Over 18 of English and Chinese electronic databases from 1937 to March, 2013, were used and different terms (e.g. Buerger’s exercise, Buerger-Allen exercise) were involved to search for relevant articles (Table 1). We also searched references with relevance in Google Scholar in all fields (no title or abstract restrictions) to filter the retrieved references. This was to ensure the search strategy was wide enough to recover all studies potentially relevant for the review. Studies of reference lists that seemed relevant were assessed based on the information provided in the title. The Cochrane Library search retrieved no systematic reviews.

Two reviewers independently screened the full text of the references selected that could not initially be excluded based on their titles and abstracts. To obtain as many relevant reports as possible, broad inclusion criteria were adopted: 1) Intervention studies, those focused on Buerger’s exercises. 2) Papers published in English and Chinese. The selection was not restricted by methodological quality, population and outcomes. Electronic search results were downloaded into the End Note bibliographic program, from which duplicates were deleted.

One reviewer extracted data from included studies about first author; year of publication; country of study; location; duration of study; sample; population characteristics; variable; outcome measures; quality indicators; interventions; characteristics of interventions and results. The second reviewer checked the accuracy of the extracted data.

Two reviewers independently assessed the eligibility of the study, quality appraisal and carried out the analysis. The quality appraisal of the included studies was assessed using criteria based on the Jadad scale [25] . The quality questions checked were the following: 1) Was the study described as randomized? 2) Was the study described as double blind? 3) Was there a description of withdrawals and dropouts? 4) Was the method to generate the sequence of randomization described and it was appropriate? 5) Was the method of double blinding described and it was appropriate? 6) Was the method to generate the sequence of randomization described and it was inappropriate? 7) Was the study described as double blind but the method of blinding was inappropriate?

The methodological quality of included studies was evaluated by two reviewers. The 1 - 5 questions were scored as “Yes” (1), “No” (0), and 6 - 7 were scored deduct 1 point if as “Yes”. Therefore, each study could score 0 to 5. Disagreements between the two reviewers were resolved by discussed until consensus.

The participants, intervention, study design, outcomes, measures and intervention characteristics were summarized with a narrative and tabular synthesis. The limited data and use of divergent variables within the studies did not allow meta-analysis.

A total of 95 potentially relevant papers were identified, of which 15 articles were identified by the systematic search in electronic database, and 79 papers were selected from the Google Scholar Website. One further study was found from a reference list. There were 86 papers that failed to reach our selection criteria, which left on 9 that could be included in the final review (Figure 1).

Table 2 showed that of the 9 studies reviewed, 6 randomized were controlled trials, two were pre-test studies with control group, and one was a prospective control study. Three of the studies were published before 1996 and 6 after 2005. Six of the studies were conducted in Mainland China, with one study each from Denmark, the United States and Taiwan. Five studies were lead by registered nurses, 3 by medical doctors and one by a physical therapist. Three papers were published in English [23] [26] [27] and 6 in Simplified Chinese [28] - [33] . These studies were conducted in a variety of settings, including vascular clinics, surgical, Chinese medicine, orthopedic, and gynecology units.

The 9 studies involved a total of 592 participants; most of whom were hospitalized patients (66.7%). Four of the studies patients were related to peripheral arterial occlusive disease (PAOD), 4 to orthopedics, and 1 to a gynecology problem. Two of the studies included diabetic patients [23] [26] . Sample size ranged from 14 [26] to 233 [31] . All studies did not calculate the sample size (power) required to see an effect. Seven studies gave the range age of participants from 10 to 83 years. Five studies recruited both male and female participants, but 2 did not mention gender. Only 2 studies mentioned informed consent and one study reported a drop-out rate [23] [26] .

Measurement variances of the selected studies were to assess Buerger’s exercises with regard to several physiological outcomes or clinical symptoms. The variables of physiological indicators included Na [24] clearance rate (Geiger counter); ¹³³Xe washout rate (NaI scintillation detector); blood flux (DRT4 Laser Doppler); walking time (treadmill); skin temperature (DRT4 Laser Doppler); hemorheology (packed cell volume, fibrinogen); and bed-rest times (days). Clinical symptoms that were measured included foot necrosis (years); degree of swelling; pain (Wong-Baker Faces Pain Rating Scale, Numeric Pain Rating Scale), and degree of cyanosis (Table 2). Most of the select studies did not investigate or report on the reliability, stability and validity of their instruments. Even when using the observational study design, 5 studies did not include an operational definition, observer training and inter-rate reliability [28] - [31] [33] .

All the studies had 2 intervention and comparison groups. Six of the 9 studies assessed the benefit of Buerger’s exercises compared with usual care, and one compared them with arterial reconstruction [26] . Two compared the same intervention, but used different participant characteristics (Table 3) [23] [27] . Two studies had certified

^*HBV = high blood viscosity, LBV = low blood viscosity, PCV = packed cell volume, EG = experimental group, CG = control group, NS = not significant.

nurse as intervener, one a physiotherapist. Most of the studies did not mention the background of interveners (Table 3).

The intervention time (period) of the 4 selected studies ranged from 3 - 14 days, 3 ranging from 28 days to one month. The 9 studies contained a divergent range of exercising programs, such as intensity, duration, frequency and procedure. Most of the exercises were undertaken each day (n = 6), the intensity ranging from 2 - 4 sections a day, with each section repeated 3 - 10 times. Regarding the duration for each step, 6 studies used 1 -3 minutes and 3 used 1 - 5 minutes. Most of the procedures of Buerger’s exercises were inconsistent, e.g., the angle of legs, duration of each step and each section.

Most of the studies (N = 8) reported a positive effect of Buerger’s exercises. Only one study showed no significant effect on PAOD patients [26] . More than half of the studies indicated significantly improved blood flow of the lower extremities and walking ability (p < 0.05), decreased hemorheology indices after operation (p < 0.05), reduced bed-rest times (p < 0.05). Six studies indicated that the exercises significantly reducing swelling, cyanosis and pain (p < 0.001) [29] [30] [33] , reduced venous embolism (p < 0.001) [31] [32] , and delayed the occurrence of foot ulcers and gangrene [28] . There was no significant effect on skin temperature [23] .

Total Jadad quality score among the 9 studies ranged from 0 to 2 (Table 4). They scored less than or equal to 2 points, including 4 studies of 0 point, 4 studies of 1 point and 1 study of 2. While 6 studies used randomized

Note. N/A = not applicable, N/R = not reported.

controlled trials, only 2 gave their method of randomization. One of these studies used table random numbers generated by a Stochastic [32] . One enrolled subjects according to their admission order, and was deducted 1 point. One study had withdrawal numbers, but no reason was given.

Despite some limitations in this study, we were able to make decisions about the strength and consistency of the evidence. To the best of our knowledge, our review is the first attempt to synthesize the evidence that Buerger’s exercises improve peripheral circulation. The results may be biased by only screening publications in the English or Chinese; thus, important studies might have been missed. The reviewed studies varied in participants, interventions, samples, methods, measures, and outcomes, giving a heterogeneity that disallowed meta-analysis, making comparisons between them difficult.

Some literatures indicate that many diabetic patients apparently had foot ulcer before their lower extremities were amputated [5] [35] . Hence, providing further diabetic care with Buerger’s exercises would be beneficial for diabetic patients who have peripheral circulation insufficiency or peripheral neurovasculopathy, e.g. ankle brachial pressure index < 0.9 or Michigan Neuropath Screening Index > 2. Therefore, it is necessary to investigate the effect of Buerger’s exercises on diabetic-related peripheral disorder through evidence-based studies. Regarding the ethical consideration, only 2 studies mentioned informed consent [23] [26] . Notably, in more than half of the selected studies, the primary researchers and background of interveners were nurses, which implied that Buerger’s exercises can be an important independent role for nurses to take care of hospitalized and community clients with lower extremity circulation disorders. However, the effect of Buerger’s exercises and the standardization of the procedures involved ought to be subject to further more rigorous scientific researches.