Recruitment of potential subject was conducted from March 1, 2004, to August 31, 2006, among patients with terminal cancer. They were referred to the Department of Family Medicine at Korea University Guro-Hospital for hospice care after anticancer-therapy had proven ineffective. We consecutively enrolled patients who were eligible for HRV testing and agreed to participate in this study. Patients with arrhythmia or a cardiac disorder capable of affecting HRV parameters, such as paroxysmal supraventricular tachycardia, atrial fibrillation, recent myocardial infarction, or myocardial ischemia, were excluded. Patients who were unable to maintain a stable supine posture for 5 min were also excluded. Finally, a total of 70 patients were enrolled and we observed their survival duration until death. However, two patients were lost to follow up. Thus, 68 patients were included in the final analysis.

Demographic variables were gender, age, primary cancer site, and previous treatment method. The Karnofsky performance status scale (KPS) and the presence of anorexia and dyspnea, which are parametric constituents of the palliative prognostic score (PaP), were also determined. In previous studies, fasting blood glucose (FBG) and total cholesterol were found to affect HRV (18-21). Therefore, we checked FBG and total cholesterol level, which were measured within one week before and after HRV test. After agreeing to participate in this study, the short term HRV was measured over five minutes using an HRV analyzer, BFM-5000 Plus (Medi-Core Co. Ltd., Seoul, Korea). By using time-domain analysis, mean heart rates, standard deviations of normal-to-normal R-R intervals (SDNN), and root mean square standard deviations of R-R intervals (RMSSD) were calculated. Frequency-domain analysis was applied to measure 5-min total power (TP), low frequency power (LFP), and high frequency power (HFP). HRV parameters were measured by the recommendations published by the Task Force of The European Society of Cardiology and The North American Society of Pacing and Electrophysiology in 1996 (22). The HRV test was explained to patients in a quiet, pleasant medical office and conducted over 5 min after applying four limb leads. Patients were advised not to smoke or consume caffeine for 2 hr prior to HRV testing, and tests were conducted more than 2 hr after a meal. Survival duration was defined from HRV testing until the date of death. This study was approved by the Institutional Review Board (IRB) of Korea University Guro Hospital (IRB No. GR0514-002).

Univariate analysis of log-rank test for categorical variables or Wald test for continuous variables were performed to identify variables that significantly related to survival duration. The relationship between HRV measures and the variables related to HRV, fasting blood glucose and total cholesterol, was examined using correlation analysis. Finally, Cox's proportional hazard regression model was used to determine the effects of dyspnea, anorexia, KPS, mean heart rates, and SDNN, which were significant in univariate analysis, on survival duration after adjusting for gender, age, fasting blood glucose, and total cholesterol. All tests were two-tailed with 5% level of significance. SAS version 9.1 (SAS Institute Cary, NC, USA) was used for statistical analysis.

We analyzed 68 patients (34 men and 34 women) aged between 26 and 84 yr old. The most common primary cancer sites were: lung (20.6%), stomach & esophagus (14.7%), hepatobiliary (14.7%), gynecologic (13.2%), and colorectal (10.3%). Previous treatment types were (in descending order) surgery+chemotherapy (29.4%), no treatment (19.1%), chemotherapy (14.7%), surgery+ chemotherapy+radiotherapy (11.8%), and chemotherapy+radiotherapy (11.8%) (Table 1). The distribution of HRV parameters are described in Table 2. HRV levels were considerably lower than the general Korean population except mean heart rate (23).

Gender, age, fasting blood glucose, and total cholesterol were independent of survival duration (P=0.584, 0.064, 0.421, 0.467, respectively). However, patients with higher KPS survived longer than those with lower scores (P<0.001). Patients with symptoms of dyspnea or anorexia had shorter survival durations than those without symptoms (P=0.037, 0.015, respectively). Fig. 1A-C showed Kaplan-Meier Survival curves of groups subdivided by these clinical parameters.

Correlation analysis between HRV parameters and fasting blood glucose or total cholesterol showed that all of the HRV parameters except mean heart rate were significantly correlated with total cholesterol (for SDNN, correlation coefficients 0.43, P<0.001).

As in previous studies on the association between HRV and prognosis, patients were dichotomized by the highest quartile (75%) of each HRV parameter (11, 14-16). However, 100 beats per min (bpm) was adopted as the reference level for grouping by mean heart rate because it is generally accepted as the definition of tachycardia. Subsequently, significant differences in survival duration were found between patients with an SDNN of <21.3 ms or ≥21.3 ms and between patients with mean heart rate of <100 bpm or ≥100 bpm (Table 3). Fig. 1D, E showed Kaplan-Meier survival curves of patients dichotomized on mean heart rate and SDNN, respectively. For HRV parameters other than SDNN and mean heart rate, survival durations after dichotomization were not significantly different.

To identify factors associated with survival duration among all variables examined, we performed multivariate regression analysis using Cox's proportional hazard model. The final model included gender, age, fasting blood glucose, total cholesterol, KPS, dyspnea, anorexia, mean heart rate, and SDNN. In addition to these variables, we added an interaction variable (inter=Cholesterol×SDNN) to the final regression model because total cholesterol level was significantly correlated with SDNN, indicating a possible interaction. Finally, SDNN, KPS, the interaction variable, dyspnea, and anorexia were significant variables by Backward selection (Table 4).