Summary of main results
Due to clinical heterogeneity, we did not perform meta-analyses. Of the 18 included studies, only three (Ouyang 2010; Wang 2010; Xue 1999) indicated that compared with antiviral drugs, Chinese medicinal herbs showed a superior effect in preventing influenza and alleviating influenza symptoms. The remain 15 studies (Chen 2010a; Chen 2010b; Jin 2010; Li 2009; Li 2010; Qian 2011; Shi 2004; Tan 2010; Wei 2010; Xie 2010; Zhang 2010a; Zhang 2011; Zhao 2010; Zheng 2010b; Zhu 2010) had a similar effect to antiviral drugs. No obvious adverse events were reported in the included studies. However, the small number of participants and studies, together with the poor quality of these studies, does not allow us to draw reliable conclusions.
Overall completeness and applicability of evidence
Definitive conclusions could not be reached as differences between the traditional Chinese medicine (TCM) formulations in the included studies lower the generalisability of the results regarding the effectiveness of TCM for patients with influenza.
The applicability of the included studies was limited, since their inclusion criteria, interventions, durations and outcome measures were different. More well-designed trials are required.
Quality of the evidence
We rated the quality of the evidence from the included studies as very low to low, and the reasons for this are listed below.
Most of the retrieved studies did not give adequate descriptions of the methodology used, which may have misled us if we had not clarified the details, for example, inclusion of non-randomised trials and classification the trials into category B rather than C. It was an exhausting but necessary process to interview every primary trial author before deciding whether to include these trials, when the methodological details were not reported. Contacting authors by telephone was more effective than writing to them because of a higher response rate. However, even after confirmation of true randomisation, we found that the methodological quality of the studies remained poor.
Allocation concealment is an important marker of trial quality. However, very few potential articles considered for our review reported or performed allocation concealment; the included trials failed to perform allocation concealment, leading to high risk of selection and confounding bias.
In one of the included trials, no blinding was used for either the participants or the investigators, which led to a high risk of performance bias. None of the studies mentioned blinding to the outcome assessors, which promotes suspicion of detection bias. Publication bias may exist as all the included studies were published in Chinese and no primary articles reporting negative results were found. The huge difference in the number of participants between the two arms raises suspicion of inadequate randomisation or a significant number of withdrawals, which may have led to high selection or attrition bias in one study (Xue 1999).
During the process of interviewing the trial authors, we understood that it was difficult for them to perform double-blinding because of certain features associated with Chinese medicinal herbs, for example, aroma and appearance. Capsules were used in one study (Xue 1999). Other methods included extracts from herbal medicines administered by injection by using an opaque cover around the fluid bag if the herb was of a particular colour. Many trials are conducted to assess the efficacy of a plant before making the expensive decision to produce it as a patented medicine and double-blinding is almost impossible.
All the patients in the included studies were diagnosed by epidemiology, clinical symptoms and routine tests. It is possible that participants with other acute respiratory infections not caused by the influenza virus, such as the common cold, may have been misdiagnosed as having influenza and were included in the trials. The disease duration on entry varied between the potential studies we retrieved for inclusion. Secondary bacterial infection or other complications that complicate influenza treatment may have been present, even if the trial authors did not find or report them.
In the practice of TCM, herbal preparations should match the type of 'zheng' which equates to a diagnosis. Trial authors are encouraged to explain each 'zheng' by using conventional medical terms, therefore making it more convenient for physicians and consumers to choose an appropriate preparation.
Regarding interventions, we considered the commonly used antiviral and antipyretic-analgesic drugs to be acceptable controls. However, there is potential for bias. If the trial author knows that a 'positive' drug was used and the study was an 'equal effect test' study, there is a potential risk that the outcome detectors will consider similar results for the two groups. In this case, even double-blinding is useless. If it is a 'superior effect test', the trial authors tend to overestimate the effect in the treatment group if allocation concealment and blinding were inadequate. When a Chinese herbal medicine combined with a supposed 'positive' intervention is found to be more effective than the 'positive' drug alone for influenza, this herbal medicine is considered effective. An alternative would be to compare Chinese medicinal herbs to a placebo (it is also recommended to compare first to placebo to test effectiveness and subsequently to compare to another treatment that was tested against placebo and proved as effective), with another 'positive' drug given to both arms.
For superior trial design, one of the key techniques for avoiding performance and detection bias is blinding. As most TCMs have particular characteristics, blinding is rather difficult. The TCM industry should develop a simulation agent when designing trials.
Although Chinese herbal medicines as a treatment for influenza and the method of manufacturing these medicines are widely accepted in China, most of the constituents of the pharmacologically prepared drugs used in trials cannot be specified. This is in marked contrast to the pharmacological agents used in conventional medicine, for which the chemical constituents, their quantities and the percentage of any impurities or contaminants are precisely known. In addition, the variation between different production batches of conventional medicines is kept within specified limits. In contrast, variation between formulations and batches of pharmacological agents is inevitable in TCM, although the Chinese government specifies the acceptable limits of variation. This variation is a factor that may contribute to any heterogeneity between different study results. The application of TCM signs is also limited as not everyone is familiar with them. However, one must accept that the overall treatment concept for TCM is different to that used in conventional medicines.
Ten included studies reported slight adverse reactions. This suggests that the TCMs used in the included studies are safe.
The definition and timing of outcome measures varied between studies. The outcome measures, defined in the primary version of this review, were based on a subjective assessment of defervescence and symptom withdrawal using dichotomous data. We may have missed additional information from studies which did not use the outcome measures stated in our original review. In this updated review, we added continuous data for duration of defervescence and symptom withdrawal, as well as influenza incidence in the prophylaxis studies. In one of the included studies (Shi 2004) ibuprofen was added temporally to patients with high fever, whereas no data were provided about how many participants in each group received the extra drug. This may have influenced the results.
TCM signs are important outcome measures in traditional practice. We will consider including TCM signs as a secondary or an additional outcome in the next update of this review. However, it is difficult to compare or quantify TCM signs as they have subjective outcomes. For example, 'mai xiang' equates to pulse presentations. Diagnosing 'mai xiang' in TCM is a complex and difficult technique, dependent on the TCM physician's experience. TCM researchers and physicians should find a gold standard method which is repeatable and easy to practice when measuring TCM signs.
In addition to the methodological limitations, the imprecision of the results is a common problem in each included study. The confidence intervals for the effects were wide in most of the results. Another problem is that most Chinese medicines were tested in one study only.
Potential biases in the review process
Most of the trials did not adequately report their methodology in the original publications, so we obtained this information by telephone communication with the authors. The studies were conducted several years ago and may be influenced by recall bias.
Agreements and disagreements with other studies or reviews
The results of well-designed randomised controlled trials (RCTs) with large sample sizes in the future may confirm or refute our current conclusions. There is no other known systematic review of TCM for influenza.