Pneumonia is the fourth most common hospital discharge diagnosis in the US with a mean (SE) length of stay (LOS) of 5.1 (0.1) days [1]. The elderly are a vulnerable population: the majority of pneumonia-related hospital admissions occur in persons 60 years and older, and the elderly have a longer mean LOS, higher severity of illness, and greater mortality than younger age groups [2–5]. Adjunctive nonpharmacologic treatments for pneumonia may enhance conventional antibiotic therapy. Chest physiotherapy, early mobilization, and continuous lateral rotational therapy have been investigated with mixed results [6–11]. Osteopathic manipulative treatment (OMT) is a nonpharmacologic manual therapy developed in the late nineteenth century before the use of antibiotics. OMT includes a number of manipulative techniques intended to enhance host defenses and physiologic function [12–16]. Many of these techniques, such as rib raising, doming the diaphragm, and the thoracic lymphatic pump, were specifically developed to treat pneumonia [15]. Two small randomized controlled trials of OMT for pneumonia suggest OMT may reduce LOS in the elderly [17, 18].

The Multicenter Osteopathic Pneumonia Study in the Elderly (MOPSE) was designed to further evaluate the efficacy of OMT. To better understand the potential therapeutic effect of the interaction between participants and providers, MOPSE included sham treatment and conventional care only control arms. The primary hypotheses were that OMT would reduce LOS, time to clinical stability [19], and a symptomatic and functional recovery score [20] in elderly patients hospitalized with pneumonia compared to light touch sham and conventional care only control groups. Analysis of secondary and safety-related outcomes was also performed.

MOPSE is a registered (http://www.clinicaltrials.gov, NCT00258661), double-blinded, randomized, controlled trial conducted from March 2004 to April 2007 with a 60-day post-admission follow-up period.

Subjects aged ≥ 60 years newly hospitalized with pneumonia were recruited for MOPSE. Ten months into the study the age criterion was lowered to ≥ 50 years to improve recruitment. Eligibility criteria included a new pulmonary infiltrate on chest x-ray and at least two of the following: new or increased cough, fever ≥ 38°C, pleuritic chest pain, new physical findings on chest examination, respiratory rate ≥ 25 breaths/min, deteriorating mental or functional status, or white blood cell count (WBC) > 12,000 cells/mm³. Exclusion criteria were nosocomial pneumonia, lung abscess, advancing pulmonary fibrosis, bronchiectasis, pulmonary tuberculosis, lung cancer, metastatic malignancy, uncontrolled metabolic bone diseases, current rib or vertebral fracture, prior pathologic fracture, previous study participation, or respiratory failure. See Figure 1 for the MOPSE flow diagram.

Subjects were recruited at seven community hospitals in one rural, two suburban, and two urban US communities. The respective institutional review boards approved the study protocol, and all participants gave informed consent. An independent data and safety monitoring board provided oversight for MOPSE.

Subjects were enrolled and randomized into three groups: conventional care only (CCO, standard care control), light-touch treatment (LT, sham control), or osteopathic manipulative treatment (OMT). All subjects received conventional treatment for pneumonia directed by their attending physicians. Subjects were stratified by study site and randomized from a computer-generated list using block sizes of 3 or 6. Allocation was concealed with opaque, sealed envelopes which were opened after enrollment. Subjects, personnel responsible for collecting data, attending physicians, nurses, and house staff caring for subjects were blinded to group assignment. Only the physicians giving the study treatments were unblinded to group assignment.

Participants randomized to the OMT or LT groups received protocol treatments for 15 minutes, twice daily (≥ 6 hours apart) beginning within 24 hours of admission and continuing until hospital discharge, cessation of antibiotic therapy for pneumonia, respiratory failure (ventilator dependent), death, or study withdrawal. All subjects were treated while supine in bed. The manipulation techniques of the OMT protocol were administered in the following sequence: thoracolumbar soft tissue, rib raising, doming of the diaphragm myofascial release, cervical spine soft tissue, suboccipital decompression, thoracic inlet myofascial release, thoracic lymphatic pump, and pedal lymphatic pump [15]. Soft tissue technique consists of massage, stretching, kneading, and direct inhibitory pressure to relax the musculature. Rib raising articulates each rib for the purpose of improving rib cage motion and theoretically stimulates the sympathetic chain ganglia. Myofascial release is a method for reducing tissue tension. Doming the diaphragm and thoracic inlet myofascial release techniques are used to improve diaphragmatic movement and lymphatic drainage. Suboccipital decompression involves traction at the base of the skull, which is considered to release restrictions around the vagus nerves, theoretically improving nerve function. The thoracic lymphatic pump with activation combines rhythmical compressions to the chest wall and the rapid removal of the hands from the chest wall during deep inhalation with the intention of enhancing lymphatic circulation and triggering a sudden expansion of airways and alveoli. The pedal lymphatic pump gently rocks the patient in a superior-inferior rhythmical motion while supine, to theoretically enhance lymphatic circulation.

Non-thrust techniques were used to treat areas unaddressed by the above techniques, limited to ≤ 5 minutes of the 15-minute session. The LT protocol, designed as a sham control treatment, applied light touch to the same body regions, in the same sequence, and for the same duration as the OMT protocol. A more detailed description of the OMT and LT protocols has been published [15].

Twenty osteopathic neuromusculoskeletal (OMT) specialists and 64 resident physicians from 12 specialties administered the protocols. An OMT specialist at each site administered one of the first two OMT or LT treatments and 3 treatments per week thereafter. Using a standardized patient, treatment skills and protocol adherence were evaluated by the principal investigator and an OMT specialist at 9 training sessions at each site during the study. A pressure mapping system (Sensor Products LLC, East Hanover, NJ) was used to standardize three OMT techniques: rib raising, suboccipital decompression, and thoracic lymphatic pump. Presentation of the pressure mapping data is beyond the scope of this manuscript.

Primary outcomes were LOS, time to clinical stability [19], and a symptomatic and functional recovery score [20]. LOS was defined by the date and time of the admission and discharge orders. Based on data recorded daily, time to clinical stability was defined as the hospital calendar day when all seven clinical parameters first met criteria for stability (ie, lowest systolic blood pressure ≥ 90 mmHg, highest heart rate ≤ 100 beats/min, highest respiratory rate ≤ 24 breaths/min, highest temperature ≤ 38°C, lowest oxygen saturation ≥ 90%, ability to eat food by mouth or by a feeding tube, and mental status back to pre-pneumonia baseline) [19]. The symptomatic and functional recovery score was calculated from a pneumonia-specific, validated questionnaire addressing five symptoms: cough, dyspnea, sputum production, pleuritic chest pain, and fatigue [20]. Higher scores indicate more symptoms. This questionnaire was administered at admission and via telephone on post-admission days 14, 30, and 60.

Secondary outcomes were duration of intravenous and oral antibiotics; treatment endpoint, including death and respiratory failure; 60-day hospital readmission rate; highest daily temperature; highest daily respiratory rate; and WBC. Baseline severity of illness was assessed using the pneumonia severity index [21]. Adverse events, whether related to the study treatments or not, were monitored daily while subjects were in the hospital. Treatment side effects were evaluated 24 hours post-discharge; subjects were asked to report the severity and character of muscle soreness, worsening of breathing, or other side effects. The success of subject blinding was assessed via questionnaire within 24 hours of hospital discharge.

Using pilot data [18] to estimate the median LOS for the OMT and CCO groups, a log-rank test for survival has 80% power to detect a difference in median LOS of 6 versus 9 days when the sample size in each group is 96 subjects discharged from the hospital. MOPSE was designed to enroll 120 subjects per group with censored values (death, respiratory failure, or study withdrawal) estimated at ≤ 20% of the subjects.

Data were entered in duplicate at the study site and the central coordinating center, and checked for agreement. Data were analyzed by intention-to-treat (ITT) analysis and by per-protocol (PP) analysis of subjects receiving 100% of prescribed treatments. Data were analyzed for all subjects (≥ 50 years) and then separately for subjects who met the original age criterion (≥ 60 years). To test the hypotheses regarding group differences, we used three statistical analysis methods that stratify by study site to account for clustering. Groups were compared on LOS, time to clinical stability, and duration of in-hospital antibiotics using stratified Cox proportional hazards models in order to include incomplete data from subjects who withdrew from the study, died, or were placed on a ventilator. Hazard ratios for LOS greater than 1 correspond to earlier discharge from the hospital for the treatment group compared to the control group. Differences in the three groups on continuous outcome measures were assessed using general linear mixed-effects models with subjects and study sites treated as random effects and group assignment, time, and the interaction of group and time as fixed effects. The Cochran-Mantel-Haenszel test for general association was used to analyze categorical outcome measures and compare the incidence of adverse events and serious adverse events between the groups. P values ≤ 0.05 were considered statistically significant. The statistical analyses were performed using SAS^© version 9.1 (SAS Institute, Inc., Cary, NC).

By ITT analysis, OMT did not improve outcomes. By PP analysis, OMT decreased LOS, duration of intravenous antibiotics, and the incidence of respiratory failure and death relative to CCO. Thus, if a subject received the OMT protocol as prescribed without missing any treatment sessions, there was significant benefit.

While the two studies conducted prior to MOPSE also suggest OMT reduces LOS [17, 18], the mean LOS for all groups has been dramatically reduced since the early 1990s, from 2 weeks in the first pilot study to less than 5 days in the current study. These changes are consistent with national trends [1, 24, 25]. The current study was powered to detect a difference in median LOS of 6 versus 9 days, estimated using data from the single site study [18] when the LOS was 33%-100% longer than current standards. This LOS reduction limited the number of treatments given and the length of time to observe the potential impact of OMT and LT on the course of the disease. While all subjects in the current study continued to receive antibiotic therapy after hospital discharge, study treatments ended at hospital discharge. OMT may show greater efficacy when continued beyond acute care or when used in chronic care settings. In addition, a larger sample size may be needed to compensate for the reduced LOS.

Consistent with other studies [19, 26], the time to clinical stability in the current study was two days, probably too short to identify between-group differences. The symptomatic and functional recovery score, a known measure of long-term recovery from pneumonia [20], did not show any between-group differences either. The 14% post-discharge loss to follow-up reduced the power and may have lessened the ability of this measure to detect between-group differences.

For the secondary outcome of treatment endpoint, death and respiratory failure were reduced for OMT relative to CCO by PP analysis. The LT group was intermediate between the other two groups and not significantly different from either. Because of the relatively low numbers involved, these findings should be interpreted with caution.

Significantly more patients receiving OMT experienced musculoskeletal soreness or pain sometime during the hospital stay. However, only one subject withdrew consequent to this side effect, and the severity was generally mild, suggesting OMT is reasonably tolerated even in this severely ill cohort.

Those making the clinical care decisions and collecting data and the subjects were blind to group assignment. The post-discharge survey suggests at least partial blinding since only half of the participants correctly identified their group assignment and 30% were unsure of their group assignment. It is surprising that only 49% of subjects in the CCO group correctly identified their group but not surprising that those in the CCO group who felt certain about group assignment more consistently identified their group. However, those in the OMT and LT groups who believed they knew their group assignment also more consistently identified their group. The impact of this partial blinding should be considered in future research.

A concern with PP analysis is that the exclusion of subjects may compromise the baseline similarity of the groups achieved by randomization, altering the risk-factor profile of the subjects within the groups [27, 28]. In the current study, almost 18% of subjects (26% of OMT, 23% of LT, and 5% of CCO subjects) were excluded from the PP analysis due to missed treatments during hospitalization (43%), delayed initiation of treatments (10%), treatment contrary to protocol (15%), and study withdrawal (32%). As a result of these exclusions, subjects who met PP criteria in all three groups had significantly less severe pneumonia than those who were excluded (P = 0.005). However, the reasons for being excluded from PP analysis (protocol violation, study withdrawal, or receiving less than 100% of prescribed treatments) were not significantly related to pneumonia severity (P = 0.77). While more subjects were excluded from the OMT and LT groups than the CCO group for the PP analysis, the composition of the groups regarding demographics, comorbidities, and pneumonia severity was not significantly different with one minor exception for current alcohol use. This similarity suggests that the PP analysis sustained the between-group characteristics from the original randomization, minimizing the potential bias that can occur with PP analyses.

Another concern many have with PP analysis is whether the results represent real world conditions. While consistency of treatment is a valid concern and does influence the interpretation of data, improved compliance with the intended protocol can be achieved with current quality assurance mechanisms. In the current study, almost 70% of the subjects excluded from PP analysis were excluded for reasons that are mostly modifiable by healthcare providers (eg, initial treatment given more than 24 hours after admission, incorrect treatment given, missed treatment).

Alternatively, a concern with ITT analysis is that lack of adherence to the treatment protocol can result in underestimation of the impact of the treatment, particularly when the proportion of subjects who did not receive the full protocol treatment is sizable [28]. While this concern may be valid for the current study, the potential degree of underestimation was minimized due to the reduced LOS, which already limited the time for OMT efficacy.

Few studies have explored how osteopathic techniques might enhance host defenses. In rats, rhythmic mechanical pressure to body regions physically distant from the location of lymphatic formation enhances lymph uptake [29]. Knott et al [30] showed in a dog model that lymphatic pump techniques increased lymph flow through the thoracic duct, independent of cardiac variables. Hodge et al [31] demonstrated that abdominal lymphatic pump in dogs increased both thoracic duct flow and the leukocyte count in the lymph, thus increasing the mobilization of immune cells. In healthy male medical students, the lymphatic pump increased antibody response to pneumococcal polysaccharide [32]. However, two studies [33, 34] in the elderly assessing the lymphatic pump on antibody response to the influenza vaccine showed no effect.

While the mechanisms are unclear, OMT may have a beneficial effect by enhancing mobilization as has been shown in early mobilization or kinetic bed therapy. Immobilization is a significant mortality-predicting characteristic in the elderly hospitalized with pneumonia [35]. Early mobilization, defined as out of bed at least 20 minutes daily from day one of hospitalization, reduced LOS in hospitalized patients with community-acquired pneumonia [9]. In patients requiring long-term assisted ventilation and in liver transplant recipients, kinetic or continuous lateral rotational bed therapy reduced the prevalence of pneumonia [8, 11]. Unlike these therapies, OMT requires little equipment and can be given to patients unable to get out of bed.

MOPSE used a three-arm design to assess the potential therapeutic contributions of touch, attention, and expectation associated with manual therapies. If the LT outcomes were the same as OMT and both were better than CCO, then the primary benefits of OMT could be attributed to the therapeutic effect of LT. If the LT outcomes were the same as CCO and both were inferior to OMT, then the primary benefits of OMT could be attributed to therapeutic benefit of OMT. However, we found that the LT outcomes generally fell between OMT and CCO, not significantly different from either. This may indicate that the effect of OMT is partly due to whatever is therapeutic in LT and partly due to something inherent to OMT. One exception is the PP60+ subgroup, where both OMT and LT had significantly shorter LOS than CCO. It is interesting that the more elderly subgroup would be responsive to LT. These results highlight the need to compare the manipulative intervention to something more than a sham control since a two-arm study design may underestimate the overall value of the manipulative intervention if the sham control has some therapeutic value. The possible therapeutic components of a sham control (such as LT) are beyond the scope of this discussion.

MOPSE has the following limitations. The sample size was calculated to compare the OMT and CCO groups based on a longer LOS than the current national norm [1]; thus, the sample size may be too small to differentiate between the OMT and control groups. Although the MOPSE PP analysis found between-group differences, an ITT analysis more closely reflects real world conditions. OMT specialists administered only 3 of 8 treatments; the rest were administered by osteopathic resident physicians trained in the protocol. While results may improve with administration by OMT specialists, use of non-specialists trained in the protocol increases the generalizability of the MOPSE results. Although regular training sessions were conducted to standardize treatments among providers, there are no validated objective tools to measure the quality and uniformity of each care provider's technique.

When OMT was administered in accordance with the protocol, reductions were seen in LOS, duration of intravenous antibiotics, and incidence of respiratory failure and death in the OMT group compared to the CCO group. Data suggest a reduced effect from LT compared to OMT in that LOS for LT was between CCO and OMT. These results suggest a potential role for OMT, and possibly for LT, to augment conventional antibiotic therapy in the treatment of pneumonia. This role may become more important with increasing antibiotic resistance, emerging pathogens, the aging global population, the cost of health care, and the likelihood of another influenza pandemic [36–38]. Since developing methods to address these issues is a major health care imperative, the effects of OMT and LT merit further investigation.