Left Ventricular Mass as a Risk Factor in the Oldest Old

Abstract

In middle‐aged and “young elderly” cohorts, higher left ventricular mass (LVM) is associated with worse outcomes. The authors examined LVM and 5‐year mortality among community‐dwelling 85‐year‐old patients. A representative sample (n=526, born 1920–1921) from the Jerusalem Longitudinal Cohort Study underwent echocardiography at age 85. LVM was indexed by body surface area (LVM‐BSA) or height (LVM‐Ht). Patients with higher LVM were less educated and sedentary and had poorer self‐rated health, functional limitations, and increased comorbidity. Five‐year mortality was 21.7% (n=114). Adjusted 5‐year mortality rates were increased for the two upper quintiles of LVM‐BSA (hazard ratio [HR], 1.8; 95% confidence interval [CI], 1.05–3.06) and LVM‐Ht (HR, 2.2; 95% CI, 1.2–3.5). A step up in mortality occurred around the third quintile corresponding with LVM‐BSA 110 g/m² or LVM‐Ht 51 g/m^2.7. Among the oldest old, elevated LVM is significantly associated with mortality.

Cardiovascular care among the oldest old presents an increasing challenge for clinicians and health care providers, especially given the relatively high frequency of morbidity and death from cardiac causes in this population. Despite the increasing burden of long‐term cardiovascular care among the growing number of oldest old, prospective data examining the incidence of abnormal cardiac structure and its association with mortality in this population are limited.1, 2 Previous studies that have used echocardiography in elderly patients to examine prognosis have included a broad range of ages, with relatively few patients aged older than 80 years.3, 4, 5 In addition, existing studies in older populations have generally been performed in the hospital or in an outpatient clinic setting, possibly contributing to a biased study population in this particular age group, as patients may have difficulty in leaving their homes.6 The recent introduction of portable echocardiographic instruments makes it possible to study patients in their homes and assess a more representative population of the oldest old.

As was shown many years ago by the late Robert Tarazi, hypertrophy of the left ventricle (LVH) is a potentially reversible risk factor.3 It has also been long known to be associated with adverse cardiovascular outcomes in both the general and elderly populations.4, 5, 6, 7, 8, 9 However, LVH is strongly affected by body size, and there is no consensus on how it should be indexed. The two most commonly used forms of adjustment of left ventricular (LV) mass (LVM) for body size are either by body surface area (BSA)10 or by height alone. Because height is unidimensional but cardiac structure is three‐dimensional, the more commonly accepted method of indexation is by height^2.7 (LVM‐Ht).11 This method has the advantage of being relatively insensitive to sex. It is important to note that the commonly used indexed definitions of normal LVM are derived from middle‐aged cohorts, with limited data from elderly cohorts. Therefore, data regarding normative values of LVH and their relationship to mortality in the “oldest old” are lacking.

The objective of the present study is to identify normal upper limits for LVM as the cutoff point beyond which mortality is found to significantly increase in this population measured with home echocardiography among a representative sample of community‐dwelling 85‐year‐old patients (“oldest old”).

Methods

Participants were recruited from the Jerusalem Longitudinal Cohort Study, which was initiated in 1990 and has followed an age‐homogenous representative cohort of West Jerusalem residents born from June 1920 to May 1921. The methods have been described elsewhere in detail.12, 13, 14 In the present study, we examined data from the third phase of data collection, which took place during 2005 and 2006. Patients were interviewed and examined in their homes on two separate occasions. The first session was composed of a structured interview that lasted about 1.5 hours. Information was collected in socioeconomic, demographic, medical, functional, cultural, and cognitive domains. The second session consisted of a medical examination. The institutional ethics committee of the Hadassah–Hebrew University Medical Center approved the study design, and written informed consent was obtained from all participants.

Patients identified from the electoral register were randomly chosen from the total sample of individuals born in 1920 and 1921 and living in Jerusalem in 2005. Random numbers were used to choose patients from the electoral register according to their ID numbers.

Patients in the study group who were not enrolled in the echocardiographic substudy had near‐identical mortality and morbidity rates as those in the echocardiographic substudy. Furthermore, the representative nature of the total study population in comparison to the overall same age stratum of the Jerusalem population has previously been described.14 Mortality‐ and hospital‐specific morbidity rates were found to be similar among patients enrolled in the study, compared with mortality and hospital morbidity data of patients who were invited to enroll in the study but declined, as well as compared with patients from the birth cohort who were not approached to enroll in the study.

Sitting blood pressure with the arm supported at the heart level was measured three times with the use of a validated electronic sphygmomanometer (Omron 705IT; Omron Healthcare, Inc, Lake Forest, IL) according to the recommendations of the European Society of Hypertension at both occasions. Hypertension was defined as treatment with antihypertensive medications or average blood pressure >140 mm Hg systolic or 90 mm Hg diastolic (average of six measurements).

Diagnosis of ischemic heart disease (IHD) was based on a history of hospitalization for myocardial infarction or acute coronary syndromes, coronary catheterization with evidence of significant coronary artery disease, positive stress test results, and a history typical for angina pectoris on exertion or previous coronary artery revascularization. Hyperlipidemia was defined as use of cholesterol‐lowering medications. Diabetes mellitus was a composite of hypoglycemic medications, personal history, or a medical record diagnosis. Congestive heart failure (CHF) was based on hospital discharge diagnosis and according to examining research physician diagnosis at the time of examination at home. Self‐rated health was assessed according to the question: “How do you rate your general health?” Possible responses were “good” and “poor,” using a Mini‐Mental State Examination (MMSE) score <24 as a cutoff point.15 Functional status was defined as dependence on another person in one or more of six activities of daily living [ADL], eg, transferring, dressing, bathing, using the toilet, eating, and remaining continent.16 Participants were questioned: “How often are you physically active?” Possible answers were: (1) <4 h/wk, (2) about 4 h/wk, (3) at least an hour per day (eg, regular physical activity such as walking), and (4) vigorous sports at least 2 times per week (eg, jogging, swimming). Physical activity was dichotomized to inactive (answer 1) vs active (answers 2, 3, or 4). This cutoff was justified statistically, accounting for distribution and frequency of responses.17 The questionnaire has been found to be predictive of both functional status as well as morbidity throughout follow‐up from age 70 to 90 years among our cohort.18

Echocardiography was performed in 526 patients, evenly distributed between new recruits and patients participating from previous phases, as previously described.12, 19 They underwent standard two‐dimensional and Doppler echocardiography at their places of residence using a portable echocardiography device (Vivid I; GE Healthcare, Haifa, Israel). All patients underwent two‐dimensional and Doppler echocardiography with M‐mode measurements of the interventricular septum, posterior wall, and LV end‐systolic and end‐diastolic diameters according to the recommendations of the European Association of Echocardiography and the American Society of Echocardiography.9 End‐diastolic measurements were performed for three consecutive cardiac cycles and averaged. Patient height and weight at the time of the study were recorded and body surface area was calculated. LVM was calculated according to a necropsy‐validated formula as LVM (g)=0.8×{1.04 [(septal thickness+LV internal diameter+posterior wall thickness)³−(LV internal diameter)³]}×0.6 and indexed to body surface area.12, 19 Given the high prevalence of basal septal hypertrophy in this population, septal thickness measurements were taken below the level of the basal septum.

Throughout the study we used two measurements for LVH. The first was defined based on LVM‐BSA >125 g/m² in men and >110 g/m² in women.20 The second definition was performed by indexing LVM to height^2.7 with a sex‐independent cutoff >51 g/m^2.7 (LVM‐Ht).20 We determined quintiles for LVM indexed by either body surface area (LVM‐BSA g/m²) or by height^2.7 (LVM‐Ht m^2.7).

Descriptive statistics were calculated, and because the cardiac parameter data were normally distributed, results are described as mean±standard deviation. Percentages were calculated as appropriate. We divided the population by quintiles of LVM and examined all‐cause mortality. For continuous variables, differences between means were calculated using t tests, and cumulative survival was assessed using Kaplan‐Meier analysis and log‐rank tests for statistical significance. Adjusted and unadjusted Cox proportional hazards models were performed. All models were adjusted for sex, education, self‐rated health, physical activity, diabetes, IHD, hypertension, and either LVM‐BSA or LVM‐Ht. When analyzing LVM‐BSA, we did not adjust for weight and height (included in the formula), and when it was indexed by height^2.7 (LVM‐Ht), only weight adjustments were made to refrain from overadjustment.

Survival status at 5‐year follow‐up was assessed through the centralized Ministry of Interior database. Follow‐up was available for all study patients.

All P values were two‐tailed and P=.05 was considered significant. Data storage and analysis were performed using SAS version 9.1e (SAS Institute Inc, Cary, NC).

Results

Of the 526 participants with available echocardiographic data, there were 248 men and 278 women. While no significant medical differences existed between patients with echocardiography and the rest of the cohort, the patients undergoing echocardiography appeared to have been in better functional condition (data not shown).

Rates of LVH were high in general and significantly higher for women than men (59.7% vs 45.2%, P=.0008) by LVM‐BSA but not for LVM‐Ht (69.8% vs 62.5%, P=.08) according to traditional cutoffs.20, 21 The same trend was seen when LVM was analyzed continuously for both methods (Table II). During 5 years of follow‐up, 114 participants (21.7%) died. Clinical characteristics of participants are shown in Table 1.

Table 1.

Baseline Characteristics of 526 Participants Aged 85 Years by Survival

	Survived (n=412)	Not Survived (n=114)	P Value
Men, %	44.1	57.9	.0094
Education ≥12 y, %	53.1	40.4	.014
Married, %	44.6	50.0	.34
Physically active, %	76.4	56.6	<.0001
Feels healthy, %	73.6	50.5	<.0001
Difficulty in ADL, %	20.6	55.0	<.0001
Hypertension, %	90.9	91.2	.87
Diabetes mellitus, %	16.5	32.5	<.0001
IHD, %	33.0	53.5	<.0001
CHF, %	7.6	24.8	<.0001
MMSE score ≥24, %	86.2	72.7	.0004

Abbreviations: ADL, activity of daily living; CHF, congestive heart failure; IHD, ischemic heart disease; MMSE, Mini‐Mental State Examination.

As expected, the survivors were more likely to be women and better educated, to regard themselves as healthy, and to be more physically active. They were less likely to be dependent with ADLs, to be cognitively impaired, and to have diabetes, IHD, or CHF. Interestingly, LVM was not related to measured systolic blood pressure (LVM‐BSA relationship to blood pressure: r=0.05, P=.2; LVM‐Ht r=0.06, P=.16).

In Table 2, LVM is presented for survivors and those who died during the 5 years of follow‐up. Survivors had significantly lower LVM according to LVMI‐BSA and LVH‐Ht for both men and women. There was good agreement in LVM‐BSA and LVH‐Ht in the quintiles (κ=0.8, P<.0001).

Table 2.

Descriptive Statistics of LVM for Different Equations by Sex and Survival

	Total	Survived	Not Survived	P Value
Women (n=278)
LVM‐BSA, g/m2	122.0±38.2	120.2±32.7	136.0±56.5	.06
Median (Q1–Q3)	118 (96–143)	116 (96–139)	129 (97–149)
LVM‐Ht, g/m2.7	63.8±20.4a	62.1±18.1	71.9±27.5	.02
Median (Q1–Q3)	61 (48–76)	60 (48–74)	68 (54–82)
Men (n=248)
LVM‐BSA, g/m2	123.1±31.7	119.5±30.4	133.0±33.2	.003
Median (Q1–Q3)	122 (99–143)	116 (98–138)	132 (109–153)
LVM‐Ht, g/m2.7	57.7±16.6	55.9±16.1	62.8±16.8	.004
Median (Q1–Q3)	57 (65–68)	54 (44–65)	62 (48–73)

Abbreviations: LVM‐BSA, left ventricular mass indexed for body surface area; LVM‐Ht, left ventricular mass indexed for height^2.7; Q, quintile. Values are expressed as mean±standard deviation. ^a P<.0002 in comparison to men.

We calculated hazards ratios (HRs) for mortality associated with LVM‐BSA and LVH‐Ht, both treated as continuous variables, using Cox regression analysis, which adjusted for covariates of sex, education higher than 12 years, diabetes mellitus, IHD, CHF, feeling healthy, physical activity, and LVM index (with LVM‐Ht, weight was also entered as a covariate). The results showed that for every unit increase in LVM index there were significantly higher mortality HRs: for LVM‐BSA the HR was 1.007 (95% confidence interval [CI], 1.002–1.013) and for LVH‐Ht the HR was 1.012 (95% CI, 1.002–1.023).

As expected, sex, diabetes mellitus, IHD, and CHF were associated with higher HRs as well, whereas perceived good health, physical activity, and weight were protective. When using the traditional cutoffs for LVH by both methods, Cox proportional hazard model yielded an HR of 1.62 (95% CI, 0.98–2.67) for LVM‐BSA and 1.73 (95% CI, 1.04–2.89) for LVM‐Ht.

For quintile analysis we compared patients with higher LVM (the three upper quintiles vs the two lower) (Table 3) and found it to be associated with hypertension, IHD, CHF, and dependence in ADL; lower subjective health perception for both indexing methods; and diabetes mellitus, higher prevalence of cognitive impairment, and higher BMI for LVM‐Ht.

Table 3.

Baseline Characteristics by LVM Level

	LVM Index‐BSA		LVM Index‐Height
	Quartile 1–2 (n=210)	Quartile 3–5 (n=316)	Quartile 1–2 (n=208)	Quartile 3–5 (n=318)
Men	99 (47.1)	149 (47.1)	98 (47.1)	150 (47.2)
Education ≥12 y	110 (52.4)	155 (49.0)	116 (55.8)	149 (46.9)
Married	96 (46.2)	144 (45.6)	96 (46.6)	144 (45.3)
Physically active	158 (76.0)	220 (69.6)	162 (78.6)	216 (67.9)a
Feels healthy	160 (77.3)	194 (62.8)b	164 (79.6)	190 (61.3)c
Difficulty in ADL	48 (23.1)	97 (31.3)b	46 (22.2)	99 (31.8)b
Hypertension	184 (88.9)	300 (95.5)a	181 (88.7)	303 (95.6)a
Diabetes	38 (18.1)	67 (21.2)	32 (15.4)	73 (23.0)b
IHD	64 (30.5)	133 (42.1)a	63 (30.3)	134 (42.1)a
CHF	16 (7.6)	52 (16.5)a	15 (7.2)	53 (16.7)a
MMSE ≤24	27 (13.2)	59 (19.0)	25 (12.2)	61 (19.6)b
BMI, mean±SD	26.8±4.1	27.2±4.3	25.8±4.0	27.9±4.2c

Abbreviations: ADL, activity of daily living; BSA, body surface area; BMI, body mass index; CHF, congestive heart failure; IHD, ischemic heart disease; LVM, left ventricular mass; MMSE, Mini‐Mental State Examination; SD, standard deviation. Values are expressed as number (percentage) unless otherwise indicated. ^a P<.01. ^b P<.05. ^c P<.0001.

When quintiles of LVM index (both methods) were plotted against rate of 5‐year mortality, a significant step up in mortality between the second and third quintiles was observed (Figure).

Figure 1.

Bar graph plotting quintiles of left ventricular mass indexed by body surface are (LVM‐BSA; open bars) or by height^2.7 (LCM‐Ht; filled bars) against 5‐year mortality. For women: first quintile LVM‐BSA/LVM‐Ht ≤89 g/m²/45 m^2.7, second quintile: 90 g/m² to 109 g/m²/46 m^2.7 to 55 m^2.7, third quintile: 110 g/m² to 128 g/m²/56 m^2.7 to 67 m^2.7, fourth quintile: 129 g/m² to 150 g/m²/68 m^2.7 to 79 m^2.7, and fifth quintile, ≥151 g/m²/≥80 m^2.7. For men: first quintile LVM‐BSA/LVM‐Ht: ≤94 g/m²/43 m^2.7, second quintile: 94 g/m² to 113 g/m²/44 m^2.7 to 52 m^2.7, third quintile: 114 g/m² to 128 g/m²/53 g/m^2.7 to 60 g/m^2.7, fourth quintile: 129 g/m² to 148 g/m²/61 m^2.7 to 70 m^2.7, and fifth quintile, ≥149 g/m²/≥71 m^2.7.

As seen in Table 3 these quintiles (three through five) had a profile of higher risk. Indeed, as shown in Table 4, they were associated with significantly higher mortality. After adjustment for confounding covariates, HRs for mortality and 95% CIs were higher for the two higher quintiles of LVM‐BSA (1.8; 95% CI, 1.05–3.06) and LVM‐Ht (2.2; 95% CI, 1.2–3.5).

Table 4.

Cox Proportional Hazard Ratios and Confidence Intervals of Mortality by LVM, Indexed by BSA and Height^2.7

	Sex‐Adjusted	Fully Adjusteda
LVM‐BSA Q4, Q5 vs Q1, Q2	2.09 (1.27–3.43)	1.8 (1.05–3.06)
LVM‐BSA Q3 vs Q1, Q2	1.26 (0.7–2.4)	1.3 (0.7–2.55)
LVM‐Ht Q4, Q5 vs Q1, Q2	2.3 (1.4–3.9)	2.02 (1.2–3.5)
LVM‐Ht Q3 vs Q1, Q2	1.6 (0.84–2.92)	1.3 (0.7–2.5)

Abbreviations: LVM‐BSA, left ventricular mass indexed for body surface area; LVM‐Ht, left ventricular mass indexed for height^2.7; Q, quintile. ^aAdjusted for sex, education, diabetes, ischemic heart disease, hypertension, subjective health perception, physical activity, congestive heart failure, and cognitive impairment.

Therefore, it seems reasonable to suggest upper limit of normal LV mass in both men and women 85 years of age, as 110 g/m² or 55 g/m^2.7 corresponding roughly to the lower end of the third quintile.

Discussion

In this study of a representative sample of community‐dwelling oldest old followed from age 85 to 90 years, we describe a high incidence of LVH and show that increased LVM is associated with mortality even after controlling for confounders that are frequently associated with both mortality and elevated LVM.

Our findings are consistent with previous epidemiologic studies showing that the prevalence of LVH increases with age.22, 23, 24 These studies have generally examined patients older than 65 years, and our findings extend to include a population followed up from age 85 to 90 years. We also confirm previous findings in younger populations of increased mortality related to LVH.8 The finding of elevated mortality related to LVH in this age group is of importance as other studies of the very elderly have shown that many traditional risk factors do not predict mortality as they do in younger elderly and middle‐aged patients.24, 25, 26, 27 Functional variables such as physical activity and reduction in ADLs may be more predictive in this population.14, 18, 28 We have previously shown that risk factors associated with the development of LVH such as obesity and hypertension were not associated with mortality in the oldest old.28, 29 The fact that high LVM predicts mortality, whereas some of its own predictors do not, underlines the importance of the association, which was significant when LVM was treated as either a continuous or discrete variable. LVM is the integrated result of hypertension, obesity, salt intake, and many other factors that have driven the myocardial changes years and decades prior to the current measurements. In addition, variables related to less access to or less‐effective utilization of health care systems appear to influence the development of LVH, as demonstrated by the finding that patients with higher LVM were less educated and more likely sedentary. The impact of lifelong accrued medical problems is evident in their poorer self‐rated health, present functional limitations, and increased comorbidities. Indeed, a recent study in a similar age group also found that what is predictive of mortality is not current hypertension but rather the changes of BP over time that have accumulated throughout adult life.30 Similarly, LVM is a marker of a variety of factors prevailing long before the current measurements.

As controversy exists regarding the most appropriate indexing method, we examined LVM according to both BSA and height. In our study, both indexing methods were in good agreement both in distribution of LVM quintiles and in their relationship to mortality. While higher indexed LVM was similarly associated with hypertension, IHD, CHF, and ADL using either method, higher LVM‐HT was associated less with physical activity and cognitive function and with higher BMI and diabetes mellitus. Otherwise, performance was similar with the exception of significantly higher LVM with the LVM‐Ht method in women. Perhaps the increased loss of height among older women generally makes it estimate higher LVM‐BSA.

Previous studies have demonstrated more pronounced LVH in women in the setting of pressure overload.31, 32 An age‐associated LVM increase could have been related to estrogen receptor polymorphism.33 Indeed, over time, middle‐aged and older patients (up to 75 years in the Framingham study) were more likely to have an increase in LVM, especially if hypertensive, obese, and diabetic,34 with a steeper increase observed in women. Men with diabetes also had accentuated LVM changes with age. Because mortality is higher in men and those with diabetes, a survival bias may be the cause (including genetic factors among others) for the frequency of higher LVM among the older women in our cohort. We did not find a sex difference in LVM‐BSA but did note higher LVM in women when indexed by height. However, significance of this finding is not clear. Nevertheless, when looked at from a prognostic point of view, the same quintiles in both sexes are similarly associated with mortality, which seems to increase around 110 g/m² and 51 g/m^2.7. Therefore, our results suggest that these cutoffs should serve as the upper limit of normal for both sexes in the very elderly. Indeed, these are very similar to those of LVM‐Ht in younger people (21 years) and to LVM‐BSA of younger women20 and slightly lower than the 125 g/m² noted in younger men.20 The discrepancy in the later finding could be related to the earlier death of younger men with highest risk, thus reducing prevailing LVM among older survivors.

Study Strengths and Limitations

The strengths of our study are that it is a representative cohort of the oldest old living in the community rather clinic‐based or hospitalized patients. Moreover, echocardiography was performed at their homes, allowing people with limitations, who might have never participated in a study that requires clinic attendance, to be included. Patients spanned a range of functional status, from completely independent to those receiving assistance in basic and instrumental ADLs and being cared for at home. Our study also allows for a relatively large age‐homogenous group to be examined with detailed geriatric‐centered data collection, including self‐perceived health, physical activity, and functional as well as cognitive assessment in addition to evaluation of comorbidities. Comparisons with similar datasets show our cohort to share remarkably similar characteristics with other cohorts from different countries.30, 35 Limitations include the retrospective data analysis for outcomes, as well as the availability of high‐quality echocardiography in only a subset of the study cohort. This was a subgroup with no significant differences in demographics and medical diagnoses among patients who underwent echocardiography and those who did not; however, the former group had significantly higher education, better functional capacity and self‐rated health, and were more likely to be married. Nevertheless, these confounders were considered in the analyses. Another issue could be the normal decline in height with age, particularly in women, caused by osteoporosis, disc problems, and spinal curvature. Such a decline may cause overestimation of indexed LVM because it is involved in both indexation formulas in the denominator. Such misclassification would only serve to underestimate the findings, and height did not change the effect of LVM on mortality in these models. If anything, these age‐related changes in height would have diluted the effect of indexed LVM on mortality, underestimating the association.

How might our findings affect clinical practice among very old people? Clearly, caution is required in extrapolating epidemiological descriptive data, such as ours, to real‐life clinical practice. Conclusions cannot be drawn from this present study as to whether LVM among the oldest old is reversible or amenable to change and whether intervention at this age aimed at reducing LVH would in any way affect clinical outcomes or mortality. While antihypertensive treatment among our cohort at age 85 has previously been shown to be unrelated to overall mortality,28 a true prospective intervention among a similar cohort is needed to shed more light on the pertinent clinical dilemma concerning the indications for treatment among the oldest old. The potential usefulness of our findings might be suggested in helping clinicians assess risk among their oldest patients and aid them in the difficult task of prognostics.

Conclusions

Elevated LVM is common and significantly associated with mortality in the oldest old even after adjustment for potential confounders. Our findings suggest cutoffs that represent the upper limit of normal LVM among 85‐year‐old patients.

J Clin Hypertens (Greenwich). 2015;17:874–879. DOI: 10.1111/jch.12594. © 2015 Wiley Periodicals, Inc.