Decreased risk of pneumonia in stroke patients receiving acupuncture: A nationwide matched-pair retrospective cohort study

Background Acupuncture treatment is common among stroke patients, but there is limited information available on whether acupuncture effectively prevents post-stroke pneumonia. The aim of this study was to analyze the differential risk of pneumonia after stroke between patients who did and did not receive acupuncture after discharge. Methods We used the Taiwan National Health Insurance Research Database to conduct a retrospective cohort study using propensity score matched-pairs of new stroke patients in 2000–2004 who did and did not receive acupuncture post-stroke. Both cohorts were followed up until the end of 2009 for new-onset pneumonia. After correcting for immortal time bias, the incidence and adjusted hazard ratios (HRs) with 95% confidence intervals (CIs) of pneumonia associated with acupuncture use were calculated using multivariate Cox proportional hazard models. Results Overall, 12557 stroke patients with 12557 paired controls were included in the analysis; pneumonia was diagnosed in 6796 (27.1%). Stroke patients receiving acupuncture had a lower incidence of pneumonia than those without acupuncture (53.4 vs. 58.9 per 1000 person-years), with an adjusted HR of 0.86 (95% CI 0.82–0.90). The association between pneumonia risk and acupuncture use was significant in men (HR 0.92, 95% CI 0.86–0.98) and women (HR 0.79, 95% 0.70–0.82) and was also observed in every age group from 20–79 years. Conclusion Stroke patients receiving acupuncture had a lower risk of pneumonia than those who did not. Further randomized control studies are needed to validate the protective effect of acupuncture on the risk of pneumonia among stroke patients.

Introduction Pneumonia remains the most common serious medical complication among stroke patients and occurs in up to 22% of all stroke cases [1]. Post-stroke pneumonia is known to affect patient outcomes including the occurrence of other complications [2], prolonged hospitalizations, and increased dependency after discharge [1,3]. Pneumonia is also closely related to mortality and highly contributes to the rate of death after stroke [1,4]. In-hospital care for stroke-associated pneumonia in the United States has previously been estimated to cost USD 459 million per year [2], and the recent costs are estimated to be more than twofold higher [3]. Sociodemographic factors including older age, male gender, admission from nursing homes, and preadmission dependency are independent predictors of post-stroke pneumonia [1,4]. Physical status including comorbid chronic obstructive pulmonary disease and coronary artery disease, as well as stroke-related parameters, such as stroke severity, dysphagia, infarct location, impairment of protective reflexes, neurological deficits, and mechanical ventilation, are also known to affect outcomes [1,4,5].
Acupuncture has been accepted as a treatment option and applied in stroke survivors worldwide [6][7][8][9][10]. However, there is limited information on the association between acupuncture and risk of post-stroke pneumonia, and the preventive effects of acupuncture on poststoke pneumonia remain unclear [11,12]. The purpose of this retrospective cohort study was to assess the protective effect of pneumonia in stroke patients who did and did not receive acupuncture after discharge.

Source of data
Reimbursement claims data from the Taiwan National Health Insurance Program, which was implemented in March 1995 and covers 23 million (> 99%) Taiwan residents, were used in this study. The Taiwan National Health Research Institutes established a National Health Insurance Research Database (NHIRD) that records all beneficiaries' inpatient and outpatient services, including their demographics, primary and secondary diagnoses, procedures, prescriptions and medical expenditures. The validity of NHIRD has been evaluated, and the results have been published globally [6][7][8][9][10].

Ethical statement
The electronic database was decoded and patient identification data were scrambled to ensure confidentiality; informed consent was thus exempted. We conducted this study in accordance with the Declaration of Helsinki. This study was also evaluated and approved by the Institutional Review Board of E-DA Hospital, Kaohsiung, Taiwan (EDA-JIRB-2014012; EDA-JIRB-2017004).

Study design
We used the same dataset from the NHIRD in our previous studies [6][7][8][9][10]. Briefly, of the 23 million people in Taiwan, 226,699 patients aged !20 years with newly diagnosed stroke who were admitted to the hospital between Jan 1, 2000, and Dec 31, 2004, were considered eligible subjects. To identify eligible stroke patients and pneumonia events, the following exclusion criteria were applied: (1) patients who died during the index stroke admission; (2) patients who stayed in the hospital for !30 days during the index stroke admission; and (3) patients who had been diagnosed with pneumonia within 6 months prior to the index stroke admission. The diagnosis of stroke was validated as described in previous studies [6][7][8][9][10]. In total, 12557 stroke survivors received at least two courses (one course consisted of six consecutive treatments) of acupuncture treatment during the follow-up period until the end of 2009; these patients were defined as the acupuncture group (those with only one course of acupuncture were excluded).
Stroke patients without acupuncture were selected using a matched pair procedure with propensity scores (exposure vs. non-exposure ratio = 1:1). For the non-acupuncture group, stroke patients were followed up from the date of discharge after stroke admission (index date) until December 31, 2009, a pneumonia event, loss to follow-up, or death. For the acupuncture group, stroke patients were followed up from the first date of acupuncture treatment after the stroke admission (index date) until December 31, 2009, a pneumonia event, loss to follow-up, or death. In the acupuncture group, the time period between the discharge date and the date of first acupuncture treatment after stroke admission represented the immortal time. Therefore, immortal time bias (resulting from an overestimation of the intervention's beneficial effects) was avoided in this study. Follow-up time, in person-years, was calculated for each stroke patient from the index date to the end point. We compared the risk of pneumonia between the matched pairs of stroke patients and the controls during the follow-up period.

Statistical analysis
We estimated the propensity scores for acupuncture use by a non-parsimonious logistic model, considering all significant clinical covariates proposed. We applied a structured iterative approach to refine this model and achieve a balance of covariates within matched pairs, using a greedy matching strategy with the nearest-neighbor algorithm. This method has been reported to remove 98% of the bias from measured covariates [13]. Chi-square tests and t tests were used to measure the distribution of covariates between stroke patients with and without acupuncture use, and a p-value <0.05 indicated a meaningful difference.
We then utilized multivariate Cox proportional hazard models to control for confounders and calculated adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for pneumonia in terms of acupuncture use. We performed further subgroup analyses for each stratum of sex, age group, and subtype of stroke using the full model without the stratifying variable. The number of sessions of acupuncture use was also considered an independent factor associated with post-stroke pneumonia in the multivariate Cox proportional hazard model. Kaplan-Meier analysis was used to calculate the cumulative probability of pneumonia in stroke patients with and without acupuncture use, and the group differences between the cumulative incidence curves were evaluated by log-rank test. All analyses were performed using Statistical Analysis Software version 9.1 (SAS Institute Inc., Cary, NC, USA). A two-sided p value <0.05 was considered statistically significant.

Results
After propensity score matching (Table 1), there were no significant differences between groups in age, sex, stroke subtype, income, urbanization, diabetes, hypertension, hyperlipidemia, mental disorder, liver cirrhosis, ischemic heart disease, urinary tract infection, heart failure, COPD, epilepsy, renal dialysis, stay in intensive care unit, neurosurgery, hospitalized rehabilitation, type of hospital, cardiovascular medications, or length of hospital stay.
During the follow-up period (Table 2), the acupuncture group had a lower incidence of pneumonia than the control group (53.4 vs. 58.9 per 1000 person-years), with an adjusted HR of 0.86 (95% CI 0.82-0.90). The log-rank test (Fig 1) indicated that the acupuncture group had a lower probability of pneumonia than the control group (p <0.0001).

Discussion
Using data from the NHIRD, we found a reduced risk of pneumonia in newly diagnosed stroke patients when they received acupuncture after discharge in this nationwide, propensity score-matched, retrospective cohort study. Our analyses revealed that the reduction in risk tend to be most significant in women, younger patients, and patients with hemorrhagic stroke. Studies focusing on the preventive effects of acupuncture on post-stroke pneumonia are relatively rare [12], and the conclusions have been inconclusive due to small sample sizes and  methodological deficiencies [11]. We previously examined the effects of in-hospital traditional Chinese medicine (TCM) on preventing post-stroke complications within 6 months after discharge from stroke hospitalization and found a significant reduction in pneumonia by adjuvant TCM therapy, 96% of which consist of acupuncture [6]. In this study, we further focused on the correlation between acupuncture and post-stroke pneumonia and extended the followup period in a significantly larger cohort. Interestingly, the tendency toward stronger effects in women, younger groups, and non-hemorrhagic stroke patients was also noted in our previous research on post-stroke acute myocardial infarction and epilepsy [8,9]. Several factors might explain the mechanisms behind our findings. First, acupuncture may increase regional cerebral blood flow [14][15][16], which may exert some neuroprotective effects [9,17]. Increased cerebral blood flow by acupuncture has been associated with enhanced perfusion mediated by endothelial nitric oxide synthase [18], attenuated angiotensin II activity [19], and angiogenesis promotion in animal models [20] and with improved carbon dioxide reactivity [16], decreased sympathetic activity [21], and increased endothelial progenitor cell count and functioning in human studies [22]. The hypothesized biophysical mechanisms of the neuroprotective effects include angiogenesis promotion, alleviation of inflammatory responses, regulation of the blood brain barrier, and inhibition of apoptosis [23].
Second, dysphagia has been considered a risk factor for post-stroke pneumonia and subsequent mortality [24]. The beneficial effects of acupuncture on improving dysphagia have been investigated in previous studies [12,25,26]. Based on the results of our study and of previous related reports, we postulate that acupuncture might help decrease post-stroke pneumonia through its beneficial effects on dysphagia.
Third, the suppression of immune function after stroke should be considered in light of three aspects. First, central nervous system injuries, including stroke, are known to reduce and disrupt peripheral immune cells and suppress their functions [27,28]. Additionally, stroke is associated with various common and persistent painful conditions [29,30]. Most patients' pain has moderate or severe effects on their daily lives and remains untreated or becomes treatmentresistant [30]. However, we have no related data to show the effects of acupuncture on poststroke pain and it needs further researches [31]. Finally, stroke survivors commonly experience varying degrees of mood disorders including depression and anxiety [32,33]. Although the prevalence of these conditions fluctuates over time [34], post-stroke emotional stresses remain a common problem several years after the event [35]. Chronic pain syndromes and emotional stresses are known to suppress immune function including decreased peripheral natural killer cell count and activities and increased metastasis of tumor cells [36]. Acupuncture can directly regulate specific [37] and nonspecific immunity [38]. Additionally, it has long been used, and proven effective, to cure chronic pains [39] and emotional disorders [40]. We postulate that acupuncture might correct for the immunosuppression that occurs after stroke either directly, by modulating the specific and non-specific immunity, or indirectly, by ameliorating post-stroke pain and emotions, thereby decreasing the incidence of post-stroke pneumonia.
The strengths of this NHIRD-based study include the large sample size, long follow-up period, and minimized selection bias. Second, we applied propensity score matching to minimize residual influences from socio-demographic factors and coexisting medical conditions, and we also used multivariate Cox proportional hazard models to control for confounding effects. Third, to reduce the immortal time bias, we corrected for immortal time in the acupuncture group when calculating person-years. Finally, we performed subgroup analyses to test the robustness of our estimates and demonstrated the trends in differential effects across the tested strata.
Some limitations of this study should be addressed. First, we used insurance claims data without detailed information on clinical risk scores, lesion characteristics, biochemical measures, or patient lifestyles. Second, information on the actual acupuncture points and manipulations applied in treatment was not available. Inadequately standardized management protocols might have affected the accuracy of the estimated treatment effects and should be considered a study limitation. Third, although the accuracy of diagnostic codes in the database has been verified in previous studies [6][7][8][9][10], the validity of comorbidity codes is another potential limitation. In addition, we cannot infer causality regarding the association between acupuncture and post-stroke pneumonia because our study was observational. Finally, although we used a propensity score matching procedure and multivariate regression models to adjust for the covariates in stroke patients with and without acupuncture, residual confounding remains a possibility.
In conclusion, we showed that stroke patients receiving acupuncture had a reduced risk of pneumonia in this nationwide retrospective cohort study. The potential beneficial effects of acupuncture as a supplemental measure to reduce the occurrence of pneumonia deserve further exploration in well-designed controlled clinical trials with stringent acupuncture protocols and blinded sham control designs.