Effects of Acupuncture on Sensory Perception: A Systematic Review and Meta-Analysis

Background The effect of acupuncture on sensory perception has never been systematically reviewed; although, studies on acupuncture mechanisms are frequently based on the idea that changes in sensory thresholds reflect its effect on the nervous system. Methods Pubmed, EMBASE and Scopus were screened for studies investigating the effect of acupuncture on thermal or mechanical detection or pain thresholds in humans published in English or German. A meta-analysis of high quality studies was performed. Results Out of 3007 identified articles 85 were included. Sixty five studies showed that acupuncture affects at least one sensory threshold. Most studies assessed the pressure pain threshold of which 80% reported an increase after acupuncture. Significant short- and long-term effects on the pressure pain threshold in pain patients were revealed by two meta-analyses including four and two high quality studies, respectively. In over 60% of studies, acupuncture reduced sensitivity to noxious thermal stimuli, but measuring methods might influence results. Few but consistent data indicate that acupuncture reduces pin-prick like pain but not mechanical detection. Results on thermal detection are heterogeneous. Sensory threshold changes were equally frequent reported after manual acupuncture as after electroacupuncture. Among 48 sham-controlled studies, 25 showed stronger effects on sensory thresholds through verum than through sham acupuncture, but in 9 studies significant threshold changes were also observed after sham acupuncture. Overall, there is a lack of high quality acupuncture studies applying comprehensive assessments of sensory perception. Conclusions Our findings indicate that acupuncture affects sensory perception. Results are most compelling for the pressure pain threshold, especially in pain conditions associated with tenderness. Sham acupuncture can also cause such effects. Future studies should incorporate comprehensive, standardized assessments of sensory profiles in order to fully characterize its effect on sensory perception and to explore the predictive value of sensory profiles for the effectiveness of acupuncture.


Introduction
Acupuncture is gaining popularity as a non-pharmacological option in pain medicine [1,2]. There is substantial evidence for acupuncture being effective in the treatment of acute [3][4][5] and chronic pain [6]. However, for several other pain conditions such as neuropathic pain [7] or fibromyalgia [8], evidence remains inconclusive. In order to specify indications for which acupuncture should be used and to optimize treatment, it is crucial to understand how the effect of acupuncture is mediated.
Various mechanisms underlying the effect of acupuncture have been suggested. Brain imaging studies have shown that acupuncture alters activation patterns in brain areas associated with pain processing [9]. It is postulated that in response to the needle stimulation mechanisms of the endogenous pain modulation such as diffuse noxious inhibitory controls (DNIC), segmental inhibition, and descending pain control pathways lead to a decrease in pain perception [10,11]. At this, various centrally and/or peripherally acting neuromodulators and neurotransmitters such as endorphins [12], serotonin [13], ATP [14], etc. have been identified to play an important role in the analgesic effect of acupuncture. In summary, one can assume that a modulation of the nervous system forms a central part of the effect of acupuncture although details are far from being understood. In special, effects on afferent nerve fibers which might be critical to the modulation of sensory perception by acupuncture remains unclear.
For investigating how acupuncture operates through the nervous system assessments of sensory threshold changes are essential. Evaluations of sensory detection and pain thresholds is referred to as Quantitative Sensory Testing (QST) and has been recognized an important tool in basic science, clinical trials, and for diagnostic and monitoring purposes [15]. QST is deemed to allow for inferences about the type of nerve fibers and about the structures of the nervous system that are affected by a disease or an intervention, according to which modality of sensory perception is changed and at which body sites these changes occur [16][17][18].
Despite the extensive use of sensory threshold assessment in acupuncture research, the impact of these data on the understanding of how acupuncture acts on the nervous system has never been systematically analyzed. Yet, there is no consensus about which modalities of sensory perception (thermal and/or mechanical thresholds, detection and/or pain thresholds) are affected by acupuncture, and whether this effect is influenced by other factors e.g. the measurement tool, the type of stimulation or the target population. The aim of this systematic review, therefore, is to give an overview about data available on the effect of acupuncture on sensory thresholds and to substantiate the respective findings by meta-analyses of high quality studies. Our work provides the first summary of knowledge about how sensory perception is modulated by acupuncture which is crucial to approach a better comprehension of its mechanisms and to improve treatment.

Selection Criteria
We included research articles published in English or German which describe the effect of manual acupuncture (MA; needle insertion with or without manipulation), electroacupuncture (EA; needle insertion with electrical stimulation), or dry needling (DN; needle insertion into myofascial triggerpoints (MTrPs)) on thermal or mechanical detection or pain thresholds in humans. Animal studies, studies using other types of pain paradigms (e.g. electrical or ischemic pain), and studies investigating acupuncture related techniques (e.g. laser stimulation or transcutaneous electrical nerve stimulation) were excluded.

Article Selection
In a first screening step, two reviewers (PIB and JF or PIB and ST respectively) independently assessed the articles retrieved from the literature search for relevance by title and abstract. Full texts of all remaining articles were obtained and screened for eligibility by two independent reviewers (PIB and JF). When the reviewers disagreed or had doubts, the full-text paper was evaluated by all authors. Remaining disagreements were resolved via discussion and consensus.

Quality Assessment
Two independent researchers (PIB, JF) evaluated the quality of the included studies by assessing the risk of bias by means of the Cochrane Collaboration's tool [20] and the quality of reporting of interventions and assessments of sensory thresholds based on the Revised Standards for Reporting Interventions in Clinical Trials of Acupuncture (STRICTA) [21]. In brief, risk of bias was assessed by answering questions about the following features of the studies with 'Yes' (low risk of bias), 'No' (high risk of bias) or 'Unclear' (lack of information or uncertainty over the potential for bias): random sequence generation, allocation concealment, blinding of participants, blinding of outcome assessment, incomplete outcome data, selective reporting and other bias. When sham acupuncture interventions were used as controls we assumed that participants were blinded for group allocation. Blinding of therapists was excluded from the risk of bias assessment since it is not possible to keep the acupuncturist unaware of the point location, stimulation and type of needle. Prior to analysis possible sources of 'other bias' were determined by consensus of the authors. This included bias due to a short washout phase in cross-over studies, questionable outcome assessment, and large baseline differences which were not taken into account in the subsequent analysis. Studies lacking a control group were a priori rated as having a high risk of bias. In addition, we report the total sample size as well as the number of study participants in each group.

Data Extraction
Articles were analyzed by two independent researchers (PIB, JF). Information from the included articles was extracted and tabulated.
Eligible studies are described according to the sensory threshold under investigation, the type of needle stimulation, the characteristics of the study population, and whether the immediate or long term effect of acupuncture was studied. The outcome of the studies is rated as positive or negative according to the authors' conclusion regarding pre-post treatment effects or group differences. Articles with elusive data presentation were rated as unclear. In addition, we compared the effects of verum acupuncture to the effects of inactive or shamcontrol procedures as well as local to distant needling effects (homo-to heterosegmental, ipsi-to contralateral).
A meta-analysis was conducted in order to compare verum and sham acupuncture in high quality studies. Studies were eligible for meta-analysis if they fulfilled the following criteria: no rating of 'high risk of bias' in none of the items of the Cochrane risk of bias tool, sham-controlled, blinding of the outcome assessment, and clear reporting of data. All studies fulfilling these criteria were grouped according to reviewers' opinion about clinical homogeneity. We were able to combine studies in which the pressure pain threshold (PPT; kPa) had been assessed before and after a series of acupuncture treatments in patients suffering from musculoskeletal pain. One of the selected studies followed a cross over design [22]. Accordingly, we only included data obtained at baseline and after the first treatment phase. In cases of multiple evaluations of the outcome at one time point, e.g. several measure sites, data were averaged in order to achieve an equal weighting of all studies for the analysis. In order to account for baseline differences, we used delta scores (post-treatment values minus pe-treatment values) for all calculations. A conservative estimator of the delta score variance was obtained according to the variance sum law without correcting for dependent samples. Meta-analytical comparison of effects of verum and sham acupuncture on the PPT was performed by using the package 'metafor' from the R-project (Version 2.15.1, www.metafor-project.org). The standardized mean difference (SMD) was calculated by dividing the delta scores of the verum and sham acupuncture group by the pooled standard deviation of the two groups. Cochran's Q test was applied to evaluate statistical heterogeneity (I 2 ). We regarded heterogeneity between studies as substantial if Tau 2 was greater than zero and either I 2 was greater than 50% or the Cochran's Q test resulted in a low P value (less than 0.10). An assessment of reporting biases did not appear meaningful due to the small number of studies included in the meta-analyses.

General Aspects
By electronic literature search we identified 3007 citations of which 2922 were excluded; 2830 were screened by title or abstract in a first selection step, and full text was obtained of 177 articles ( fig. 1). Eighty five articles published between 1974 and 2012 met our inclusion criteria (table 1 & 2). More than half of these articles (50 out of 85, 58.8%) were issued after 1999.
Most of the studies were performed in the US and Europe (49 out of 85, 57.6%) and were published in English (81 out of 85, 95.3%). The majority (77 out of 85, 90.6%) describe the effect of acupuncture on a single sensory threshold , whereas eight studies (9.4%) assessed acupuncture evoked changes in more than one sensory threshold [99][100][101][102][103][104][105][106]. About half of the studies (44 out of 85) were conducted with healthy volunteers and evaluated the immediate effect of one acupuncture session (table 1). Forty-one studies (48.2%) included subjects suffering from pain conditions and assessed either the effect of one single acupuncture treatment (13 out of 41; 31.7%) or a series of treatments (29 out of 41; 70.7%; table 2).
In 65 studies (76.5%) an effect of acupuncture on at least one sensory threshold was observed. A statistically significant effect of acupuncture was found in 60 studies (70.6%). In contrast, 15 studies (17.6%) showed no effect on any sensory  threshold. In five of the included publications (5.9%), authors report a pronounced threshold change after acupuncture, but no statistical analysis was provided. Still, we classified the outcome of these studies as positive, in terms of an effect of acupuncture, on the basis of individual reasoning. The overall review results were not affected by these studies and the individual reasoning will be stated in the respective chapter [53,54,74,81,94]. The results of another five studies (5.9%) were rated as unclear due to poor reporting [35,55,106], or because only a pooled analysis of all study arms was reported (table3). Forty studies (47.1%) investigated the effect of EA, 41 studies (48.2%) the effect of MA. Both were equally frequent shown to be effective in changing sensory thresholds (EA 32 out of 40, 80.0%; MA 30 out of 41, 73.2%; Chi-squared-test n.s.). In three studies effects of EA tended to be larger than those of MA [60,76,100]. DN was evaluated only with regard to its effect on the PPT, which was found to be elevated in six out of seven studies (8.2%). In four studies (4.7%) different types of needle stimulation were combined, with two showing a significant effect and two showing no effect. In six studies (7.1%) acupuncture was combined with another intervention. Among these, five resulted in altered sensory thresholds.
Results of the risk of bias assessment are presented in table 4 and in summary in fig. 2. Only six studies (7.1%) were rated as 'low risk of bias' in six or more items of the Cochrane risk of bias tool [33,37,40,47,49,62]. A judgment as 'high risk of bias' resulted most frequently from methodological shortcomings within the      randomization process and the choice of comparison groups not allowing for blinding of participants (8.2% random sequence generation, 14.1% allocation concealment, 11.8% blinding of participants). Blinding of the outcome assessment was assured in less than half of the included studies (42.4%). The overall large proportions of studies with 'unclear risk of bias' reflects the low quality of reporting. In contrary, overall reporting of assessments of sensory thresholds was good and treatment regimens were reported in a mediocre manner (table 4). In 23 of the included publications (27.1%), information was missing for three or more items of the STRICTA guidelines [21]. About a quarter of all studies (24.7%) met all of the nine treatment related items. Threshold measurements were well described in 62 studies (72.9%), but especially combinations of different threshold assessments lacked standardization. Among the eight studies (9.4%), that evaluated acupuncture evoked changes in more than one modality of sensory perception, there was a single study which applied a comprehensive and standardized test battery for sensory testing [100].

Heat Pain Threshold (HPT)
All 26 studies (30.6%) that assessed whether acupuncture had an effect on the HPT evaluated immediate treatment effects. Nineteen of these studies (73.1%) showed that HPT was elevated through acupuncture. In five articles (19.2%), no statistical analysis was provided. Nevertheless, in three cases we agree with a positive rating of the results because of the following reasons. In two studies case numbers were too low to allow for statistical calculations, but HPT elevations following acupuncture were observed in all subjects [53,54]. In another study, two thirds of all subjects (6 out of 9, 66.6%) showed an elevated HPT after EA, but not after stimulation at non-acupuncture points [94]. In contrast, the results of two studies which claim to indicate an effect of acupuncture on the HPT were rated as unclear, because mean HPT changes were not calculated [55], or data were depicted in a graph only [35].
Price et al. assessed low back pain patients and found a significant effect of EA on the HPT [75]. Lundeberg et al., however, did not find any effect of MA or EA on the HPT neither in patients suffering from sinus pain nor in healthy volunteers [104].
The methods used to determine the HPT varied largely, and some studies evaluated several measures. In 16 out of 26 studies (61.5%), the HPT was defined as the time period that subjects were able to tolerate a defined heat stimulus. Eleven of these studies showed a positive outcome. Defined heat stimuli were produced by a projection lamp held on blackened skin [30, 34-36, 42, 53-55, 66, 69, 85, 90], by a hot plate [32,67], or by a thermode containing a peltier element [31]. In one study no information was given about the method used for the application of heat stimuli [94]. In six studies (23.1%) the HPT was defined as the temperature that was first experienced as painful when increasing heat was applied with a thermal sensory analyzer [99-102, 104, 105]. Results of three of these studies (50%) indicated an effect of acupuncture on the HPT [100,102,105]. Six of the 26 studies (23.1%) assessed the pain intensity evoked by a heat stimuli applied with a thermode [60,61,75,80,101,102]. All six found a reduction of the pain intensity after acupuncture.

Cold Pain Threshold (CPT)
Of the 85 included studies, 17 (20%) addressed the effect of acupuncture on the CPT. In 11 of these 17 studies (64.7%), subjects were less sensitive to cold pain after than before acupuncture. All studies focused on the immediate effect after one single acupuncture treatment. Sixteen studies (94.1%) were conducted with healthy volunteers and one with chronic pain patients. When assessing healthy volunteers, eight studies (50%) showed a significant change of the CPT through EA [25,57,58,71,84,92,97,105], while four studies (25%) found no effect of EA [56,59,101,102]. Three studies (18.8%) conducted with healthy volunteers investigated the effect of MA. Two of these showed a desensitizing effect [24,26] and one no effect of acupuncture [99]. Lang et al. assessed both, MA and EA, and did not find any effect on the CPT [100]. Ulett et al. showed that sensitivity to cold pain of chronic pain patients was decreased after EA [89].
Different methods were used for the CPT measurements. In ten studies (58.8%), the CPT was assessed by immersion of the subjects' fingers or forearm into ice-water and documenting either the time until subjects withdrew their limb [24,26,71] or the pain intensity that was experienced during a defined period of immersion into ice water [25, 56-59, 84, 89]. Eight of these ten studies (80%) showed a significant reduction of the CPT after acupuncture. In one study, contact to an ice cold cylinder was tolerated longer after than before acupuncture [105]. In six studies (35.3%), cold stimuli were applied with a thermode including a peltier element and either the intensity of the pain evoked by defined cold stimuli [97] or the temperature that was considered painful [92,[99][100][101][102] was evaluated. Four of the latter showed no changes of the CPT [99][100][101][102].

Warm Detection Threshold (WDT)
The effect of acupuncture on the WDT was investigated in six studies (7.1%) with five evaluating the immediate effect in healthy subjects. Among these, significant changes of the WDT were shown in one study after MA [29] and in two studies after EA [101,102]. No effect of acupuncture in healthy volunteers was reported in two studies, one on MA [99] and one assessing both MA and EA [100]. Ahn et al. included patients suffering from painful diabetic neuropathy (PDN) and measured the WDT before and after a series of treatments [106]. The results of this study are rated as unclear due to the lack of a statistical analysis and the fact that data are depicted in a graph only.

Cold Detection Threshold (CDT)
Five of the six studies that investigated the effect of acupuncture on the WDT also assessed changes in the CDT after acupuncture. Correspondingly, results of one study assessing the CDT in PDN patients were rated as unclear [106]. Of the four remaining studies, three found no change of the CDT after MA [99,100] or EA [101], and two found a decrease of the CDT after EA [100,102].
All but one study, for which the measuring tool was not described, used a thermode including a peltier element to assess the WDT and the CDT.

Thermal Sensory Limen (TSL)
Three of the included acupuncture studies (3.5%) used the TSL as an outcome measure. All three investigated immediate treatment effects. In two studies by Lundeberg et al., the TSL was found unchanged after EA as well as after MA in healthy volunteers and in patients suffering from sinus pain [103,104]. In contrast, Lang et al. found a significant elevation of the TSL of up to 3˚C through EA but not through MA in healthy volunteers [100]. In all studies, TSL assessments were performed by using a sensory thermal analyzer. The thermode applied on the subjects' skin increases or decreases its temperature. As soon as the subject indicates the feeling of cool or warm, a switch towards the opposite direction of the temperature change is induced.

Pressure Pain Threshold (PPT)
Almost half of the articles (42 out of 85, 49.4%) included in this review describe changes of the PPT through acupuncture.

PPT assessed in Healthy Subjects
Seven of these 42 studies (16.7%) evaluated the immediate effect of acupuncture on the PPT in healthy subjects. All showed a statistically significant elevation of the PPT through acupuncture. Two studies assessed the effect of EA only [28,39], three studies solely the effect of MA [65,95,98], and two the effect of both, MA and EA . Either a manual algometer [39,65,95,100] according to Fischer et al. [107] or an electronic algometer [28,76,98] was used as measuring tool (tip size 1 cm 2 ). Changes ranged from 59 to 392 kPa [28,39,76,98,100] or from 10% to 27%, respectively [65,95].
Ten of the 35 patient studies (28.6%) evaluated the immediate effect of one single acupuncture treatment. Nine of these reported the PPT to be significantly elevated [33,40,41,43,51,64,78,79,83], and one study found no significant PPT increase [50]. Among the 26 studies (74.3%) that assessed the effectiveness of an acupuncture series, 18 (69.2%) showed a PPT increase after treatment [37,38,46,50,52,62,63,68,70,72,74,77,81,82,86,87,91,93] while six studies (23.1%) observed no effect of acupuncture on the PPT [22,27,[47][48][49]96]. The outcome of two studies was rated as unclear, because the analysis was performed by combining data of all groups [44,45]. Although no statistics were provided, the outcome of two other studies was rated as positive. Singh et al. found that, after a MA treatment of two months, the majority of fibromyalgia patients felt less pain, although greater pressure than at baseline was applied [81]. Perez-Palomarez et al. included a large number of patients and showed a prominent mean increase of the PPT (74.5 kPa-202.0 kPa) after DN and EA [74].
The effect of EA on the PPT was evaluated in four out of the 35 studies (11.4%) conducted with patients [37,64,74,93]. All four showed an increase of the PPT after EA. MA was investigated in 22 studies (62.9%) of which 17 (77.3%) revealed an increase of the PPT . In two of these studies, MA was applied as an add-on treatment to stretching [70] or tricyclic antidepressants and exercise [87] for myofascial pain or fibromyalgia, respectively. Seven of the 35 studies (20.0%) investigated the effects of DN in the treatment of myofascial pain. DN was shown to be effective in increasing the PPT in six studies [33,38,40,41,74,83], but one study found no effect of DN on the PPT at MTrPs of the neck [48]. Four studies used a combination of either EA and MA [22,46,68] or of MA and DN [49]. Two of these studies showed a change of the PPT [46,68] and two no treatment effect [22,49]. The observed PPT changes varied largely between studies (22.5 kPa to 245.2 kPa) and tended to increase during follow up if assessed.
Three studies also assessed the pressure pain tolerance (PPTo) which was defined as the time subjects tolerated painful pressure [45,64,74]. Two studies showed the PPTo to be elevated after EA [64] and DN [74], respectively. Results of one study on MA were rated as unclear [44].

Meta-Analysis
Considering the previously defined quality criteria and clinical homogeneity, only studies investigating the effect of acupuncture on the PPT were eligible for metaanalyses. Four studies could be combined in an analysis regarding the immediate treatment effect [22,37,49,93], and two studies were analyzed regarding longterm effects of acupuncture in pain conditions [49,52]. The results of these metaanalyses confirm our descriptive findings. The effect of acupuncture directly after a series of treatments ( fig. 3A)

Mechanical Pain Threshold/Sensitivity (MPT/MPS)
Three of the 85 articles (3.5%) included in this review investigated the effect of acupuncture on the MPT and/or the MPS. All three studies were conducted with healthy volunteers assessing the MPT/MPS before and immediately after a single acupuncture treatment. All showed a desensitizing effect of EA as well as of MA [73,100,105]. Just Lang et al., who assessed both, the MPT and the MPS, did not find an effect of MA on the MPT [100]. However, different measuring tools were used; scaled forces applied with pin-pricks [100], increasing pressure applied by a blunt needle [73], or gauged forceps [105].

Mechanical Detection Threshold
We identified two acupuncture studies (2.4%), in which the MDT was one of the outcome measures. Both studies were conducted with healthy volunteers, used Von Frey Filaments in order to evaluate the MDT, and focused on immediate effects of acupuncture; one on EA [102] and one on both EA and MA [100]. Results of these two studies indicate that neither EA nor MA has an impact on the MDT in healthy subjects.

Vibration Detection Threshold (VDT)
Four articles (4.7%) describe the effect of acupuncture on the VDT. In two studies the impact of a MA treatment series on the VDT in patients suffering from diabetic peripheral neuropathy was explored. Tong et al. noted an improvement of the ability to detect vibration [88] while Abuaisha et al. found no changes of the VDT after MA [23,88]. Lang et al. and Lundeberg et al. reported that a single application of neither EA nor MA had an effect on the VDT in healthy volunteers [100,103]. Vibration stimuli were applied with a Rydel-Seiffer tuning fork [100] or an electromechanical device [23,88,103].

Effect of Needle Location relative to the Sites of Measurements
Twenty (23.5%) studies compared ipsi-to contralateral and/or close to distant needling either in healthy subjects (15 out of 20; 75%) or in pain patients (5 out of 20; 25%). Of the ten studies comparing close to distant needling, eight (80%) showed a larger increase of at least one sensory threshold close to the needling location [35,51,53,67,83,101,102,105]. In six out of eleven studies (54.5%), threshold changes were more pronounced after ipsi-than after contralateral needling [25,34,67,[100][101][102]. In one high quality study bilateral needle placement was superior to unilateral needling [65], and results of three studies suggest that needle stimulation at LI 4 is more effective in changing pain thresholds than needling at other acupuncture points [39,53,65]. Four studies (20%) found significant effects of acupuncture on sensory perception independent of the needle location [28,60,66,75,95], and in two studies no change of any sensory threshold was observed [103,104].

Responder versus Non-responders
Six of all included studies (7.1%) distinguished between subjects that responded to acupuncture and those who did not [32,35,42,55,60,94]. All of these studies assessed the HPT in healthy volunteers and all were conducted with relatively small case numbers (11.8¡3.0). Proportions of responders ranged from one third to two thirds. Chae et al. found genetic difference between acupuncture responders and non-responders [32], but their results have not been reproduced by further investigations. Furthermore, the role of hypnotic susceptibility in responsiveness to acupuncture was investigated in three studies, with contradictory results [56,58,89]. The influence of expectancy was found to be substantial in two studies [57,61]. Knox et al. found no effect of EA neither in oriental nor occidental subjects [59].

Result Interpretation
Our results revealed that in 76.5% of 85 eligible studies at least one sensory threshold was changed after acupuncture, indicating an activation of neuromodulatory mechanisms. However, results displayed substantial heterogeneity, which is illustrated for the PPT by results of the meta-analyses ( fig. 3).
Over half of the sham-controlled studies found larger effects in the verum group than in the sham group. However, a quarter of these studies found significant threshold changes also after sham acupuncture, while only one of the included articles reported changes of sensory thresholds after an inert control procedure. These facts go in line with the previously drawn conclusion that there are effects specific to acupuncture, but that sham acupuncture may cause physiological reactions exceeding pure placebo effects [6,108,109]. It can be assumed that this in part explains the clinical effects of sham acupuncture interventions observed in acupuncture randomized controlled trials.
Most studies conducted with patients used the PPT as an outcome measure. This reflects the frequent use of acupuncture in clinical practice for treating pain conditions e.g. musculoskeletal disorders, in which the PPT correlates well with clinical status [110]. More than 80% of the studies -i.e. 27 out of 35 clinical studies and all seven studies conducted with healthy volunteers -showed that acupuncture reduced pain evoked by blunt pressure which is mainly mediated by deep tissue nociceptors (Ad-and C-fibers) [110]. PPT reductions of up to 245.2 kPa as observed in some studies can be interpreted as clinical relevant. In addition, results of two meta-analyses show significant short-and long-term effects of acupuncture on the PPT in pain conditions ( fig. 3). Thus, these findings provide a physiological basis for the growing body of evidence for the effectiveness of acupuncture in locomotor conditions associated with tenderness [111][112][113].
Pain thresholds were elevated after acupuncture also when painful pressure was exerted on a rather small skin area (#1 cm 2 ) and when pin-prick like stimuli (MPS and MPT) were applied. This finding is derived from few studies, but suggests that acupuncture also affects mechanical pain evoked by punctate objects which is primarily mediated by intra-epidermal nociceptors (mainly A-d fibers) [114].
Studies investigating whether thermal pain is reduced through acupuncture are abundant but almost exclusively conducted with healthy volunteers. Results of such investigations are more ambiguous than data of included studies on changes of the PPT, MPT and MPS. After acupuncture, sensitivity to painful heat was reduced in 19 out of 27 studies, while sensitivity to painful cold was reduced in 11 out of 17 experiments. The transmission of heat pain is mainly evoked by C-fiber mechano-heat nociceptors (CMH) responding to heat stimuli ranging from 41˚C to 49˚C [115] and is linked to the capsaicin sensitive vanilloid receptor VR1 which is also found in type II Ad-nociceptors [116]. In contrast, the transmission of cold pain is mediated by both cold sensitive C-and Ad-fibers [117,118] which are insensitive to vanilloid compounds [119]. Our results suggest that both types of nociception are likely to be affected by acupuncture in healthy subjects. Reasons for the heterogeneity of data might partly be explained by methodological issues. Studies in which painful cold was applied through a thermode found changes of the CPT less frequently (two versus four) than other studies (eight versus two), in which e.g. the subject's hand or arm was immersed into ice-water. It is conceivable that differences in study outcome are related to the size of the skin area to which cold stimuli were applied or to the intensity of these stimuli (see 4.2. Limitations for further discussion).
In contrast to pain perception, data on the effect of acupuncture on sensory detection are sparse. The ability to detect temperature changes (WDT, CDT, TSL) was reduced after acupuncture in half of all experiments. Mechanical detection (VDT and MDT), in contrast, was not affected by acupuncture in five out of six trials. These findings provide a first hint that acupuncture might not affect mechanical detection which is mainly mediated by Ab-fiber signaling [120,121], while the influence on thermal detection which is linked to signaling of warm sensitive C-fibers or cold sensitive Ad-and C-fibers [122,123] remains unclear.
Another important finding is that ipsilateral needling and needling close to the measure sites were found to exhibit stronger effects on sensory thresholds than needling at contralateral or remote body sites, respectively. This underscores the importance of local mechanisms such as the release of neuromodulators at the needling site [14] and spinal mechanisms such as segmental inhibition [11]. It is supposed, that activation of A-fiber afferents results in the activation of spinal inhibitory interneurons, those achieving primary analgesia within the same segment [124,125]. Nevertheless, more studies adopting a sophisticated selection of measure sites in combination with imaging studies are needed to clearly differentiate between local, spinal and supraspinal mechanisms of acupuncture. In contrast, it seems to be of minor importance whether the needle is stimulated electrically or manually.

Limitations
The overall poor study quality ( fig. 2) and the consequently low number of studies included in the meta-analyses are the major limitations to our findings. However, limitations resulting from the quality assessment itself need to be taken into account. Many publications were characterized by poor reporting which renders an estimation of the real number of high quality studies/studies with low risk of bias impossible. In particular details of blinding and randomization procedures as well as treatment regimens were often missing. Second, the STRICTA guidelines provide an essential tool for assuring the quality of reporting in acupuncture trials. Nevertheless, it bears the striking disadvantage that the quality of reporting of treatment related items might be underestimated in studies investigating the effectiveness of individualized acupuncture regimens since precise instructions for such studies are missing.
A further limitation comes with the methodology of QST itself. Although when performed with standardized methods, QST involves subjective ratings of perceptions and, therefore, is susceptible to bias due to psychological factors such as expectation and conditioning [126]. These should be assessed and/or controlled for [127]. Comparability of studies was also limited due to methodological variability with regard to treatment, study population, and outcome assessment. Appropriate selection of the acupuncture points needled, number of needles and stimulation technique are, according to the traditional concept of acupuncture, crucial for achieving an optimal treatment effect. These concepts are in part supported by research findings. For example, current evidence suggests an association between the number of needles applied and the clinical outcome [128], but overall there is still no consensus within the scientific community about how to appraise these parameters correctly. The same holds for the socio-cultural background of the study population. Previous acupuncture experience and social valuation of acupuncture is very likely to have an impact on treatment outcome; likewise to any other treatment. Yet, there is very little emphasis on sociocultural aspects in acupuncture research. There were striking differences between acupuncture interventions applied in the included studies and, although not reported in detail, very probably also between the different study populations. However, due to the still poor knowledge about the impact of these factors, there are no guidelines on how to appraise them in a review apart from pure reporting as performed in table 1 and 2.
Additionally, study outcome might in part be influenced by the measuring tool. Investigations comparing different methods evaluating the same sensory thresholds are missing. It also remains unclear to what extent the test stimuli might interfere with the treatment effect. For example, strong noxious stimulation is known to activate aspects of the endogenous pain-modulating network [129][130][131]; a phenomenon known as counter irritation. It is postulated that the main underlying mechanisms namely DNIC is also involved in the analgesic effect of acupuncture [98,[132][133][134]. This is however controverted by recent work and needs to be further explored [98,134].
Furthermore, the interpretation of the clinical relevance of the effect of acupuncture is limited by the fact that studies -besides those assessing the PPTwere almost exclusively conducted with healthy volunteers. There is also limited amount of data encoding the effect of acupuncture on stimuli above pain thresholds, which have been shown to activate other nerve fibers than stimuli around the pain threshold [135].

Future Perspectives
Recently, efforts are increasing to comprehensively characterize painful conditions by means of QST [136][137][138][139]. Attempts are made to classify patients on the basis of symptoms, signs, or patterns of somatosensory abnormalities [139][140][141]. Those might reflect the underlying pathological mechanisms [142] and might, therefore, be related to different treatment responses [137,143]. Such subgroup analysis should also be adopted in acupuncture research since responsiveness to acupuncture varies largely between individuals. This inter-individual variability in response to acupuncture was also demonstrated by studies included in this review, but proposed hypotheses explaining this phenomenon have not been verified. Comprehensive QST assessments applied in large scale studies might provide hints about whether individual patterns of somatosensory alterations might be of predictive value for the effectiveness of acupuncture. It is striking that only one study applied a comprehensive, standardized QST test battery [100].
Comprehensive QST assessments might, furthermore, provide information about the extent to which signs of peripheral and central sensitization may be reduced by acupuncture. For example, it has not yet been subject to clinical trials whether thermal and mechanical hypaesthesia or increased wind-up, which can be associated with chronic pain conditions [144][145][146], might vanish along with clinical improvement after an acupuncture therapy.

Conclusion
This review provides evidence for the effect of acupuncture on sensory perception, especially pain perception. Most compelling evidence supports the reduction of pain evoked by blunt pressure through acupuncture; particularly as a measure of tenderness in pain conditions. Moderate evidence was found that acupuncture reduces the perception of noxious heat or cold. The outcome of these studies seemed to depend on measuring methods. Little but consistent evidence was found that pin-prick like pain (MPT, MPS) is reduced but that mechanical detection (VDT, MDT) is not influenced by acupuncture. No conclusions can be drawn about whether acupuncture affects the ability to detect temperature changes.
Sham-acupuncture approaches can evoke such effects as well. Thus, it is questionable to classify them as pure placebos, a fact that needs to be taken into account when conducting and interpreting acupuncture studies.
Data support the importance of local and spinal mechanisms involved in the neurophysiological effect of acupuncture. More high quality studies are needed to characterize the effect of acupuncture on the whole sensory profile by means of comprehensive QST assessments. In addition, investigating defined patient populations accordingly might also clarify whether certain characteristics of the somatosensory profile are of predictive value for the analgesic effect of acupuncture and to which extent pathologic alterations of sensory perception vanish along with clinical improvement achieved by acupuncture.
Supporting Information S1 Table. PRISMA Checklist. Overview of the reporting items in accordance with the PRISMA statement doi:10.1371/journal.pone.0113731.s001 (DOC)