Abstract
Context: Osteopathic manipulative treatment (OMT) is a promising adjunctive treatment for older adults hospitalized for pneumonia.
Objective: To report subgroup analyses from the Multicenter Osteopathic Pneumonia Study in the Elderly (MOPSE) relating to hospital length of stay (LOS), ventilator-dependent respiratory failure rate, and in-hospital mortality rate.
Design: Multicenter randomized controlled trial.
Setting: Seven community hospitals.
Participants: Three hundred eighty-seven patients aged 50 years or older who met specific criteria for pneumonia on hospital admission.
Interventions: Participants were randomly assigned to 1 of 3 groups that received an adjunctive OMT protocol (n=130), a light touch (LT) protocol (n=124), or conventional care only (CCO) (n=133).
Main Outcome Measures: Outcomes for subgroup analyses were LOS, ventilator-dependent respiratory failure rate, and in-hospital mortality rate. Subgroups were age (50-74 years or ≥75 years), Pneumonia Severity Index (PSI) class (I-II, III, IV, or V), and type of pneumonia (community-acquired or nursing–home acquired). Data were analyzed by intention-to-treat and per-protocol analyses using stratified Cox proportional hazards models and Cochran-Mantel-Haenszel tests for general association.
Results: By per-protocol analysis of the younger age subgroup, LOS was shorter for the OMT group (median, 2.9 days; n=43) than the LT (median, 3.7 days; n=45) and CCO (median, 4.0 days; n=65) groups (P=.006). By intention-to-treat analysis of the older age subgroup, in-hospital mortality rates were lower for the OMT (1 of 66 [2%]) and LT (2 of 68 [3%]) groups than the CCO group (9 of 67 [13%]) (P=.005). By per-protocol analysis of the PSI class IV subgroup, the OMT group had a shorter LOS than the CCO group (median, 3.8 days [n=40] vs 5.0 days [n=50]; P=.01) and a lower ventilator-dependent respiratory failure rate than the CCO group (0 of 40 [0%] vs 5 of 50 [10%]; P=.05). By intention-to-treat analysis, in-hospital mortality rates in the PSI class V subgroup were lower (P=.05) for the OMT group (1 of 22 [5%]) than the CCO group (6 of 19 [32%]) but not the LT group (2 of 15 [13%]).
Conclusion: Subgroup analyses suggested adjunctive OMT for pneumonia reduced LOS in adults aged 50 to 74 years and lowered in-hospital mortality rates in adults aged 75 years or older. Adjunctive OMT may also reduce LOS and in-hospital mortality rates in older adults with more severe pneumonia. Interestingly, LT also reduced in-hospital mortality rates in adults aged 75 years or older relative to CCO. (ClinicalTrials.gov number NCT00258661)
Pneumonia is a major infectious disease afflicting elderly patients. While antibiotic therapy is the mainstay of treatment, the emergence of resistant bacteria is worrisome
1-3 and provides an impetus to explore adjunctive treatments. There are a few promising adjunctive nonpharmacologic treatments for pneumonia. Early mobilization (getting the patient out of bed daily from the first day of hospitalization) modestly reduced hospital length of stay (LOS) in community-acquired pneumonia.
4,5 A 2014 review
6 of 26 articles found that early mobilization reduced LOS and medical complications in an intensive care unit. One hospital-based clinical trial found that blowing bubbles into a bottle containing 10 cm of water, 10 times per day, reduced LOS in community-acquired pneumonia.
7 However, reviews of incentive spirometry have shown no definite benefit for preventing pulmonary complications after upper abdominal surgery
8 or complications after coronary artery bypass graft.
9 A meta-analysis of rotational bed therapy found that it decreased the incidence of pneumonia in selected critically ill patients but had no effect on duration of mechanical ventilation, number of days in intensive care, or hospital mortality.
10 Studies of classic chest physiotherapy for pneumonia have generally failed to show benefit.
11,12
Osteopathic manipulative treatment (OMT) is another nonpharmacologic adjunctive treatment for patients with pneumonia. Preliminary randomized controlled clinical trials suggested that OMT in hospitalized older patients reduced LOS and duration of intravenous antibiotic therapy.
13,14 Two systematic reviews of manual therapies, from perspectives outside the osteopathic medical profession, concluded that OMT is a promising treatment for patients with pneumonia.
15,16 A distinctive aspect of OMT is that many techniques were developed in the pre–antibiotic era specifically for the management of pneumonia.
17,18 The Multicenter Osteopathic Pneumonia Study in the Elderly (MOPSE) sought to assess the efficacy of OMT as an adjunctive treatment for pneumonia in hospitalized older patients.
19 A detailed description of the MOPSE protocol
18 and the primary study outcomes
19 have been previously published. For the primary MOPSE outcomes, no between-group differences in LOS were found when intention-to-treat (ITT) analysis was used (n=387). When per-protocol (PP) analysis was used (n=318), OMT showed a reduction of 1 day in LOS and 0.5 days in the closely related duration of intravenous antibiotic therapy compared with conventional care only (CCO).
19 Treatment end points of ventilator-dependent respiratory failure and in-hospital mortality rates were 8% and 6% lower, respectively, in the OMT group compared with the CCO group for the PP analysis.
The purpose of the current article is to report subgroup analyses from MOPSE relating to LOS, ventilator-dependent respiratory failure rate, and in-hospital mortality rate. Because the benefits of OMT may vary by age or severity of illness, subgroup analyses were conducted to provide insight into characteristics of patients who were more likely to benefit from OMT. Differences between MOPSE treatment groups—OMT, light-touch (LT), and CCO—were evaluated for subgroups of participants based on their age, Pneumonia Severity Index (PSI), and type of pneumonia. Our hypothesis was that participants in MOPSE will have differing outcomes for LOS, ventilator-dependent respiratory failure rate, and in-hospital mortality rate for different ages, severity of illness, and type of pneumonia.
A multicenter randomized controlled clinical trial (ClinicalTrials.gov number NCT00258661), MOPSE was conducted between 2004 and 2007 at 7 community hospitals and enrolled patients aged 50 years or older who met specific criteria for pneumonia on their admission to the hospital. Details of the MOPSE protocol
18 and methods
19 have been previously published. The key methodologic elements related to our subgroup analyses are summarized here.
After providing informed consent, participants were randomly assigned to 1 of 3 groups: OMT, LT (sham control), or CCO (standard care control). Those participants assigned to the OMT or LT groups received standardized treatment protocols twice daily
18 for the duration of their hospital stay in addition to conventional care. The OMT protocol techniques included thoracolumbar soft tissue, rib raising, doming of the diaphragm with myofascial release, cervical spine soft tissue, suboccipital decompression, thoracic inlet myofascial release, thoracic lymphatic pump, and pedal pump. The LT protocol applied light touch to the same body regions in the same sequence and duration as the OMT protocol. The duration of each treatment session was approximately 20 minutes. The outcomes used for these subgroup analyses were LOS (defined by the date and time given on the admission and discharge orders), ventilator-dependent respiratory failure rate, and in-hospital mortality rate.
Data were analyzed by ITT analysis of all participants and by PP analysis of participants who finished the study without missing any protocol treatments. Data were analyzed for subgroups based on age (50-74 years or 75 years or older), PSI class (I-II, III, IV, or V), and type of pneumonia (community-acquired or nursing home–acquired). All analyses were stratified on the 5 study sites. The distribution of the participants across the treatment groups within each subgroup was compared with the distribution of the total participants across treatment groups using χ2 goodness-of-fit tests. To test the hypotheses regarding treatment group differences in LOS (measured in days), we used stratified Cox proportional hazards models, stratifying by study site. The LOS for participants who died in the hospital, had ventilator-dependent respiratory failure, or withdrew from the study was censored. The only variables included in the Cox proportional hazards model were the interaction of the treatment group with the subgroup, the main effect for the treatment group, and the main effect for the subgroup. The proportional hazards assumption was met within each of the strata; residuals were not evaluated. The treatment groups were compared for each level of the subgroup variable, and all pairwise comparisons between treatment groups were performed when a significant difference was observed. Cochran-Mantel-Haenszel tests for general association, stratifying on study site, were used to compare treatment groups on ventilator-dependent respiratory failure and in-hospital mortality rates for each subgroup separately, and all pairwise comparisons between treatment groups were performed when a significant difference was observed. P values less than .05 were considered statistically significant. The statistical analyses were performed using SAS version 9.1 (SAS Institute, Inc).
Sample sizes for the treatment groups within each subgroup are provided in
Table 1. The distribution of the participants across the treatment groups within each subgroup was consistent with the distribution of the total participants across the treatment groups (
P≥.50).
By ITT analysis, there was no significant difference in LOS between the treatment groups for either the age 50 to 74 years (
P=.94) or 75 years or older (
P=.48) subgroups (
Table 2;
Figures 1A and
1C, respectively). By PP analysis, there was a significant difference between the treatment groups for the age 50 to 74 years subgroup (
P=.006) (
Table 2;
Figure 1B); LOS was shorter for the OMT group than the LT and CCO groups. There was no significant difference by PP analysis in LOS between the treatment groups for the age 75 years or older subgroup (
P=.18) (
Table 2;
Figure 1D). When only those with community-acquired pneumonia were included in the analysis of LOS for the age subgroups, the results were consistent with the results for all participants.
There were no significant differences by ITT analysis between the treatment groups on ventilator-dependent respiratory failure and in-hospital mortality rates for the age 50 to 74 years subgroup (
P=.60 and
P=.77, respectively) (
Table 3;
Figure 2A). By ITT analysis of the age 75 years or older subgroup, the treatment groups were not significantly different for ventilator-dependent respiratory failure rate (
P=.20) but were significantly different for in-hospital mortality rate (
P=.005) (
Table 3;
Figure 2C); rates were lower for the OMT and LT groups than the CCO group. Results for the PP analysis were similar to the ITT analysis (
Table 3;
Figures 2B and
2D), as were the results when only those with community-acquired pneumonia were included in the analysis.
By ITT analysis, there was no significant difference in LOS between the treatment groups for any of the PSI class subgroups (
P≥.06) (
Table 2;
Figures 3A,
3C,
3E, and
3G). By PP analysis, there was a significant difference between the treatment groups in LOS for the PSI class IV subgroup (
P=.01) (
Table 2;
Figure 3F); LOS was shorter for the OMT group than the CCO group, but neither was different from the LT group. There was no significant difference in LOS between the treatment groups for PSI class I-II (
P=.14), III (
P=.26), or V (
P=.40) subgroups (
Table 2;
Figures 3B,
3D, and
3H, respectively). When only those with community-acquired pneumonia were included in the analysis of LOS for the PSI class subgroups, the results were consistent with the results for all participants.
There were no significant differences by ITT analysis between the treatment groups in ventilator-dependent respiratory failure rate for any of the PSI class subgroups (
P≥.07) (
Table 3;
Figures 4A,
4C,
4E, and
4G). By ITT analysis comparing the treatment groups in in-hospital mortality rate, the treatment groups were significantly different for the PSI class V subgroup (
P=.05) (
Table 3;
Figure 4G); the rate was lower for the OMT group than the CCO group, but neither was different from the LT group. There was no significant difference between the treatment groups in in-hospital mortality rate for the other PSI class subgroups (
P≥.18) (
Table 3;
Figures 4A,
4C, and
4E). Results for the PP analysis were similar to the ITT analysis with one exception (
Table 3;
Figures 4B,
4D,
4F, and
4H). For the PSI class IV subgroup, there was a significant difference between the treatment groups in ventilator-dependent respiratory failure rate (
P=.05) (
Table 3;
Figure 4F); the rate was lower for the OMT group than the CCO group, but neither was different from the LT group.
When only those with community-acquired pneumonia were included in the ITT analysis, there was a significant difference between the treatment groups in ventilator-dependent respiratory failure rate for the PSI class IV subgroup (P=.05); the rate was lower for the OMT group (0 of 38 [0%]; 95% CI, 0%-9%) than the CCO group (5 of 37 [14%]; 95% CI, 5%-29%), but neither was different from the LT group (2 of 31 [6%]; 95% CI,1%-21%). There were no significant differences between the treatment groups in ventilator-dependent respiratory failure rate for the other PSI class subgroups (P≥.47) or for in-hospital mortality rate for any of the PSI class subgroups (P≥.29). For the PP analysis of those with community-acquired pneumonia, there was a significant difference between the treatment groups in ventilator-dependent respiratory failure rate for the PSI class IV subgroup (P=.05); the rate was lower for the OMT group (0 of 40 [0%]; 95% CI, 0%-9%) than the CCO group (5 of 50 [10%]; 95% CI, 3%-22%), but neither was different from the LT group (1 of 35 [3%]; 95% CI, 0%-15%). There were no significant differences between the treatment groups in ventilator-dependent respiratory failure rate for the other PSI class subgroups (P≥.47). There was a significant difference between the treatment groups in in-hospital mortality rate for the PSI class V subgroup (P=.05); the rate was lower for the OMT group (0 of 12 [0%]; 95% CI, 0%-26%) than the CCO group (6 of 17 [35%]; 95% CI, 14%-62%), but neither was different from the LT group (2 of 10 [20%]; 95% CI, 3%-56%). There were no significant differences between the treatment groups in in-hospital mortality rate for the other PSI class subgroups (P≥.06).
By ITT analysis, there was no significant difference in LOS between the treatment groups for the community-acquired pneumonia subgroup (
P=.79) or the nursing home–acquired pneumonia subgroup (
P=.53) (
Table 2). The results for the PP analysis were similar to the ITT analysis.
There were no significant differences by the ITT analysis between the treatment groups in ventilator-dependent respiratory failure and in-hospital mortality rates for the community-acquired pneumonia or nursing home–acquired pneumonia subgroups (
P≥.14) (
Table 3). Results for the PP analysis were similar to the ITT analysis.
Even though the average hospital LOS for pneumonia has dramatically decreased since the implementation of the prospective payment system in the 1980s,
20 it remains a clinically important outcome measure.
21,22 For patients hospitalized for pneumonia, our findings suggest that OMT reduced LOS for those aged 50 to 74 years (
Figure 1B) and lowered the risk of death while hospitalized for those aged 75 years or older (
Figures 2C and
2D). Osteopathic manipulative treatment also reduced the in-hospital mortality rates of those with the highest severity of illness (
Figures 4G and
4H). The mortality data are particularly robust because these differences were shown by ITT and PP analyses. The results for the community-acquired pneumonia subgroup were similar to overall results; OMT lowered ventilator-dependent respiratory failure rates for those with the second highest severity of illness by PP analysis and reduced in-hospital mortality rates for those with the highest severity of illness by ITT and PP analyses. These findings support previous systematic reviews
15,16 that suggested OMT has potential as an adjunctive modality to conventional antibiotic therapy. Likewise, our finding that OMT reduced LOS in the younger subgroup is encouraging. Overall, our findings appear to support the greater use of OMT as an adjunctive modality for pneumonia and the need for more research.
When interpreting these results, it is important to understand the characteristics of ITT and PP analyses. Intention-to-treat analysis has better generalizability to real-world populations because in clinical practice patients will miss or decline treatment sessions for various reasons. Thus, this analysis reflects the overall effect of an intervention on a population because everyone who was assigned to a treatment is included in the analysis regardless of whether everyone adhered to the treatment regimen. As a result, ITT analysis is a stricter assessment and is considered the primary analysis. An advantage of PP analysis is that it is a better measure of what a treatment can do for a patient because it represents the effect of treatment for those who received the treatment as prescribed. Nonadherence to the treatment protocol is common in clinical trials. A recent analysis of 100 clinical trials found that 98 reported nonadherence to the treatment protocol, and most clinical trials report their findings using some variant of PP analysis.
23 The use of both ITT and PP analyses demonstrates the rigor with which the current study was performed.
Interestingly, LT reduced in-hospital mortality rates from pneumonia in those aged 75 years or older relative to the CCO group. Considering that LT was intended to be a sham control, this finding was unexpected. It was the only instance in which LT showed clear benefit in our subgroup analyses, so it could be an erroneous finding. If the finding is not erroneous, then it needs to be interpreted within the context of the current evidence. Other clinical trials of OMT have reported beneficial findings from the sham arm of the study. In a chronic low back pain clinical trial, both OMT and LT showed greater improvement in back pain and function relative to the no-treatment control group and no significant benefit of OMT over LT.
24 In a clinical trial of OMT in pregnancy, OMT and placebo ultrasound groups had significantly improved back pain and back-related functioning relative to the usual-care group.
25 These sham arms generate a significant amount of touch or superficial stimulation that could have some physiological therapeutic response separate from a placebo mechanism. For MOPSE, a placebo effect is an incomplete explanation because only 44% of LT participants could correctly identify their group assignment,
19 thus minimizing the influence of anticipation or belief in the treatment. However, when considering the area of attention, the OMT and LT groups had 2 more physician visits per day than the CCO group.
18 During these visits the clinician touched the patient for a total of 30 minutes daily.
18 This additional attention could have been a placebo factor that influenced the MOPSE study outcomes. It is also possible that there was a direct, physical factor in the stimulation of touch that caused a therapeutic response rather than a cognitive placebo-type response.
The extra attention from being randomly assigned to either the OMT or LT group also means that these participants were more likely to be moved in and out of bed relative to the CCO group. Participants who were out of bed had to return to bed and those sitting up or lying prone in bed had to be repositioned. After OMT or LT treatment, the participants could return to their previous position. The net effect is that increased activity may have created an early mobilization effect relative to the CCO group. Evidence suggests that early mobilization (getting out of bed daily from the beginning of the hospital stay) has a therapeutic effect in patients with pneumonia.
4-6 Patients with community-acquired pneumonia who got out of bed to sit in a chair once per day from day 1 of hospitalization modestly reduced their LOS.
4 However, the frequency of bed transfers and repositionings are not known because they were not recorded during MOPSE; thus, the magnitude of any early mobilization effect will remain unclear.
Given the very short mean national LOS for pneumonia (about 4.5 days), it is challenging for any therapy to reduce LOS. Before the introduction of the prospective payment system in the United States in the 1980s, the mean national LOS for pneumonia was approximately 12 days.
20 By 2007, the prospective payment system had reduced the mean national LOS to 5 days for all ages and to 5.4 days for those aged 65 years or older.
26 Three previously published studies of OMT in hospitalized older patients have reflected this progressive decrease in LOS.
13,14,19 In those studies,
13,14,19 where the mean LOS was longer than the current average, OMT demonstrated a more obvious reduction in the number of days patients were hospitalized: reductions in LOS from OMT are more clearly recognized when LOS is longer. All therapeutic interventions require a minimal duration of treatment for optimal effectiveness. For example, for most infectious diseases, only 1 day of antibiotics would be too short for optimal effect. Fortunately, antibiotic therapy can easily be extended after hospital discharge, but it is more difficult to arrange for adjunctive OMT after hospital discharge. The optimal duration of treatment for adjunctive OMT in pneumonia is likely to be more than 4 days because OMT theoretically works by enhancing host defenses rather than by directly attacking the pathogen.
27 Given the current short duration of LOS for pneumonia, the effect of extending OMT after hospital discharge should be investigated to fully understand the benefit of adjunctive OMT in the management of pneumonia.
Adjunctive OMT has been shown to reduce LOS for other medical conditions. In a small but well-designed clinical trial, OMT reduced hospital LOS for pancreatitis.
28 A more recent retrospective cohort study found that OMT reduced LOS for patients with postoperative ileus.
29 In another exploratory study,
30 OMT reduced LOS and the occurrence of gastrointestinal symptoms in premature infants.
An inherent limitation of subgroup analysis is that sample sizes are reduced because the original cohort is subdivided into smaller groups, which may reduce the statistical power to detect between-group differences. The increased number of comparisons increases the possibility of erroneous findings, and the smaller subgroups increase the possibility of randomization failure. Another limitation is that MOPSE relied solely on randomization to produce comparable participants in each group and did not use a risk adjustment model from claims data. Although the PSI serves as a measure of successful randomization, it is not designed as a risk adjustment tool. Also, although the mortality outcomes are interesting, they need to be interpreted with particular caution because the subgroup sample sizes are smaller and more open to bias. Although PP analysis is a good reflection of what an intervention can do if followed, it has less generalizability to the real world and deviations from a protocol can introduce bias. Because of these limitations, these MOPSE subgroup findings should be interpreted with caution.
Our subgroup analyses suggested that adjunctive OMT for pneumonia reduced LOS in participants aged 50 to 74 years when administered as prescribed and lowered in-hospital mortality rates for those aged 75 years or older. Adjunctive OMT also appeared to be beneficial for participants with a higher severity of illness, reducing LOS in PSI class IV when administered as prescribed and reducing in-hospital mortality rates in PSI class V pneumonia relative to the CCO group but not to the LT group. These findings are encouraging but should be interpreted with caution because the subgroup sample sizes are relatively small. Subgroup analysis also showed that LT reduced in-hospital mortality rates in participants aged 75 years or older. Possible explanations for both OMT and LT reducing in-hospital mortality rates in participants aged 75 years or older include anticipatory belief, a physiologic response to attention and touch, or an early mobilization effect. The findings indicate that more research is needed for niche applications of adjunctive OMT for pneumonia, such as for patients at risk for longer hospital stays and those with higher severity of illness. Likewise, there is a need to solve the logistical and availability challenges of extending OMT beyond the acute-care hospital setting to outpatient and subacute rehabilitation settings.
We thank MOPSE project manager, Kenneth Pamperin, MS, and the numerous physicians and staff who were instrumental in the performance of this study. We also thank Deborah Goggin, MA, scientific writer at A.T. Still University, for editorial assistance.