Avoiding overfitting

METHODOLOGY

STUDY LOCATION

The study was carried out at the Cardiology Unit of the Department of Medicine of the University College Hospital, Ibadan.

SUBJECTS

Cases for the study were patients with congestive cardiac failure secondary to hypertensive, rheumatic, valvular, ischemic heart diseases and the cardiomyopathies on hospital admission at the University College Hospital, as well as those recruited from the medical outpatient clinics. The controls were subjects of comparable age and sex, not in heart failure and with no overt cardiovascular disease. This second group was recruited from volunteers among healthy hospital staff as well as non-cardiac follow-up cases from the outpatient clinics.

INCLUSION CRITERIA

Patients with congestive heart failure (defined by the Framingham criteria⁷⁸ ) secondary to hypertensive, rheumatic, valvular and ischemic heart diseases as well as the cardiomyopathies, and controls with echocardiographically-determined normal cardiac structure and function, of both sexes, were recruited for the study.

EXCLUSION CRITERIA

Subjects excluded from the study were those with chronic renal failure, diabetes mellitus, and history of significant alcohol intake, as these are known independently to affect HRV.

SAMPLE SIZE

The sample size was calculated using the formula⁷⁹: n =

(

) (

)

/

Where Δ = |μ2– μ1|. The means and variances of the respective groups are (μ1, σ1 2) and (μ2, σ2 2). The level of significance α used was 0.05, while β was 90%. This study was powered to detect at least a 15 ms difference in SDNN and SDANN between the group with congestive cardiac failure and normal controls, using the standard deviation from a similar study⁸⁰. This brings the sample size to 55 subjects per group or 110 for the two groups. Ultimately, 74 patients with heart failure and 69 normal subjects were recruited into the study.

CONSENT

Informed consent was obtained from each of the subjects prior to participation in the study (see Appendix III).

ETHICAL APPROVAL

This study was performed in accordance with the declaration of Helsinki and approval was obtained from the University College Hospital/University of Ibadan joint ethical committee.

CLINICAL ASSESSMENT

The bio-data and relevant history were obtained, and thorough physical examination carried out on all patients who gave informed written consent and satisfied the inclusion criteria for the study. Diagnosis of congestive heart failure was based on the Framingham’s criteria for diagnosis of heart failure⁷⁸. Major criteria were paroxysmal

Cardiomegaly was deemed present when cardiothoracic ratio on postero-anterior chest radiograph was greater than 0.5. Minor criteria were ankle edema, night cough, dyspnea on exertion, hepatomegaly, pleural effusion and tachycardia.

The subjects were weighed without shoes and in light clothing on a standard beam balance. Height was measured to the nearest centimeter using anthropometrical plane with subjects not putting on shoes or headgear. Body mass index (BMI) was calculated using the formula:

BMI= Weight (kg)/ [Height (m)] ²

.

The BP was recorded in the right arm of subjects who have rested for at least 5 minutes, with a mercury sphygmomanometer (Accosson) and an appropriately sized cuff, according to the standard guidelines⁸¹. Systolic and diastolic blood pressure were measured at Korotkoff sounds phases I and V respectively. Hypertension was deemed present if systolic blood pressure (BP) ≥ 140 mmHg and/or diastolic BP ≥ 90mmHg on at least 2 occasions, or if patient was receiving anti-hypertensive medications.

Ischemic heart disease was considered present and a cause of heart failure when there was evidence of definite myocardial ischemia from history, ECG or Echocardiogram, or if there is a long history of stable angina pectoris. Valvular heart disease was considered the primary cause of heart failure when there was echocardiographic evidence of valvular heart disease of more than moderate severity. Dilated cardiomyopathy was diagnosed based on dilated left ventricle without any discernible cause⁸².

ELECTROCARDIOGRAPHIC MONITORING

24-hour Holter electrocardiographic monitoring was conducted on all subjects using Vision 5L^® Digital Holter Recorder (Spacelabs Burdick Medical Instruments, Deerfield, WI, USA). Patients with CHF in NYHA Grade II-IV underwent continuous 24-hour Holter monitoring while resting in bed during the procedure, whereas normal controls were permitted limited ambulation around the hospital premises, the latter measure being largely a safeguard against possible loss of the monitoring device. Three channel (V1, V5, and aVF) recordings were obtained and analyzed according to published guidelines⁹. Computer-assisted rate and arrhythmia analyses were performed using Vision Premier^® software (Spacelabs Burdick Medical Instruments, Deerfield, WI, USA). HRV measures were obtained from computerized analysis. The time domain HRV measures obtained were SDNN, SDANN, SDNNDX, rMMSD and PNN>50. The frequency domain HRV measures obtained were Total Power, VLF Power, LF Power, LFNu, HF Power, HFNu, and LF/HF.

ECHOCARDIOGRAPHY

Trans-thoracic echocardiographic assessment was carried out with ALOKA-SSD-1700 (Aloka Co. Ltd., Tokyo, Japan) echocardiographic machine which was equipped with a 3.5 MHz linear array transducer. Two-dimensional and M-mode echocardiography was performed on each subject (heart failure cases as well as controls) in the left lateral decubitus position. All measurements were made according to the American Society of Echocardiography guidelines⁸³ using leading edge to leading edge criteria.

Measurements were taken with simultaneous ECG recording in three cardiac cycles,

The left atrium was measured at end-ventricular systole. Left ventricular internal diameters were measured at end diastole and end systole. Left ventricular (LV) systolic performance was assessed using the fractional shortening and the ejection fraction of the left ventricle.

LV ejection fraction was calculated using the Teichholz calculation formula⁸⁴: [(EDV-ESV)/ EDV] x 100.

Where EDV = End diastolic volume ESV = End systolic volume

Fractional shortening was calculated from the LV internal dimensions in diastole and systole using the formula:

[( LVIDd – LVIDs)/ LVIDd] x 100

Where LVIDd = Left ventricular internal diameter in diastole.

LVIDs = Left ventricular internal diameter in systole.

DATA MANAGEMENT AND ANALYSIS

All data generated were collected in a standard proforma. Statistical analysis was performed using SPSS software version 12.0 for Windows (SPSS, Inc. Chicago Illinois). Means were expressed as means (standard deviation) while proportions were expressed as count (percentages). Continuous variables were compared with the student’s t test for independent groups. A stepwise regression analysis was carried out, with the HRV time domain values as dependent variables and Age, Gender, Ejection Fraction, Body Mass Index and NYHA status as independent variables, with a probability of entry of 0.05 and a probability of removal of 0.1. Data was log transformed when necessary to fulfill the criteria of normality. Non-parametric analyses were carried out when assumptions for normality were not met. A two tailed p value < 0.05 was considered to be significant.