In this meta-analysis, we investigated the effect of Iscador on the survival rates of cancer patients. We were able to demonstrate that adjuvant treatment with Iscador was associated with a significant overall enhancement of survival rates. In particular, the benefit of Iscador with respect to survival was clearly seen in matched pair studies, but notably less in randomized trials. We also identified the design of the study as a source of variation among the studies on the survival of cancer patients using a meta-regression approach.
With the rapid evolution of both surgical and chemotherapeutical therapies, the question of survival of cancer patients is extensively debated. While prognosis of some cancer entities, i.e. recurrent breast cancer, has improved within the last decades [30
], other cancers only tend to be moderately impressive, i.e. gall bladder cancer [31
]. As a consequence different oncological treatment regimes were test - and not all of them were convincing. Other attempts were the application of low-dose chemotherapeutic regimen in distinct patients, i.e. advanced unresectable hepatocellular carcinoma, which were quite effective without worsening the quality of life of the patients [32
]. Thus, more tolerable treatment regimes for cancer or the combination of conventional chemotherapy with adjuvant therapies gain more and more attention in the light of the primary aim to improve the situation of cancer patients. From that point of view it is worth to mention that a similar meta-analysis investigated the effects of VA-E used as an adjuvant in the treatment of cancer patients showed moderate improvements on quality of life (Büssing et al., in preparation); however, the methodological quality of these studies was quite heterogeneous.
Quite similar, the studies included in this meta-analysis were of moderate or even poor quality (additional file 1
). Although studies in this area face a number of unique challenges and therefore might be difficult to conduct, most of the studies were on a low level of documentation quality. In particular in older studies, traceability and transparency was missing and thus has to be considered when looking at our results.
The prospect of randomizing patients to an adjuvant mistletoe treatment often raises ethical debates and negatively impacts the feasibility of studies [33
]. Not surprisingly, the published randomized trials (not only in this area of research) are often small and consist of a selected patient populations compared to observational studies [34
]. Furthermore, due to limited resources, a properly conducted long-term follow-up in survival is often omitted.
Although the benefit of adjuvant mistletoe treatment has been demonstrated in some randomized and observational studies, a comprehensive meta-analytical approach like the present one has not been previously conducted. In 2003, Ernst et al. [12
]published a systematic review on randomized clinical trials (RCTs) and stated that "statistical pooling was not possible because of the heterogeneity of the primary studies. Therefore a narrative systematic review was conducted." We can confirm the heterogeneity of studies on the clinical effects of VA-E, but nevertheless were able to extract data from 41 studies which provided enough data to calculate HRs and their standard errors. Ernst et al. [12
] stated that the weaker studies implied benefits of VA-E, particularly in terms of quality of life, while none of the methodologically stronger studies were able to verify a benefit with respect to survival or quality of life. A Cochrane Review of Horneber et al. [13
] published in 2008 analyzed RCTs on various mistletoe extract preparations. The authors found weak evidence from RCTs that VA-E application impacts the survival of cancer patients, but that it could be effective with respective to quality of life during chemotherapy for breast cancer. In contrast, Kienle and Kiene provided a different point of view [15
]. In their systematic review of RCTs from 2003 they identified 23 studies which met their inclusion/exclusion criteria [15
]. Most of these studies reported statistically significant positive outcomes (or at least positive trends) for survival or tumour remission and quality of life, while several studies reported no effect on survival, recurrence, remission and QOL; just one study showed a negative trend for disease-free-survival [15
]. Also Kienle et al. agreed that the methodological quality of several studies was "far below the standard that is today regarded as optimal or necessary" [15
]. In 2007, Kienle and Kiene [16
] published a systematic review of prospective clinical trials on Anthroposophic mistletoe extracts and identified 16 randomized and 9 non-randomized trials. Among them, 8 of 17 trials stated a significant benefit in favour of the VA-E with respect to survival; remission of tumour and malignant effusion in 2 of 4 controlled trials; for quality of life in 3 of 5 studies, and for quality of life and reduction of side effects of cytoreductive therapies in 5 of 7 trials [16
]. They concluded that the best evidence for efficacy of VA-E exists for improvement of quality of life and reduction of side effects of cytotoxic therapies, while the survival benefit was a matter of critique [16
Methodological quality of studies on the clinical effects of VA-E has improved over the last years; however, it is not surprising that particularly the older studies did not meet the current methodological standards. Indeed, most of the identified studies did not report data on compliance and completeness of follow up, intention to treat analysis was rarely mentioned, clear description of Iscador
usage (duration, dosage) was reported in just a few cases, etc. For this analysis we did not judge RCTs as methodological 'better' that non-randomized; each methodological design has its unique weakness. It can not be ignored that results from RCTs, despite of their higher internal validity, often have a lower external validity. Particularly the studies of Grossarth-Maticek [35
] tried to address this problem and used a mixed design, i.e. they combined a randomized matched-pair study with a non-randomized matched pair study within the same trial. In most cases, both study designs exhibited similar results. Although we were aware that it is quite problematic to separate these nested RCT from the non-RCT, we nevertheless decided to do so. We noticed that the survival observed in randomized studies was lower than in the non-randomized studies, and that matched-pair studies gave significantly better results than others. However, several positive reports (and strata) were from the same origin [35
] and thus had the same methodological problems. These studies had a matched pair design, either with or without randomization; the description of the methodology was good, the discussion of potential bias factors was profound. It is obvious that the strict matching process significantly affected the number of patients enrolled in the evaluation (all studies had sample sizes of <200 subjects). Potential bias factors which might contribute to the overall positive effects described in the studies of Grossarth-Maticek et al. [35
] were discussed in detail by the authors themselves [35
], i.e., selection bias and loose matching, etc. However, because these studies started in 1973, several relevant study objectives were not available, i.e., exact dates of first diagnosis, operation, initial and follow up-data assessments and matching, socio-economic status, social support, spirituality etc. In these studies, attrition bias was less important because with the drop out of any study patient, the matching partner was also excluded and thus the balance of the groups was not severely affected [35
]. Altogether, the internal validity of these study results was limited by selection bias and confounding. Moreover, there was no written protocol and no statistical hypotheses, the sample sizes were in most cases very small, and no sample size calculation. Another intriguing fact could be that the non-randomized studies of Grossarth-Maticek's group nevertheless might have a lower external validity (generalisability), because the inclusion and exclusion criteria were not very precise and not all of them explicitly formulated in advance [35
]. Moreover, apart from the matching criteria, there were no explicit procedures for building pairs. The most important fact was raised by the authors themselves [35
], as they can not exclude the possibility that preferentially patients with a good prognosis were enrolled, since patients from both groups who died shortly after the diagnosis would not have entered the study.