Low back pain (LBP) is defined as pain located below the costal margin and above the inferior gluteal folds[1]. Specific causes of LBP are uncommon, accounting for less than 15% of all back pain[2]. About 85% of patients with isolated LBP cannot be given a precise pathoanatomical diagnosis[3]. Nonspecific LBP has been defined as tension, soreness, and/or stiffness in the lower back region for which it is not possible to identify a specific cause of the pain[4]. It commonly leads to a loss of function and limitations in activities and participation in social life[5]. Because LBP pain is common in Western industrial societies, the economic consequences of back pain are enormous[6], and the effect on quality of life is substantial[7].

Back pain in pregnant and postpartum women is also common. The prevalence of LBP in pregnancy ranges from 24% to 90%, although it is most commonly estimated at 40%-50%[8–10]. Prevalence increases with the duration of pregnancy and is at the highest point in the third trimester[11, 12]. The cause of pain appears to be nonspecific and may be related to changes in body posture with the development of joint, ligament, and myofascial dysfunctions[13, 14]. The prevalence of LBP in postpartum women increases in the year after delivery, with estimates from 28% after 3 months to over 50% after 5 months and 67% after 12 months[15–18].

Osteopathy is a health approach that emphases the role of the musculoskeletal system in health and promotes optimal function of the tissues of the body by using a variety of manual techniques to improve the function of the body[19]. In the United States, practitioners are known as osteopathic physicians and have full medical licence. Osteopathic manipulative treatment (OMT) typically involves an eclectic range of manual techniques, which may include soft tissue stretching, spinal manipulation, resisted isometric ‘muscle energy’ stretches, visceral technique, or exercise prescription, for example. Treatment is characterised by a holistic approach to the patient, and OMT may be applied to many regions and tissues of the body, sometimes remote from the symptomatic area and at the clinical judgement of the practitioner[19–21].

Patients with LBP visit osteopaths for treatment, although the number of patients consulting osteopaths is not clear. In the United Kingdom, osteopaths were estimated to perform 4.38 million treatments in 1998[22]. Lumbar symptoms are the most common presentation in osteopathic practice in the United Kingdom, and, in a national pilot survey[23] and a snap-shot survey[24], accounted for 36% and 46% of presenting symptoms in patients, respectively. In Australia, the osteopathic profession is relatively small, and of those patients with LBP who chose to see a practitioner, medical practitioners (22.4%) and chiropractors (19.3%) are the most popular care providers, with only 2.7% of patients seeing osteopaths[25]. Despite this, LBP is the most common patient presentation in osteopathic practices[26].

In the United States, osteopathic physicians are more likely to provide LBP care than their allopathic medical counterparts[27].

To our knowledge, only 2 systematic reviews exist for osteopathic treatment of LBP. In 2005, Licciardone et al.[28] published the first systematic review of OMT for LBP and concluded that OMT significantly reduced LBP. This review had a number of limitations and was criticized because it did not differentiate between OMT and single manual techniques[29] and because single techniques do not reflect osteopathic clinical practice. Further, it combined dichotomous and continuous outcomes, combined studies with specific and nonspecific back pain, lacked a risk of bias evaluation, and contained a unit of analysis error. Given these shortcomings, reservations remain concerning the authors’ conclusions[29].

In 2012, Orrock and Myers[30] published another systematic review of OMT for LBP. This review only included studies of chronic nonspecific LBP and was limited to those published in English[30]. Only 2 studies met their specific search criteria, so no meta-analysis or robust conclusions were possible.

The objective of the current review was to examine the effectiveness of OMT for improving pain and functional status in LBP patients as compared to control treatments (no treatment, sham, and all other treatments) for adults in randomized clinical trials. Although 2 systematic reviews have been published on this topic, we were aware of recent, non-English studies that were not included in the previous reviews. Further, we intended to search the non-published ‘grey’ literature for studies which have not been published in journals or books, as recommended by the Cochrane Collaboration[31] for preparing and updating high-quality systematic reviews. We expected that using rigorous methodology and an extensive search without language restriction would provide a more comprehensive insight into the effectiveness of this intervention.

To our knowledge, the current review is the first systematic review with meta-analyses examining the effect of osteopathic management for acute and chronic nonspecific LBP based only on studies that used an authentic osteopathic approach. This approach required clinical judgment to individualise the treatment to each patient, rather than applying a single technique or predetermined set of techniques. Because our review had no language or publication restrictions, it is likely the most comprehensive review to date. When included studies were grouped and analysed using meta-analyses, a significant effect for OMT was found for LBP (acute and chronic), chronic LBP, LBP in pregnant women, and LBP in postpartum women. The significant effects were also found to be clinically relevant according to the criteria recommended by the Cochrane Collaboration[38].

The risk of bias in the studies was generally low, with all but 3 of the 15 included studies found to have low risk. None of the studies had a high risk of bias in the randomisation and allocation procedure, but every study had problems with the 3 blinding criteria in the risk of bias assessment. For studies of manual therapy, blinding will usually be an issue because patients tend to be aware of the manual treatment and practitioners cannot be easily blinded from the treatment intervention they deliver. When using methodology assessment tools such as the Risk of Bias instrument, the difficulty of blinding creates a disadvantage for manual therapy studies compared to studies using other interventions like pharmaceuticals which can easily be blinded.

The 2013 study by Licciardone et al.[50] was the largest RCT included in the current review, assessing 455 patients with chronic LBP. The data in the study was not normally distributed and the authors reported medians and interquartile ranges, which were not easily used for meta-analyses. We contacted the authors of this study[50] several times for additional data that could be used in the current analysis, but unfortunately this data was not made available. Subsequently, we needed to transform these data to determine means and standard deviations. We used simulation calculations recommended by Hozo et al.[37], where the median was the best estimator for the mean for sample sizes greater than 60. For the estimation of standard deviations, we calculated average standard deviations based on 3 studies[46, 51, 52], which were similar in outcome, comparison and duration of pain. For the estimation of standard deviations for functional status, we based calculations on two studies[46, 51]. For the margin of error for every estimation, it was possible that a greater difference between our estimation and the real data (i.e., the data was more in favour of the control group) could change our results regarding functional status in chronic and acute and chronic back pain. However, our results for the comparisons were almost identical regardless of whether the transformed data from Licciardone et al.[50] were included or not.

Two previous systematic reviews examined the effect of OMT on LBP. In a 2005 review by Licciardone et al.[28], studies were included if they were performed by an osteopath or osteopathic physician, but the authors also included interventions based on single manual techniques. In the current review, we wanted to examine the effect of studies that used an authentic osteopathic intervention where the clinician was free to use clinical judgment for each patient, as occurs in clinical practice. Consequently, we excluded 2 studies[58, 59] that were included in the 2005 Liccardone et al. review[28] because they involved single techniques. Further, we did not include studies with specific causes of LBP[68]. Although our review did include studies of LBP associated with pregnant and postpartum women, these studies were pooled and analysed separately. Despite these differences, the results and conclusions of our study and of the Licciardone et al.[28] study are similar: both suggested that OMT may be an effective treatment for LBP.

The findings of the current review differ from the results of a recent review by Orrock and Myers[30], largely due to different inclusion criteria. The Orrock and Myers review[30] was restricted to chronic nonspecific LBP and consequently fewer studies met their inclusion criteria. The current review was not restricted to the English language or the published literature, and we located 6 unpublished studies in German[42, 46, 48, 53–55] and 1 study in Italian[52]. Searching the unpublished grey literature for relevant studies is recommended by the Cochrane Collaboration for a more comprehensive search and to avoid publication bias[31]. In addition, the limited number of studies retrieved by Orrock and Myers prevented statistical analysis, whereas we were able to conduct meta-analyses to determine the effect of OMT interventions on LBP. In another systematic review, Posadzki and Ernst[69] examined the effect of osteopathy for musculoskeletal pain. However, Posadzki and Ernst[69] did not specifically address LBP, had only 5 studies that focused on LBP, and had no quantitative analysis, so this study is not comparable to the current review.

OMT appeared to have a larger effect on pain than functional status. Given that our analyses used results from the included studies recorded 3 months after the initial intervention, the subjective experience of pain may be quicker to respond to treatment than function.

It is difficult to assess the relative effectiveness of OMT compared to other specific interventions commonly offered to people with LBP using the available studies. The comparison interventions of the studies included in this review were varied, including sham treatment, usual medical care, physiotherapy, and no treatment; and it was not possible to group and analyse these studies according to the comparison intervention. Rubinstein et al.[32] found that there was high-quality evidence that spinal manipulation, a technique used by osteopaths and other manual therapists, had a small short-term effect on pain, but the effect was not clinically significant. The current review suggested that the effect of OMT was clinically relevant, and it may be that an individualised approach with different techniques contributed to greater effectiveness. However, Rubinstein et al.[32] had access to a greater number of studies with a total of 6070 participants, and the authors were able to examine different time periods for longevity of effectiveness. Walker et al.[70] reviewed studies of chiropractic management of LBP when combined with other interventions, as represents typical practice for many chiropractors, rather than spinal manipulation alone. Chiropractic interventions were found to improve pain in the short and medium term, but not in the long term, compared to other interventions. For functional status, there were short-term improvements, but not in the medium and long term. This review included 12 studies involving 2887 participants, substantially more than the current review on OMT, but only 3 studies had low risk of bias. Given the differing comparison groups in the studies of both reviews, it is not possible to directly compare the effects of OMT and chiropractic management.

Two important limitations of the current review were the sample sizes and comparison groups of the included studies. When studies include few participants and have wide confidence intervals in the analysis, or have small confidence intervals with effects in different directions, heterogeneity is evident and the rating of the quality of evidence should be downgraded according to the GRADE approach recommended by the Cochrane Handbook[31]. Although the majority of the included studies had relatively small sample sizes[39, 42, 46, 48, 51–56], each comparison for chronic and acute pain and for chronic pain contained over 400 participants. However, the comparisons for the conditions of LBP in pregnant and postpartum women contained fewer than 400 participants, which indicated likely imprecision of results and a resultant downgrading of the level of evidence[31]. Future studies with larger samples sizes may change our estimates of effect size for all these comparisons, particularly for LBP in pregnant and postpartum women. There were also a number of different comparison groups in the included studies, including placebo control, usual medical care, and untreated patients.

Considerable heterogeneity was evident in many of the forest plots, which indicated variability and poor overlap in the confidence intervals of the studies. This heterogeneity may be related to the small sample sizes of the studies, as well as the different comparison interventions, which may have had differing effects on pain and functional status. The small sample sizes of many of the studies, the different comparison interventions, and the heterogeneity are limitations of the current review and cause for caution concerning the conclusions. Although we performed meta-analyses on patient groups with different chronicity of symptoms, this did not appear to be a major source of heterogeneity. All patient groups together had a substantial heterogeneity of I² = 85%, but the heterogeneity of only the mixed acute and chronic groups (I² = 81%) and only the chronic groups (I² = 89%) were similar.

It should be noted that the broad widths of the 95% CIs in the forest plots indicate imprecision of the results. This is often the case with systematic reviews of RCTs with small sample sizes. We have interpreted clinical relevance based on the scores of the MD and SMD, but it is necessary to consider lower and higher bounds of the CI and that the true value may lie in this range. The true value could be higher or lower than our point estimator from which we have calculated the clinical relevance, and future studies, using larger samples and robust methodology, may clarify the true point estimate and the clinical effectiveness of OMT for LBP.

The delivery of OMT, which can include a range of manual techniques, is not standardized between practitioners and requires individual clinical judgment for each patient. Most of the included studies provided an indication of the range of manual techniques used for OMT, but the exact interventions performed for each patient were generally unknown. For instance, OMT interventions in the included studies may emphasize different manual treatment approaches. Unfortunately, this lack of specific information from the included studies does not enable us to identify whether responder and non-responder patient groups received different treatments or to understand what the most effective components of OMT interventions are for LBP.

The pain and functional status outcomes analysed in the current review were measured in each study close to 3 months after the initial treatment. Therefore, the longevity of the effect of OMT for LBP cannot be determined, and most of the included studies did not have a longer follow-up period for assessment of pain and functional status. Details about the treatment approach used and clearly reported adverse events should also be included in studies. Future studies should examine the long-term effects of OMT, clearly describe the treatment approach, and report adverse events. Because of the small sample sizes in the majority of the included studies and the heterogeneity in our analyses, these future studies should also have larger sample sizes. Larger RCT studies are expensive to conduct and most of the reviewed studies were unfunded. In order to produce large RCTs examining the effect of OMT on LBP, there must be willingness from osteopathic professional organizations and national funding bodies to support such research.

To our knowledge, the current systematic review used the most comprehensive search for studies of OMT for nonspecific LBP. The studies we reviewed generally had a low risk of bias, but most had relatively small sample sizes of patients. Our results suggest that OMT improves pain and functional status in acute and chronic nonspecific LBP, in chronic nonspecific LBP, and in pregnant and postpartum women with LBP. Given the small sample sizes, different comparison groups in different studies, heterogeneity, and lack of long-term measurement, larger, high-quality RCTs with robust comparison groups are needed to provide firm conclusions regarding the effectiveness of OMT for LBP.