Journal of Musculoskeletal Disorders and Treatment
The Effects of Whole-Body Vibration Training for Rehabilitation after ACL Reconstruction: A Systematic Review
Y. Osawa1,2*, Y. Oguma2,3, T. Hashimoto2 and Naokata Ishii4
1Center for Supercentenarian Medical Research, Keio University School of Medicine, Japan
2Sports Medicine Research Center, Keio University, Japan
3Graduate School of Health Management, Keio University, Japan
4Graduate School of Arts and Sciences, The University of Tokyo, Japan
*Corresponding author: Y. Osawa, Center for Supercentenarian Medical Research, Sports Medicine Research Center, Keio University School of Medicine, 35 Shinanomachi, Shinjuku-ku, Tokyo, 160-8582 Japan, Tel: 813-3353-1211, E-mail: email@example.com
J Musculoskelet Disord Treat, JMDT-1-001, (Volume 1, Issue 1), Review Article
Received: August 07, 2015: Accepted: September 18, 2015: Published: September 21, 2015
Citation: Osawa Y, Oguma Y, Hashimoto T, Ishii N (2015) The Effects of Whole-Body Vibration Training for Rehabilitation After ACL Reconstruction: A Systematic Review. J Musculoskelet Disord Treat 1:001
Copyright: © 2015 Osawa Y. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
We investigated the effects of short-term WBVT program for rehabilitation after ACL reconstruction on postural control, knee position sense, and knee laxity. This review was restricted to randomized controlled trials, which investigated the rehabilitation effects of WBVT on physical function compared with conventional rehabilitation program. Data from a total of 99 participants with ACL reconstruction in 3 studies were included. Articles with high risk of bias were included based on the criteria of Cochrane Review Guideline. Furthermore, based on the International Society of Musculoskeletal and Neuronal Interactions recommendation, the lack of information on the WBV settings was found in them. Meta-analysis showed that, after WBVT intervention, no significant standardized mean difference (SMD) was observed in postural stability (n = 3; -0.58[-1.32-0.16], p = 0.13) with large heterogeneity (I2 = 67%) and knee laxity (n = 2; -0.10[-0.50-0.70], p = 0.74) with no heterogeneity (I2 = 0%), whereas a significant SMD was observed in position sense (n = 2; SMD = -1.44[-2.22- -0.67],p = 0.0003) with large heterogeneity (I2 = 85%). We concluded that WBVT might be a promising alternative exercise therapy for rehabilitation after ACL reconstruction. However, to clarify the effects of WBVT on prevention from re-injury after returning to sports activities, future studies with longer term WBVT with better quality are strongly needed.
Vibration, Exercise Therapy, Balance, Proprioception, Review
Anterior cruciate ligament (ACL) injuries frequently occur during sports activities . After the ACL reconstruction, rehabilitation programs play a key role in returning to sports. Although, to date, optimal rehabilitation program is not established, the review by van Grinsven et al.  indicated that the time to returning to sports would depend on the rehabilitation program.
Whole-body vibration training (WBVT) has recently been introduced in fitness clubs, clinics and professional sports teams as an alternative or supplementary to conventional training. While a vibration source can be directly applied to the muscle belly or tendon of target muscles in locally applied vibration, vibration can be indirectly applied to the target muscles by first transmitting vibration through a body part in WBVT. Because quite a few studies have widely investigated the effects of WBVT on physical function or morphological characteristics in healthy subjects including older individuals without any serious side effects, the effects of WBVT as an exercise therapy on physical function have more recently been investigated in patients with various diseases such as Parkinson disease, post stroke, cerebral palsy, muscular dystrophy, and myofibrosis [3-5].
WBVT has four unique characteristics. First, physically, the force generated by a vibration platform becomes workload in an exercise program with WBV . Second, during exposure to vibration, the Ia afferent neural pathways from muscle spindles are more sensitive to vibration than other afferent neurons, such as group II afferent neurons and Golgi tendon organs . The vibration-induced stretch reflex, which proceeds by way of mono- and poly-synaptic pathways, is termed the 'tonic vibration reflex' [8,9]. As an acute effect of vibration, continually activated group Ia afferent neural pathways of muscle spindles together with activation of the γ-loop (known as α-γ co-activation) . Third, WBVT can be a time-efficient training modality to improve muscle fitness compared with conventional exercise modalities (e.g., resistance training, aerobic exercise, and balance training) . Last, WBVT may have an act on relieving musculoskeletal pain [12,13].
Taken together, it can be expected that exercise therapy comprising of WBVT would be suitable for rehabilitation programs after ACL reconstruction to improve physical function. However, because WBVT is relatively new exercise for rehabilitation, it is still controversial whether using WBV in rehabilitation program would be meaningful for physical function enhancements compared with conventional exercise therapy after ACL reconstruction. Here, to investigate the effects of WBVT-based rehabilitation program on physical function compared with conventional exercise therapy after ACL reconstruction, we systematically reviewed recently published reports on short-term effects of WBVT on physical function.
Literature search strategy
Electrical databases of MEDLINE (PubMed), EBSCO (CINAHL Plus with Full Text), PEDro, and Web of ScienceSM were accessed online in December 2014 and searched using the following key words: ("whole body vibration" OR "vibration exercise" OR "vibration training" OR "vibration therapy") AND (anterior cruciate ligament [MeSH]).References lists of potentially useful articles were also scanned for additional articles.If the study title was related to WBVT, the article was selected as the first selection round. In the second selection round, we read the full articles.
Eligibility criteria: The eligibility criteria for this review were: (a) the participants were individuals experienced ACL reconstruction, (b) a randomized controlled trial (RCT) that had at leastone exercise group with WBV group, (c) intervention was comprised of several exercise sessions with at least 4-week , and (d) any outcome measurements related to physical function.
Exclusion criteria: The exclusion criteria for this meta-analysis were: (a) locally applied vibration, (b) bed-rest studies, (c) clinical controlled trials, (d) case-control studies, (e) proceedings, (f) animal studies, and (g) when double (or triple) publications of single trial were identified.
Assessment of methodological quality
Methodological quality was assessed based on the guidelines for systematic reviews established by the Methodological Guidelines Cochrane Review . Briefly, risk of bias was evaluated based on responses to seven questions inquiring about the random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other bias. In 'other bias', we mainly assessed the risk of bias, regarding compliance of exercise intervention, co-intervention, and baseline characteristics of each study group. These seven criteria were scored with 'yes', 'no', or 'unsure' based on the criteria indicated by the Cochrane Handbook for Systematic Reviews of Interventions .
In addition, we also evaluated the quality of each study based on the recommendation of the International Society of Musculoskeletal and Neuronal Interactions (ISMNI) for reporting WBV intervention studies, consisting of 13 factors . Briefly, we evaluated whether each article adequately described the WBV-related factors based on responses to 13 questions inquiring about the WBV parameters (e.g. frequency, peak-to-peak displacement, and acceleration) and participants' position (e.g. holding bar, exercise position, and foot wear condition). Whether the article adequately described each of the above was scored with 'yes', 'no', or 'unsure'. 'Vibration displacement' was scored as 'unsure' if it was unclear whether the described displacement was peak-to-peak, and if the landmark of foot position was unclarified in case of side-alternating WBV. If we knew bar holding and foot-wear condition by figures, we scored these with 'yes'.
Participant and disease-specific characteristics (age, body weight, sex, and athletic performance level, ACL reconstruction method, and the severity of injury), WBV parameters (frequency, peak-to-peak displacement, and if applicable, accelerations), exercise program, and outcomes were extracted.
Standardized mean difference (group contrasts) was calculated using the Review Manager version 5.3.5 (Copenhagen, Nordic Cochrane Center, The Cochrane Collaboration, 2014). Intervention effects were calculated as 'post-trial mean minus pre-trial mean' for each intervention group. The standard deviation of the difference scores from the standard deviation of each intervention group was calculated using the following equation:
Where N represents the number of participants. We then calculated standardized mean differences (SMD = the Hedges' correction g). In addition, to clarify the effects of WBVT on postural stability, position sense, and knee stability in comparison with conventional exercise therapy (CON) group, meta-analysis was performed.
Test for heterogeneity
Heterogeneity among included studies was assessed using the Cochrane Q statistic. P values were obtained by comparing the Q statistic with a χ2 distribution and κ - 1 degrees of freedom, where κ represents the number of studies included. Because heterogeneity is, to a certain degree, inevitable in meta-analysis, particularly for exercise trials, we reported the I2 statistic using the following equation:
Where Q and df are Cochran's heterogeneity statistic and the degrees of freedom, respectively. I2 = 0%-40% indicates the absence heterogeneity, and I2 = 30%-60%, I2 = 50%-90%, and I2 = 75%-100% indicate the presence of moderate, large and extremely large heterogeneity, respectively15. In this meta-analysis, I2 of > 50% was used as the indication of significant heterogeneity. If significant heterogeneity was observed, a random effect meta-analysis model was applied.
As the first process, 21 potentially relevant references were identified. After reading these articles, 18 articles were excluded in this review as following reasons: review articles (n = 7); animal study (n = 1); case study (n = 1); cross-sectional studies (n = 3); not related articles (n = 4); and not individuals experienced ACL reconstruction (n = 2). Three articles were RCTs which satisfied the eligible criteria, and all were included in this review (Figure 1).
Figure 1: Flow of the meta-analysis. View Figure 1
The methodological quality scores and support information of included trials are shown in Appendix. The overall mean score was 2.0 ± 0.8(range: 1 to 3) of 7 points. The quality scores of each study followed by the ISMNI recommendation and support information are shown in Appendix. The overall mean score were 7.3 ± 1.2 (range: 6 to 9) of 13 points.
Of these studies, two studies adequately described the randomization methods [17,18] while another study did not state the randomization procedure at all . Table 1 shows the study design and WBV parameters in these included studies, respectively.
Table 1: Characteristics of the studies included View Table 1
A total of 99 participants were included in the analysis. The mean age in each study ranged from 22.7 to 28.0 years (Table 1). The physical performance level of the participants in each study was competitive athletes of national or international level , non-competitive athletes , and unclear , respectively.
ACL reconstruction methods of the included studies were as follows: bone-tendon-bone method [17,19]; a single-bundle method . It was unclear the severity of injury, but all participants did not experience any previous or concomitant injuries or surgeries of the knee [17-19].
Figure 2 shows the timing of introduction of WBVT for rehabilitation after ACL reconstruction surgery in each study. During WBVT program, a conventional rehabilitation program was simultaneously conducted in the study by Fu et al. , whereas only WBVT was performed in the other two studies [17,19]. Mean duration of training period and number of total training sessions in all included trials were 7.3 ± 2.5 (range: 4 to 10) weeks and 22.0 ± 11.4 (range: 12 to 38) sessions, respectively. The type of WBV device varied among included studies: synchronous type [18,19]; side-alternating . For progressive overloading, both WBV settings and exercise program (i.e., exercise, the duration for exercise, and the number of the training sets) were changed during the intervention period [17-19]. All participants received physical therapy immediately after ACL reconstruction surgery, and were randomly allocated to either WBVT or CON group.
Figures 2: The timing of introduction of whole-body vibration training for rehabilitation after ACL reconstruction surgery. ACL, anterior cruciate ligament; Ex, exercise therapy; Rehab, rehabilitation; WBVT, whole-body vibration training View Figures 2
Reported side effects
Two of the included studies reported that no severe side effects were observed in WBVT [18,19], whereas Berschin et al. reported that 60% of the participants experienced pain or swelling after WBVT rehabilitation program .
Physical function tests were as follows: postural stability measured by Biodex Stability System (eyes-opened); knee joint reposition sense; anterior laxity of the knee using KT-1000 (MEDmetric Corp, San Diego, California); knee extension/flexion torques; other physical function tests such as single-legged and triple hop tests, shuttle run, and carioca tests.
Standardized mean difference
Pooled data showed that no significant differences were found between WBVT and CON group in postural stability (n = 3; p > 0.05; Figure 3a) and in knee joint laxity (n = 2; p > 0.05; Figure 3c), but a significant difference was observed in knee position sense (n = 2; p < 0.01; Figure 3b).
Figure 3: Meta-analysis for (a) postural stability, (b) knee position sense, and (c) knee joint laxity. CI, confidence interval; CON, conventional rehabilitation program; WBVT, whole-body vibration training View Figure 3
The methodological quality scores were poor in the included studies. In exercise therapy studies including WBVT, blinding of participants and personnel would not be realistic. Selective reporting bias was 'unclear' in almost all articles as the Cochrane review group stated, "it is likely that the majority of studies will fall into this category (unclear)" . Meanwhile, all articles reported compliance and side effects, which should be important information in exercise therapy. Based on the recommendation of the Cochrane Back Review Group that drop-out rate in 1 group should be < 20% for short-term interventions (i.e., 4 to 12 weeks) , all included studies had no serious flaws. For the future studies, it will require configuring a study with allocation concealment, blinding assessors, and reporting the risk of co-interventions in WBV studies.
When we evaluated the quality of each study according to the ISMNI recommendations , included studies did not adequately report some factors related to the acceleration. First, no study measured the actual acceleration of the WBV platform. Second, because no study adequately described the information for the displacement of WBV device, it was not unsure whether the described displacement was peak-to-peak, and how a specific landmark was defined to ensure consistent targeting displacement of WBV in side-to-side alternating platform-type WBV, such as with a Galileo platform. The acceleration induced by the WBV platform affects training effects , it should be strictly adhered to these guidelines in future studies.
Despite the same measurement of postural stability using a Biodex device, a significant heterogeneity was observed among the included studies. The heterogeneity would be due to the timing of introduction of WBVT, exercise program and duration, and the diversity of WBV device. While Moezy et al.  found a significant improvement in postural stability by WBVT compared with CON program, Fu et al.  did not find a remarkable difference between WBVT and CON programs. In both studies, the participants performed body weight exercises on synchronous type WBV devices with similar vibration frequency and displacement (Table 1), whereas the sex of the included participants and the timing and the duration of WBVT program were apparently different. Either one or both of them might be important factor(s) to obtain higher improvement in postural stability by WBVT than by CON program.Pel et al.  reported that the adding weight loading to the WBV platform alters acceleration generation, and mean body weights of each study were likely to be different: Moezy et al., 74.3 kg  vs. Fu et al., 66.7 kg . If body weight of participants might make difference in the enhancement of postural stability by synchronous type WBVT, participants with light body weight like women should use resistive exercise with WBV, although the effects of resistive exercise with WBV on muscle performances such as muscle strength or jump performance are equivocal in healthy young individuals [22-24]. Because higher prevalence rate of ACL injury is reported in female athletes than in male ones , further research is needed to investigate the influences of body weight or additional weight loading on WBVT effects after ACL reconstruction surgery.
Knee position sense was significantly improved in WBVT compared with CON. Position sense is mainly adjusted by afferent signals from proprioception, and is one of the most essential factors for sports performance [25,26]. Not only injured knee but also non-injured one would decrease position sense after ACL injury . The results of this review suggest that exercise program mainly configured by bipedal exercises such as squatting might improve position sense in both knees. Although the improvement of knee position sense would contribute to prevention from re-injury, sport specific movements should be also included in rehabilitation of sports injuries . Although what sports activities the participants did was not clarified in the included studies [17-19], it was promising to investigate the effects of WBVT coupled with sport specific movements on the physical performance or the prevalence of re-injury in future studies. Meanwhile, no significant difference in knee laxity between WBVT and CON programs, but one study, as previously reported [11,29], demonstrated that the time for one rehabilitation session was significantly shorter in WBVT than that of CON(40.2 ± 2.3 vs. 85.0 ± 4.4 min/session) . Taken together, WBVT might be a time-efficient training modality compared with traditional resistance training program, and be a promising rehabilitation program after ACL reconstruction surgery.
To date, no standard WBVT regimens for rehabilitation after ACL reconstruction have been established, and therefore, the timing and duration of WBVT varied among the included studies. However, WBVT was likely to be regarded as a part of a whole rehabilitation program, which was mainly comprised of conventional physical therapy in all studies. Namely, one study configured a combination conventional rehabilitation with WBVT in the same training period for 8-week , whereas the other two studies did a WBVT-only training period after short-term conventional physical therapy in their rehabilitation program [17,19]. Future studies are needed to establish the rehabilitation program and the optimum time to introduce WBVT in the program, which would make the most of WBVT effects on physical function.
Previous review described that side effects of WBVT, particularly transient itching and erythema, muscle soreness, headache, forefoot pain, groin pain, fear, and knee pain were reported, and stated that these side effects were observed within the first 3-10 WBVT sessions . Although more than half of the participants experienced pain or swelling after WBVT rehabilitation program in a study by Berschin et al. , these symptoms might not be due to WBVT per se because the participants of the CON group also experienced the same symptoms. The combination of ACL reconstruction method, the timing of exercise intervention, and the exercise intensity might be related to these symptoms. Further research is needed to report the side effects of exercise therapy, particularly within the first couple of WBVT sessions.
Several limitations to the present study are to be mentioned. First, it remains unclear whether WBVT can speed up the time to return to sports activities. Moreover, it is still premature for judging whether WBVT can play a role in the prevention from re-injury after rehabilitation because the included studies in this review were too much short to prevent patients from re-injury [31,32].Therefore, future research is needed to accumulate supporting information on the effects of rehabilitation program using WBVT on the preventive effects from re-injuries. Second, the present study did not conduct thoroughly meta-analysis and perform testing for funnel plot asymmetry, as the power of test is too low to distinguish chance from real asymmetry if less than 10 studies are included . Therefore, the risk of publication bias could not be excluded. Third, the present study did not suggest the optimal vibration parameters or exercise prescription due to the lack of consistency in methodologies. Last, because this review excluded unpublished results (i.e., grey literature), the effects of WBVT on physical functions may possibly be overestimated .
In conclusion, this review suggests that WBVT might lead to improve some of the physical function tests, but its effects did not prove to be more effective compared to conventional rehabilitation program. Future studies with longer term WBVT with better quality are strongly needed.
The present study was financially supported by the Research Fellowships of Japan Society for the Promotion of Science for Young Scientists.
Prodromos C, Joyce B, Shi K (2007) A meta-analysis of stability of autografts compared to allografts after anterior cruciate ligament reconstruction. Knee Surg Sports Traumatol Arthrosc 15: 851-856.
van Grinsven S, van Cingel RE, Holla CJ, van Loon CJ (2010) Evidence-based rehabilitation following anterior cruciate ligament reconstruction. Knee Surg Sports Traumatol Arthrosc 18: 1128-1144.
Yang X, Wang P, Liu C, He C, Reinhardt JD (2015) The effect of whole body vibration on balance, gait performance and mobility in people with stroke: A systematic review and meta-analysis. Clin Rehabil 29: 627-638.
Sharififar S, Coronado RA, Romero S, Azari H, Thigpen M (2014) The effects of whole body vibration on mobility and balance in Parkinson disease: a systematic review. Iran J Med Sci 39: 318-326.
del Pozo-Cruz B, Adsuar JC, Parraca JA, del Pozo-Cruz J, Olivares PR, et al. (2012) Using whole-body vibration training in patients affected with common neurological diseases: a systematic literature review. Journal of alternative and complementary medicine 18: 29-41.
Cochrane DJ (2011) Vibration exercise: the potential benefits. Int J Sports Med 32: 75-99.
Fallon JB, Macefield VG (2007) Vibration sensitivity of human muscle spindles and Golgi tendon organs. Muscle Nerve 36: 21-29.
Burke D, Hagbarth KE, Lofstedt L, Wallin BG (1976) The responses of human muscle spindle endings to vibration during isometric contraction. J Physiol 261: 695-711.
Hagbarth KE, Eklund G (1966) Tonic vibration reflexes (TVR) in spasticity. Brain Res 2: 201-203.
Kouzaki M, Shinohara M, Fukunaga T (2000) Decrease in maximal voluntary contraction by tonic vibration applied to a single synergist muscle in humans. J Appl Physiol 89: 1420-1424.
Bogaerts A, Delecluse C, Claessens AL, Coudyzer W, Boonen S, et al. (2007) Impact of whole-body vibration training versus fitness training on muscle strength and muscle mass in older men: a 1-year randomized controlled trial. J Gerontol A Biol Sci Med Sci 62: 630-635.
Iwamoto J, Takeda T, Sato Y, Uzawa M (2005) Effect of whole-body vibration exercise on lumbar bone mineral density, bone turnover, and chronic back pain in post-menopausal osteoporotic women treated with alendronate. Aging Clin Exp Res 17: 157-163.
Rittweger J, Just K, Kautzsch K, Reeg P, Felsenberg D (2002) Treatment of chronic lower back pain with lumbar extension and whole-body vibration exercise: a randomized controlled trial. Spine (Phila Pa 1976) 27: 1829-1834.
Furlan AD, Pennick V, Bombardier C, van Tulder M; Editorial Board, Cochrane Back Review Group (2009) 2009 updated method guidelines for systematic reviews in the Cochrane Back Review Group. Spine (Phila Pa 1976) 34: 1929-1941.
Higgins JPT, Green S (2008) Cochrane handbook for systematic reviews of interventions. Wiley Online Library.
Rauch F, Sievanen H, Boonen S, Cardinale M, Degens H, et al. (2010) Reporting whole-body vibration intervention studies: recomme dations of the International Society of Musculoskeletal and Neuronal Interactions. J Musculoskelet Neuronal Interact 10: 193-198.
Berschin G, Sommer B, Behrens A, Sommer HM (2014) Whole Body Vibration Exercise Protocol versus a Standard Exercise Protocol after ACL Reconstruction: A Clinical Randomized Controlled Trial with Short Term Follow-Up. J Sports Sci Med 13: 580-589.
Fu CL, Yung SH, Law KY, Leung KH, Lui PY, et al. (2013) The effect of early whole-body vibration therapy on neuromuscular control after anterior cruciate ligament reconstruction: a randomized controlled trial. Am J Sports Med 41: 804-814.
Moezy A, Olyaei G, Hadian M, Razi M, Faghihzadeh S (2008) A comparative study of whole body vibration training and conventional training on knee proprioception and postural stability after anterior cruciate ligament reconstruction. Br J Sports Med 42: 373-378.
Rittweger J (2010) Vibration as an exercise modality: how it may work, and what its potential might be. Eur J Appl Physiol 108: 877-904.
Pel JJ, Bagheri J, van Dam LM, van den Berg-Emons HJ, Horemans HL, et al. (2009) Platform accelerations of three different whole-body vibration devices and the transmission of vertical vibrations to the lower limbs. Med Eng Phys 31: 937-944.
Preatoni E, Colombo A, Verga M, Galvani C, Faina M, et al. (2012) The effects of whole-body vibration in isolation or combined with strength training in female athletes. J Strength Cond Res 26: 2495-2506.
Wang HH, Chen WH, Liu C, Yang WW, Huang MY, et al. (2014) Whole-body vibration combined with extra-load training for enhancing the strength and speed of track and field athletes. J Strength Cond Res 28: 2470-2477.
Carson RG, Popple AE, Verschueren SM, Riek S (2010) Superimposed vibration confers no additional benefit compared with resistance training alone. Scand J Med Sci Sports 20: 827-833.
Relph N, Herrington L, Tyson S (2014) The effects of ACL injury on knee proprioception: a meta-analysis. Physiotherapy 100: 187-195.
Riemann BL, Lephart SM (2002) The sensorimotor system, part I: the physiologic basis of functional joint stability. J Athl Train 37: 71-79.
Arockiaraj J, Korula RJ, Oommen AT, Devasahayam S, Wankhar S, et al. (2013) Proprioceptive changes in the contralateral knee joint following anterior cruciate injury. Bone Joint J 95-95B: 188-191.
Puddu G, Giombini A, Selvanetti A (2013) Rehabilitation of sports injuries: current concepts. Springer Science & Business Media.
Delecluse C, Roelants M, Verschueren S (2003) Strength increase after whole-body vibration compared with resistance training. Med Sci Sports Exerc 35: 1033-1041.
Merriman H, Jackson K (2009) The effects of whole-body vibration training in aging adults: a systematic review. J Geriatr Phys Ther 32: 134-145.
Noyes FR, Barber-Westin SD (2014) Neuromuscular retraining intervention programs: do they reduce noncontact anterior cruciate ligament injury rates in adolescent female athletes? Arthroscopy: the journal of arthroscopic & related surgery 30: 245-255.
Zech A, Hübscher M, Vogt L, Banzer W, Hänsel F, et al. (2009) Neuromuscular training for rehabilitation of sports injuries: a systematic review. Med Sci Sports Exerc 41: 1831-1841.
McAuley L, Pham B, Tugwell P, Moher D (2000) Does the inclusion of grey literature influence estimates of intervention effectiveness reported in meta-analyses? Lancet 356: 1228-1231.