Summary of main results

On the basis of the available study data P. sidoides extract can be considered as effective for symptom resolution of acute bronchitis in adults and in children. Statistical heterogeneity was partly explained by the use of different preparations. Only the drops of P. sidoides extract showed superiority, referred to the cure rate at day seven. The tablet preparation was not effective. There was consistency across different outcomes (all symptoms, key symptoms) for the alcoholic extract. However, there were only two studies each per age group on the alcoholic extract, and one dose‐finding, multi‐arm study each for tablets. Overall the quality of the evidence for these outcomes is considered low (summary of findings Table for the main comparison; summary of findings Table 2).

There were only single studies available for the assessment of efficacy in sinusitis and the common cold in adults, and the overall quality for these two entities was considered very low (summary of findings Table 3; summary of findings Table 4). Patients with sinusitis might benefit from P. sidoides extract (higher cure rate at day 21). For the treatment of the common cold there are only few data. Only half of the collected data for this condition are published. The reported results show efficacy after 10 days, but not after five days.

Based on a pooled analyses of all studies, there was a slightly raised risk for adverse events in the treatment group. None of the reported events were serious. Besides gastrointestinal complaints such as nausea, vomiting, diarrhea or heartburn, allergic skin reactions with pruritus and urticaria were reported in the included trials.

Overall completeness and applicability of evidence

Although we performed a broad search for all acute respiratory infections (ARIs) in both children and adults, only a limited number of conditions were covered by clinical trials of sufficient quality. These included acute bronchitis in adults and children, and acute sinusitis and the common cold in adults only. For other frequent diseases such as rhinosinusitis, non‐streptococcal pharyngitis or otitis media there were no trials, or only with high risk of bias (e.g. not blinded).

Very few studies reported on time to complete resolution of symptoms (preferred primary outcome), and none of the studies originally reported the number of patients free of symptoms at a prespecified point in time (alternative primary outcome). Therefore, we had to abstain from meta‐analysis of the preferred primary outcome. All studies used a non‐validated symptom score to show efficacy. Results of the individual studies were recalculated for this review by the investigator under the assumption that a score of 0 corresponds to complete resolution. Thus, in conclusion, although not available from the published reports, evidence could be pooled as prespecified and may thus be considered complete. However, data on quality of life and days lost from work were insufficiently reported.

Some concerns arose as all studies were performed based on a similar protocol, by the same investigator (i.e. the manufacturer) and in similar settings (multi‐center outpatient, Ukraine or Russia). This narrow spectrum compromises the representativeness of the evidence. More studies including those from other investigators and/or other parts of the world (Germany, Bulgaria, Japan, Korea) are underway (see Characteristics of ongoing studies table and Studies awaiting classification). So far, results are not available from these studies.

With respect to the assessment of safety, the relatively small studies of short treatment duration included in this review cannot be expected to pick up rare but severe events with sufficient precision. To solve this issue, other data sources may be necessary to complement our findings.

Shorter duration of symptoms or of illness due to acute respiratory infection by using herbal medicines would need to be translated into fewer lost days of work and less inappropriate antibiotic use. Studies based on (often self reported) symptom improvement remain difficult to assess because of the lack of valid, responsive outcome measures (Linder 2003).

Quality of the evidence

All included studies used similar methods and all were individually considered of moderate risk of bias (Figure 3). While randomization procedures and concealment of allocation were standard for all included studies, we had some concerns about the effectiveness of the blinding, the use of unvalidated scores for outcome assessment, minor attrition problems and, in one instance (Lizogub 2007), selective reporting. This was reflected by downgrading (‐1) for trial quality in the GRADE system for all main outcomes (summary of findings Table for the main comparison; summary of findings Table 2; summary of findings Table 3; summary of findings Table 4).

There were also concerns with respect to positive outcome, or small study bias. While the number of studies was insufficient for formal statistical testing, the combined funnel plot based on the main outcomes for all conditions and age groups was severely asymmetrical, suggesting publication bias. This resulted in further downgrading of the overall quality of the evidence (‐1 for publication bias).

Statistical heterogeneity was present in several instances, even when different preparations were accounted for by sub‐analyses. However, the direction of the results was rather uniform.

Lastly, the overall quality of the results was compromised by the low number of studies. In particular, this led to further downgrading of the quality of the evidence when only small, single studies were available (common cold and acute sinusitis). The results for these two conditions should be regarded with particular caution (very low quality).

Potential biases in the review process

A very broad search strategy was used in this review, including tailored search strategies for a large spectrum of different sources. Also, authors and investigators were very co‐operative in providing information on unpublished studies. Despite this, there is some evidence of publication bias. This resulted in downgrading the overall quality of the evidence (see above).

As all authors are German, there may be a bias in that studies published in German, or German journals, may have been more likely to be identified than articles in other non‐English languages, even though language restrictions were not applied.

Some evidence is still missing from a number of studies for which results are not yet available. This will be included in the next update of this review. At this time, all unpublished data used in the first version of this review have been published.

Quality scoring of individual studies may have been overly restrictive. For example, there was downgrading for failure to assess the success of blinding, since we considered blinding particularly important in view of the nature of the outcome assessment (symptom scores). Differential drop‐out rates led to exclusion from the analyses due to the high risk of bias of two individual studies. Both were strongly positive so inclusion would have resulted in stronger effects favoring treatment.

Agreements and disagreements with other studies or reviews

Reducing the duration of illness due to acute respiratory infections is, of course, desirable. These infections are very common and are a major cause of lost days of work and of inappropriate antibiotic use (Gonzales 2000; Wenzel 2006).

Among the different treatment options studied so far for acute bronchitis, none have been reliable or highly effective in reducing the duration of illness. The use of ß2‐agonists in the treatment of acute bronchitis in people without pulmonary disease is not supported by clinical data (Becker 2011). Although data from clinical trials are lacking, in some countries mucolytic or antitussive agents are commonly used, often in combination with antibiotics (Chalumeau 2013; Poole 2012; Wenzel 2006). Different meta‐analyses of antibiotic treatment in acute bronchitis suggested that symptoms of acute bronchitis were reduced by less than a day (Bent 1999; Fahey 1998; Smith 2011). A recent large study of antibiotics versus placebo for acute lower respiratory tract infection was unable to show differences in the duration of symptoms rated "moderately bad" or worse (Little 2013). The limited effect of antibiotic treatment in acute bronchitis is accompanied by a trend toward increased adverse events and must be weighed against potential adverse event effects such as medialization of a self limited condition, the treatment costs of antibiotics and the risk of an increasing microbial resistance (Smith 2011). Similarly, antibiotic treatment cannot be recommended in acute laryngitis, otitis media in children and the common cold (Kenealy 2013; Reveiz 2013; Venekamp 2013).

The role of herbal medicines for symptom relief of acute respiratory infections is unclear. Studies with over‐the‐counter (OTC) medications in general did not show convincing results based on a recent Cochrane review (Smith 2012). According to a systematic review, the use of Chinese herbs for acute bronchitis cannot be recommended due to limitations in trial design and an unknown safety profile of Chinese herbs (Jiang 2012). Some evidence for limited effects of Echinacea on the incidence and duration of the common cold was reported (Shah 2007), but the meta‐analytic approach in this study can be criticized. The authors of an earlier review summarized that there is no clear evidence that Echinacea may prevent common cold and, if there is any effect on the duration of the illness, only preparations of Echinacea purpurea might have this after the onset of cold symptoms (Linde 2008). A recent review showed again that Echinacea purpurea extract cannot prevent a common cold. For treatment of the common cold the extract might be effective, but issues surrounding dose and formulation require clarification (Nahas 2011). There is one study reporting that a preparation of thyme dry extract and primrose root dry extract was associated with a mean reduction in coughing fits (Kemmerich 2006). The same was shown for a fixed combination of thyme fluid extract and ivy leaves fluid extract (Kemmerich 2007). A further study showed higher cure rates for the fixed combination of a thyme fluid extract and primrose root tincture than placebo in adult patients with acute bronchitis (Gruenwald 2005). None of these study results has been replicated. Treatment effects have also been reported with a preparation containing nasturtium herb and horseradish root (Goos 2006; Goos 2007). Some of these studies have methodological flaws, and interpretation of the results is difficult. Recently a study with a more sophisticated methodology evaluated the same phytotherapy combination for prevention of acute respiratory tract infections in patients with a medical history of at least two such episodes in the last cold season (Fintelmann 2012). Compared with placebo the preparation was reported to be associated with a lower infection rate after 84 days, but the study was sponsored by the manufacturer and therefore the results have to be replicated by independent researchers.

For acute rhinosinusitis a study reported a higher rate of complete remission with a complex homeopathic medication versus placebo (Zabolotnyi 2007), but no differences in paracetamol concomitant use in the two arms. Since the study showed a large difference in drop‐outs in the two arms, the last observation carried forward analysis approach may have overestimated symptom resolution in the active treatment arm. In a systematic review, oral glucocorticoids, given as adjunctive therapy to oral antibiotics, potential effects on short‐term relief of symptoms in acute sinusitis were found, but the data are limited and of minor quality (Venekamp 2011). A subsequent randomized study by the same author indicates that the anti‐inflammatory effect of monotherapy with oral corticosteroids seems to be of no benefit in clinically diagnosed acute rhinosinusitis (Venekamp 2012). Antibiotics have no role in uncomplicated acute rhinosinusitis, due to lack of major benefits (Ah‐See 2007; Lemiengre 2012; Masood 2007; Williamson 2007).

To date, two other systematic reviews of P. sidoides extract have been published (Agbabiaka 2008; Ulbricht 2010). Agbabiaka et al reviewed the available data for acute bronchitis and came to conclusions similar to those presented here. The systematic review included six randomized controlled trials in adult patients with acute bronchitis. Two of the studies were unpublished at that time. The studies included by Agbabiaka were also reviewed in the present report. All but two studies (Blochin 1999; Matthys 2003) were eligible for the present review under the terms of our methodological requirements. In Ulbricht et al unpublished data were not reported. The authors included the results from blinded and unblinded observational study reports. They summarized that on basis of these data the positive evidence for the effects of P. sidoides in acute bronchitis is highly suggestive, albeit not definitive.

In the present analysis, adverse events were slightly more frequent with P. sidoides. There have been case reports of allergic reactions, liver toxicity and hemorrhage after treatment with the commercially available preparation (de Boer 2007). Recently, the Drug Commission of the German Medical Association (Arzneimittelkommission der deutschen Ärzteschaft AKDÄ) announced a possible relation between the consumption of Umckaloabo®, the proprietary medicinal product of P. sidoides on the German market, and the occurrence of hepatitis (AKDÄ 2011). On the basis of 19 case reports the commission concluded that a causal relationship between Umckaloabo® and very rarely occurring hepatitis cannot be ruled out. In contrast, the European Medicines Agency (EMA 2012) refers to a re‐evaluation (Teschke 2012a; Teschke 2012b) of all available data concerning the hepatotoxic effects of P. sidoides. This assessment showed a lack of convincing evidence for a hepatotoxic risk associated with the treatment of P. sidoides when used as recommended. In none of the 15 analyzed cases could Pelargonium hepatotoxicity be confirmed as the final diagnosis.

A more recent report summarized the adverse events data of herbal medicines, published in systematic reviews (Posadzki 2013). The authors defined three classes of phytotherapeutics: one class comprises plant extracts, such as Piper methysticum and Cassia senna, which can cause serious side effects. The second class consists of plant extracts such as Aloe vera,Viscum album and Cimicifuga racemosa, with moderately severe side effects. The third class comprises phytotherapeutics with minor side effects such as Crataegus ssp., thyme and Echinacea spp. P. sidoides was assigned to the second class with moderate side effects, such as severe nausea or pruritus. Hepatotoxic effects of P. sidoides were not reported in this review.