "trial settings" of EVITA can be defined.

Trial setting A represents clinical research which addresses patient relevant outcomes including mortality, reduced progression rate, event reduction, restoration or preservation of functionality, control of symptoms and quality of life. Within this A level, trial setting A₁ applies to testing the potential superiority of a new drug over established therapy by one or more head-to-head RCTs. The clinically relevant therapeutic benefit can also be tested by comparison with placebo either directly, if there is no established therapy, or administered on top of a given established therapy, i.e. the add-on situation (trial setting A₂). If new treatment strategies are investigated with respect to acceptable surrogate outcomes only, trial setting B is attributed, with B₁ for the head-to-head situation and B₂ for the placebo comparison (including add-on situations), as described for the A trial settings. In situations with insufficient trials for a justified EVITA calculation the trial setting is defined as N/A (EVITA calculation not available).

The compound of interest gains positive score points if it is shown to be superior to the current established therapy. In the absence of such established therapy, superiority over placebo can be interpreted as therapeutic benefit. Furthermore, the compound of interest receives a positive rating if the investigational drug demonstrates a more favorable risk profile than its comparator.

Conversely, negative rating points are attributed to the drug if the clinical trials result in inconclusive data, lack of superiority or even proof of inferiority and/or a less favorable risk profile than the comparator. Both the benefit and the risk profile of the drug are evaluated on the basis of RCTs.

The point scores for efficiency are modified by the number of clinical trials, the trial setting encountered and the therapeutic aim the agent is used for (Table 2 and 3). Superiority or inferiority proven by two or more studies result in a higher positive or negative score than proof from only one study (Table 3). Studies demonstrating non-inferiority or equivalence of a new drug compared with the given established therapy have to be considered in a more differentiated way. In the case of trials designed to demonstrate the superiority of an investigational agent, a resultant positive score will be reduced by tests showing non-inferiority or equivalence. Conversely, in the case of inferiority trials, a resultant negative score will be attenuated in favor of the investigational agent by further studies showing non-inferiority or equivalence. In respect of the superiority or inferiority, two or more non-inferiority/equivalence trials will have a higher numeric impact than only one (Table 3).

Point scores are multiplicated by two when evidence is drawn from trials investigating patient-relevant outcomes (trial setting A) than from trials investigating surrogate outcomes only (trial setting B) (Table 3).

The efficiency score also takes into account the strength of the therapeutic effect, the measure of which is the number needed to treat (NNT). As can be seen from Table 2, the modification of the efficiency score due to the therapeutic strength is differentiated according to the treatment goals. For preventive situations, the NNTs may increase the benefit score, leave it unchanged or even reduce it, which is reflected in the requirement of prudently low NNTs to outweigh the potentially negative consequences of long-term drug therapy. In contrast, drugs for the therapy of acute disorders with treatment aims such as healing and symptom relief would, in general, have low NNTs. In this therapeutic situation it seems sufficient to have a positive correction factor only for agents with a particularly low NNT (Table 2).

The risk score of an investigational drug is based on information about adverse effects and interactions with other interventions. To ensure a balanced evaluation and a justified comparison with the established therapy or placebo, only data from controlled trials is included. The adverse effects are scored on the basis of severity and frequency (Table 4). EVITA considers the highest occurrence of adverse effects (AE) from each of the three severity groups Grades 4 and 5 (disabling AE or life-threatening AE or death related to AE), Grade 3 (severe and undesirable AE) and Grades 1 and 2 (mild AE or moderate AE) according to the Common Terminology Criteria for Adverse Events (CTCAE Version 3.0, 2006) 4 . CTCAE grades are accepted criteria for the evaluation of adverse effects in clinical studies, including oncology trials. Frequency is evaluated according to the guideline on Summary of Product Characteristics (SPC) 5 in the following classes: rare/very rare (below 0.1%), less common (0.1 through below 1.0%), common (1.0 through below 10.0%), and very common (10.0% and above). It should be noted that very rare adverse effects (< 0.01%) cannot usually be detected in clinical trials because of limited subject numbers. Therefore, this frequency is taken together with the frequency class rare.

The risk profile of an investigational drug cannot be evaluated in isolation. It rather has to be compared with that of the therapeutic alternative or the basal frequency of adverse effects of the underlying disease. Therefore, the score points for the risk profile are opposed to those observed in the control groups in the clinical trials, either the established therapy or placebo if applicable. Thus, the eo ipso negative value of the risk score of the investigational drug can be counterbalanced or even reversed to give a positive value if the adverse effects of the comparator dominate.

Interactions, mainly drug-drug-interactions, are a potential source of risk which can emanate from a drug. Since drug interactions are scarcely registered in clinical trials, this information is taken from the SPC.

The final stage of EVITA is adding up the efficiency and the risk score, both of which are evaluated in comparison to the established therapy, to obtain the EVITA score. Values considerably above zero indicate a likely potential that the new drug features a clinically relevant improvement in treatment; values around zero indicate an ambiguous state which might be clarified by further clinical trials.

It is unlikely that a drug will be introduced into the pharmaceutical market if its inferiority has been proven by two or more RCTs investigating patient relevant outcomes (-7.5) (Table 3) under established therapy or placebo in a preventive situation (modifying factor 2.0) (Table 2), and with an unfavorable risk and interaction profile in all severity groups (-8.0 and -2.0, respectively). Therefore, the lowest calculable EVITA score of -25.0, is not to be expected and should not arise. On the other hand, the maximum calculable EVITA score of +25.0 also appears unlikely since it would result from proof of superiority over the comparator in a treatment situation in at least two RCTs investigating patient relevant outcomes (+7.5) with low NNT (modifying factor 2.0) and a complete absence of adverse effects and interactions, while the comparator featured an unfavorable risk and interaction profile. Scores in the range of -5 to +10 are therefore to be expected as the typical EVITA result. It should be noted that EVITA may demonstrate an additional therapeutic benefit of a new compound despite a lack of superior efficiency because of a better risk and interaction profile.

Since exact score values might suggest a more precise evaluation than EVITA can provide, the EVITA score is displayed as a color-coded bar graph, with green (positive values) indicating a compound that is likely to be innovative, yellow (values around zero) indicating an unclear result at the time and red (negative values) indicating an unlikely probability that the compound will offer an additional therapeutic benefit (Figures 2a-d). Further, a traffic-type yield sign incorporating a question mark is attributed for situations in which available data are unsuitable for an EVITA calculation indicating N/A trial settings (Figure 2e).

Transparency is a major concern of the EVITA project. To accomplish this, EVITA evaluation results are presented online http://www.evita-report.de together with the complete list of the source RCTs and the reasons for excluding any particular RCT from the evaluation. This also allows for rapid revision in response to justified criticism and simplifies the inclusion of new data e.g. from a new RCT publication.

Lenalidomide (Anatomic Therapeutic Chemical (ATC) Classification: L04AX04), a structural analogue to thalidomide, was introduced in 2007 (Germany), in combination with dexamethasone for the treatment of refractory or relapsing multiple myeloma (International Classification of Diseases Version 10 (ICD-10): C90) after at least one completed previous therapy 6 7 . Numerous therapeutic approaches are currently considered established treatment alternatives for this situation: high-dose dexamethasone, anthracycline ± vincristine-containing chemotherapeutic schemes, bendamustine, cyclophosphamide, bortezomib, or a second autologous stem cell transplantation can be taken as established therapy for this situation 8 9 10 11 . Without doubt, the therapeutic aim is "treatment" of a life-threatening disease indicating disease category I.

Two studies were found which met the EVITA requirements 12 13 . However, since both trials meticulously used the identical study protocol, were performed in the identical number and profile of patients and during the same time period, they have to be considered as one clinical trial. All patients received high dose dexamethasone and were randomly assigned to either additional lenalidomide or placebo.

Outcome was primarily measured using time to progression (TTP). Strictly speaking, this variable is a surrogate outcome. However, since, the overall survival (OS), measured as a secondary outcome, showed a clear tendency in favor of the study drug in this trial, TTP was accepted as a clinically relevant outcome. Thus, EVITA trial setting A₂ applies, because lenalidomide was compared with placebo in an add-on situation with an established therapy using patient relevant outcomes (Figure 1). In the light of further established therapeutic alternatives the conclusion must be restricted to patients for whom dexamethasone appeared as the only therapeutic option.

The statistically significant outcome advantage demonstrated in one study attributes +5.0 points to lenalidomide (Table 3). As no NNT calculation is possible for the variable TTP, the higher proportion of patients, by over 38%, with an overall response was used instead, assuming a positive correlation between these variables. The NNT of 2.6 in 1.5 years (indicating 4 per year) results in a modifier factor of 1.5 (Table 2). The resulting efficiency score is +7.5.

The risk profile was calculated on the basis of observations of the adverse effect in the clinical trials used for the efficiency evaluation. There was a very similar frequency of adverse effects in the two treatment groups. Only grade 3 severity group adverse effects occurred slightly less frequently in the placebo group; in addition, there is a clinically relevant drug interaction risk, while for placebo interaction potential must be considered zero. The resultant risk score for lenalidomide is -1.5 (Table 4).

Taking efficiency and risk score together, gives an EVITA score of +6.0, indicating a rather likely chance that lenalidomide represents a patient relevant innovative drug for the specific subgroup of patients investigated in the trials cited (Figure 2a).

Pioglitazone (ATC: A10BG03) is a thiazolidindione in the treatment of diabetes mellitus type 2 (DM2) (ICD-10: E10-E14) 14 . The evaluation was carried out twice, once based on the knowledge available in 2001, i.e. shortly after drug approval, and the other based on the knowledge of 2005, i.e. after publication of the PROactive study (PROspective pioglitAzone Clinical Trial In macroVascular Events) 15 , a large outcome trial in diabetic patients.

Oral antidiabetics seek to protect the patient from macro- and microvascular sequelae, thus the therapeutic aim is "prevention". There are widely accepted established oral antidiabetic therapies, mainly using metformin and sulphonylurea compounds. In the pivotal studies 16 17 , pioglitazone was tested in an add-on trial setting, metformin or sulphonylureas being employed as basal therapy. The outcome variable was glycosylated hemoglobin A1c (HbA1c), i.e. a surrogate variable. In view of the fact that the therapeutic efficiency of alternative oral antidiabetics, mainly metformin, was assured by clinical trials using patient relevant outcomes, a trial using surrogates only should no longer be accepted according to the EVITA evaluation strategy. However, since metformin was given as a basal drug in both groups, the therapeutic benefit of other antidiabetics on top of metformin is not satisfactorily ensured, and lowering blood glucose in diabetic patients is widely accepted as a necessary therapeutic goal, the trials using HbA1c as primary efficacy outcome were accepted. Hence, EVITA trial setting B₂ applies (Figure 1).

For HbA1c, pioglitazone showed significant benefit over placebo in two trials, resulting in an efficiency score of +3.75 points, rounded to +4.0 (Table 3). The NNT for such a benefit cannot be calculated from this outcome. Therefore, since no other outcomes closely associated with the primary outcome were reported in this study the efficiency score was not modified further. The risk profile evaluation taken from the two trials showed an occurrence of less than 10% of both severe (-2.0 points) and mild to moderate (-1.0 points) adverse effects, but this applied to both the pioglitazone and the placebo treatment arm, so that the net risk for pioglitazone resulted in zero. According to the SPC, severe interactions with gemfibrozil and rifampicin may require dose changes of pioglitazone for which one negative score point is given, while for the comparator (placebo) a zero interaction potential is assumed (Table 4). The efficiency point score is +4.0 and the risk score -1.0, resulting in an EVITA score of +3.0 points which promises an additional therapeutic value of this new pharmacological candidate (Figure 2b).

Five years after approval, the results of the PROactive study were published 15 . In this trial, the effects of pioglitazone on the macrovascular outcome of over 5,000 DM2 patients treated with pioglitazone or placebo, both on top of established therapy, were investigated. The primary outcome - a composite of seven clinical events - is considered to be patient relevant so that EVITA trial setting A₂ applies, i.e. a comparison of the patient relevant outcomes of an investigational drug with those of a placebo in an add-on situation (Figure 1). This trial failed to show the superiority of pioglitazone over placebo. According to the EVITA strategy, RCTs with patient relevant outcomes rule out trials with surrogate variables. Thus, unlike the earlier evaluation, the efficiency score decreases to zero (Table 3). Even if the individual risk profiles, as extracted from the Dormandy study, were changed from -7.5 points to -6.5 points, the reconciliation of both risk profiles would remain the same (-1.0 points) (Table 4) and the EVITA score would become negative (-1.0 points) (Figure 2c). According to the EVITA philosophy, point scores around zero are considered inconclusive for the interpretation of the innovative value of a new pharmacological candidate. Thus the aforementioned promise of an additional value is negated.

It should be noted that in the same publication 15 the authors communicated an additional outcome variable comprising only three of the seven original outcome components. This new outcome variable showed a statistically significant superiority of pioglitazone over placebo. However, since this variable was not mentioned in the original protocol 18 , it cannot be considered predefined and thus cannot be included in the EVITA evaluation.

Bupropion (ATC: N06AX12, N07BA02) is a reuptake inhibitor of norepinephrine, dopamine, and, to a lesser degree, serotonine; it is also a nicotinic receptor antagonist. Originally marketed in the United States in 1984, bupropion was not introduced as an antidepressant in Germany until 2007 19 20 . Major depression (ICD-10: F32-F39) is a severe if not life-threatening disease due to its elevated risk of suicide. Therefore category I applies. Although prevention is one aspect of antidepressive therapy, its main focus in the majority of patients is on the reduction of disease burden, thus defining the therapeutic aim as "treatment".

Bupropion has been on the pharmaceutical market for over two decades, so there is a plethora of clinical trial information. The majority of the older studies do not, however, meet the quality requirements of modern trial methodology such as conformity with Good Clinical Practice (GCP) or they focused on side effects such as disorders in sexual function as primary outcome. Therefore, they are unsuitable for EVITA evaluation of the therapeutic efficiency of bupropion. Some of the recently published studies 21 22 23 used placebo as the comparator. This cannot be accepted in indications where established therapies exist. Thus, only a minority of trials met the requirements for EVITA evaluation.

To fully assess all studies of antidepressant efficiency with an acceptable comparator, not only the trials published in peer-reviewed scientific journals were inspected, but also those published as abstracts on the company website, "GlaxoSmithKline (GSK) Clinical Trial Register". Finally, three trials were identified as suitable for EVITA evaluation. In a six-week study of 100 depressed patients 24 , bupropion did not show superiority with respect to the Hamilton-scale (HAMD-17) over paroxetin, an antidepressant acceptable as established therapy. In two three-arm studies, only available online in the GSK Clinical Trial Register 25 26 , bupropion was compared to placebo and venlafaxin using the "Montgomery-Asberg Depression Rating Scale"(MADRS) measuring changes after 8 or 10 weeks. The two studies showed divergent results. In one 25 , while patients treated with bupropion showed a larger MADRS-score reduction than those under placebo, there was no difference in comparison to venlafaxine. In the other trial 26 bupropion induced similar MADRS-score reductions compared to placebo, albeit a smaller one when compared to venlafaxin.

Since each of the three trials represents a direct comparison with an established therapy and since scale-scores which quantify the severity of a disease generally have to be considered as surrogate, EVITA trial setting B₁ applies (Figure 1). Therapeutic superiority over established therapy was not seen in any trial (0); however, inferiority in one study (-2.5) and equivalence in two studies (+1.25), resulted in a efficiency score of -1.25 which does not need to be adjusted for treatment strength (Table 3).

Due to the lack of complete publication of the three studies taken for the efficiency evaluation, there was insufficient information on the adverse effects profile of bupropion. Therefore, the respective SPC information was used to evaluate the risk profile. For comparison, the risk profile information in the SPCs of venlafaxin and paroxetin were considered. Since all compounds have been on the market for many years now, this procedure appeared to be justified. Although not all adverse effects (AE) were attributable to all the compounds, bupropion and its comparators paroxetin and venlafaxin, featured similar AE frequencies in all severity groups. While taking these antidepressants, patients experienced mild AE or moderate AE (Grades 1 and 2) very commonly, severe and undesirable AE (Grade 3) very commonly, and disabling AE or life-threatening AE or death related to AE (Grades 4 and 5) less commonly. Drug interactions occur occasionally and can also lead to clinical consequences during treatment with all of the three compounds. Thus, a point score of (-1.5) + (-2.5) + (-2.0) + (-1.5) = (-7.5) was calculated for the risk profiles of both bupropion and its comparators (venlafaxin and paroxetin), resulting in a risk score of zero for bupropion (Table 4).

The EVITA score for bupropion, obtained by adding up the -1.25 points for efficiency and the zero value for the relative risk profile was -1.25, indicating that an additional therapeutic benefit for bupropion in the treatment of major depression can hardly be expected (Figure 2d).

Zoledronic acid (ATC: M05BA08) is a nitrogen-containing bisphosphonate which inhibits osteoclast-mediated bone resorption 27 28 . In Europe, intravenous zoledronic acid (4 mg administered every 3 to 4 weeks) was approved in 2001 for the prevention of skeletal complications in patients with bone metastases and for the treatment of tumor-induced hypercalcaemia 29 . It was introduced in 2005 as an annual intravenous formulation (5 mg/y) for the treatment of Morbus Paget 30 . In 2007, it was approved in the same dosage for the treatment of patients with PMO on high risk of fracture 31 32 (ICD-10: C80-C81), for which many other intravenous as well as oral bisphosphonates have been granted market authorization. PMO is a severe disease due to its elevated risk fractures, which if affecting the hip might be life-threatening. Therefore "treatment" and "category 1" apply for therapeutic aim and disease severity, respectively. From evidence-based perception, bisphosphonates, particularly alendronic and risedronic acid, raloxifene and estrogenes, are considered established therapies in the treatment of osteoporosis, as they have been proven to reduce the occurrence of clinically relevant bone fractures.

Two studies of zoledronic acid (5 mg once per year) treatment of PMO have been published so far. A dose-efficiency trial 33 investigated the influence on bone mineral density, a surrogate outcome. The pivotal trial HORIZON (Health Outcomes and Reduced Incidence with Zoledronic Acid Once Yearly) 34 investigated two primary efficiency outcomes, one of surrogate quality (risk reduction of morphometric vertebral fracture) and one of patient relevant quality (risk reduction of hip fracture). However, both trials were conducted in comparison with placebo. Against the background of existing established therapies, these studies did not fulfill the requirements for EVITA evaluation. As a consequence, zoledronic acid (5 mg annually) in the indication PMO had to be rated N/A (Figure 2e).