Osteopathic manipulative treatment of patients with chronic low back pain in the United States: a retrospective cohort study

Study design and participants

A retrospective cohort study was utilized to measure the clinical outcomes of OMT over 12 months. Participants were selected from the Pain Registry for Epidemiological, Clinical, and Interventional Studies and Innovation (PRECISION) from April 2016 through April 2022. Registry participants were screened and recruited from the 48 contiguous states and the District of Columbia, primarily through social media. Registry data were self-reported by participants at quarterly encounters utilizing a digital research platform and electronic data capture. Eligible registry participants ranged from 21 to 79 years of age at enrollment, had sufficient English language proficiency to complete case report forms independently or with staff assistance, were not pregnant, and did not reside in institutional facilities. For inclusion, registry participants must have reported CLBP according to criteria established by the National Institutes of Health Task Force on Research Standards for Chronic Low Back Pain (NIH-RTF) [16]. These criteria require having low back pain for at least the past 3–6 months and with a frequency of at least half of the days in the past 6 months. Participants were also required to have a physician who regularly provided CLBP treatment. This research was approved by the North Texas Regional Institutional Review Board (protocol 2015-169), and all participants provided informed consent prior to contributing data. This study is reported following the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines [17]. Further information about PRECISION is available at ClinicalTrials.gov [18].

Treatments for low back pain

The use of OMT was determined by participant reporting of their physician type (osteopathic or allopathic) and current or prior use of spinal manipulation to treat low back pain at registry enrollment. Consistent with terminology espoused by the National Center for Complementary and Integrative Health [19], the generic term “spinal manipulation” is utilized by the registry because it is more universally recognized than the term “osteopathic manipulative treatment,” or OMT, although the latter includes manual techniques other than spinal manipulation. Because the registry does not collect data on the specific type of spinal manipulation provider, participants who were currently treated by an osteopathic physician and reported utilizing spinal manipulation were classified as OMT users. Participants were classified as OMT nonusers if they were treated by an allopathic physician or by an osteopathic physician without current or prior use of spinal manipulation. Since September 2016, registry participants have been required to have a physician (i.e., osteopathic or allopathic physician) who explicitly treats their low back pain as a condition of enrollment. Thus, the registry does not collect data on the use of chiropractic, as this modality is unlikely to be provided by either osteopathic or allopathic physicians. The current use of NSAIDs and opioids for low back pain was also measured at registry enrollment.

Outcome measures

The four outcome measures were recommended by the NIH-RTF [16]. A numerical rating scale ranging from 0 to 10 was utilized to measure average low back pain intensity during the 7 days prior to a registry encounter. Pain impact was measured utilizing nine items on the Minimum Dataset recommended by the NIH-RTF, including pain intensity and eight items derived from the Patient-Reported Outcomes Measurement Information System (PROMIS) [20, 21]. These included four items in each of two PROMIS scales involving physical function and pain interference with activities within the 7 days prior to an encounter. Pain impact scores may range from 8 to 50. The Roland-Morris Disability Questionnaire (RMDQ) was utilized as a legacy measure of physical function [16]. It consists of 24 items that measured back-related disability on an encounter date, with scores ranging from 0 to 24 [22]. Health-related quality-of-life measures included four PROMIS scales in the Minimum Dataset (sleep disturbance, pain interference, depression, and low energy/fatigue) and a fifth PROMIS scale to measure anxiety. Collectively, 20 items (four items in each PROMIS scale) comprise the SPADE cluster (sleep disturbance, pain interference, anxiety, depression, and low energy/fatigue) that measures health-related quality-of-life deficits. All SPADE scale scores, except sleep disturbance, are normed according to the general US population and have a mean of 50 and standard deviation (SD) of 10 [20]. The sleep disturbance scale is similarly scored; however, it is normed with a calibration sample enriched for chronic illness. The SPADE cluster score is the mean of its five scales and may range from 38 to 77. Each of the four study outcomes was measured at registry enrollment and at up to four quarterly encounters over 12 months. Higher scores on each measure represent worse outcomes.

Statistical analysis

Descriptive statistics were utilized to characterize participants at registry enrollment. Logistic regression was utilized to compute unadjusted and adjusted odds ratios (ORs) and 95% confidence intervals (CIs) associated with OMT use. The multivariable model included age, gender, race, ethnicity, cigarette smoking status, chronic widespread pain, comorbidities (herniated disc, sciatica, depression), low back surgery, and current medication use for low back pain (NSAIDs, opioids). These variables were largely measured utilizing elements of the Minimum Dataset [16].

Repeated measures analysis of variance (ANOVA) was utilized to assess each of the four outcomes during 12 months of follow-up. Outcomes were measured as between-group differences in means (i.e., in grand means over 12 months) for OMT nonusers minus OMT users. Hence, positive values favored OMT in all analyses. Cohen’s d statistic was utilized to assess the clinical relevance of between-group differences, with d values ≥ 0.2 considered clinically important [23]. Participants who withdrew from the registry, reported OMT crossover (i.e., entered the cohort in the OMT group and then switched to a physician who did not provide OMT, or vice versa), or missed any quarterly encounter during 12 months of follow-up were excluded from these analyses. Both unadjusted and propensity-score adjusted analyses were performed. The latter included propensity scores for OMT use upon entry to the cohort, computed with the aforementioned multiple logistic regression model to control for confounding by indication. This is particularly useful in observational studies of treatment effect, wherein randomization is not possible, to control for important factors that may differentiate treatment users and nonusers [24]. In a series of sensitivity analyses, linear mixed models were utilized to study outcomes among participants without OMT crossover, but with either complete or incomplete quarterly encounter data.

A series of subgroup analyses were performed, each including the unadjusted and adjusted analyses described above, to identify OMT interaction effects involving the 12 prespecified variables in the multiple logistic regression model. Data management and analyses were performed with the IBM SPSS Statistics (Version 28) software. Hypotheses were generally tested at the 0.05 level of statistical significance utilizing two-sided tests. However, because the subgroup analyses estimated OMT effects for four outcomes (pain intensity, pain impact, physical function, and health-related quality of life) within 25 subgroups (100 comparisons), a Bonferroni-corrected significance threshold of p<0.001 was utilized to interpret the results. Similarly, a Bonferroni-corrected significance threshold of p=0.001 was utilized to interpret the OMT interaction effects involving the four outcomes and 12 prespecified subgroup variables (48 comparisons).

Participant characteristics and OMT use

The mean (SD) age of 1,358 participants was 53.2 (13.1) years, and 1,010 (74.4%) were female. There were 187 (13.8%) participants who utilized OMT at registry enrollment. Black race, cigarette smoking, depression, and NSAID use were each inversely associated with OMT use (Table 1). In the multivariable analysis that controlled for potential confounders, participants who were Black (OR, 0.36; 95% CI, 0.21–0.63; p<0.001), cigarette smokers (OR, 0.56; 95% CI, 0.33–0.93; p=0.02), and current NSAID users (OR, 0.59; 95% CI, 0.43–0.81; p=0.001) remained less likely to utilize OMT.

Flow of participants through the study and unadjusted outcomes

Of 1,358 participants who entered the cohort, 913 (67.2%) attended the 12-month encounter, 348 (25.6%) were in various stages of follow-up, and 97 (7.1%) had withdrawn (Figure 1). There were 797 participants without OMT crossover at the 12-month encounter, including the 753 (94.5%) participants with complete follow-up data in the primary analyses. The OMT users reported better outcomes pertaining to pain intensity, pain impact, physical function, and health-related quality of life (Figure 2). The mean between-group differences were: 1.02 (95% CI, 0.63–1.42; p<0.001) for pain intensity; 5.12 (95% CI, 3.09–7.16; p<0.001) for pain impact; 3.59 (95% CI, 2.23–4.95; p<0.001) for physical function; and 2.73 (95% CI, 1.19–4.27; p<0.001) for health-related quality of life. These results were clinically important, with Cohen’s d statistics of 0.60, 0.58, 0.61, and 0.41, respectively.

Figure 1:

Flow of participants through the study. ^aRegistry participants were required to have a physician who treated their low back pain beginning in September 2016.

Figure 2:

Unadjusted outcomes during 12 months of follow-up. (A) Pain intensity was measured with a numerical rating scale (range, 0–10). (B) Pain impact was measured utilizing pain intensity and the physical function and pain interference scales on the Patient-Reported Outcomes Measurement Information System (PROMIS; range, 8–50). (C) Physical function was measured as back-related disability with the Roland-Morris Disability Questionnaire (RMDQ; range, 0–24). (D) Health-related quality of life was measured utilizing the SPADE cluster (sleep disturbance, pain interference, anxiety, depression, and low energy/fatigue) scales on the PROMIS (range, 38–77). Higher scores reflect worse outcomes on each measure. The means and p values for each outcome are based on 753 participants without OMT crossover and with complete data for all five encounters during 12 months of follow-up. Error bars represent 95% confidence intervals. OMT, osteopathic manipulative treatment.

Propensity-score adjusted outcomes

The use of OMT remained associated with better outcomes after adjusting for propensity scores (Figure 3). The mean between-group differences were: 0.62 (95% CI, 0.24–0.99; p=0.001) for pain intensity; 3.12 (95% CI, 1.16–5.08; p=0.002) for pain impact; 2.24 (95% CI, 0.93–3.56; p<0.001) for physical function; and 1.55 (95% CI, 0.03–3.07; p=0.045) for health-related quality of life. The corresponding Cohen’s d statistics were estimated as 0.37, 0.35, 0.38, and 0.24, respectively.

Figure 3:

Adjusted outcomes during 12 months of follow-up. (A) Pain intensity was measured with a numerical rating scale (range, 0–10). (B) Pain impact was measured utilizing pain intensity and the physical function and pain interference scales on the Patient-Reported Outcomes Measurement Information System (PROMIS; range, 8–50). (C) Physical function was measured as back-related disability with the RMDQ (range, 0–24). (D) Health-related quality of life was measured utilizing the SPADE cluster (sleep disturbance, pain interference, anxiety, depression, and low energy/fatigue) scales on the PROMIS (range, 38–77). Higher scores reflect worse outcomes on each measure. The means and p values for each outcome are based on 753 participants without OMT crossover and with complete data for all five encounters during 12 months of follow-up and are adjusted for the propensity score for reported OMT use upon entry to the cohort. Error bars represent 95% confidence intervals. OMT, osteopathic manipulative treatment.

Sensitivity and subgroup analyses

The results of sensitivity analyses that captured available data for 797 participants without OMT crossover, but with either complete or incomplete follow-up, were not materially different than the unadjusted and adjusted results reported above (Table 2). The direction of the patient subgroup results favored OMT in 98 of 100 unadjusted analyses and in 97 of 100 adjusted analyses. However, only 43 results in the unadjusted analyses and 10 results in the adjusted analyses met the Bonferroni-corrected significance threshold (p<0.001) (Supplemental Table 1). No subgroup comparison demonstrated an OMT interaction effect at the Bonferroni-corrected significance threshold (p=0.001).