Both M-mode and two-dimensional imaging can be employed to calculate left ventricular mass. M-mode imaging allows better endocardial border definition as it has greater resolution due to higher frame-rate, as long as adequate ultrasound beam positioning is ensured and ventricle shape approaches normality. Two-dimensional imaging, on the other hand, depicts the "real" ventricular shape and identifies regional motion abnormalities. However, the quality of two-dimensional imaging may be limited due to both lower lateral resolution and frame-rate. Additionally this option is more time consuming, limiting its use in epidemiological studies. Two-dimensional images are usually acquired both in paraesternal and apical views, depending on the geometrical formulas that are used.

Technological advances have joined both methods and partially minimized their limitations There are standardized and validated recommendations for the clinical use of two-dimensional determination of LV mass 6, However, two-dimensionally oriented M-mode, obtaining images perpendicularly from the longitudinal axis slightly above the papillary muscle level is widely employed in clinical practice and is accepted as an adequate alternative in epidemiological studies. Digital imaging has also made it possible to reconstruct diverse M-mode planes from two-dimensional images (so called anatomical M-mode), allowing better positioning, although the final image resolution is that of two-dimensional images. Although accurate 7, LV mass estimation using anatomical M-mode has not been adequately validated in clinical studies. Two-dimensional mode has also improved due to refined imaging processing technology, particularly second harmonic imaging. Also, built-in software for automatic border detection has been developed allowing calculation of real-time volumes. Although most of these newer technologies are widely available in commercially available ultrasound equipment, their potential for providing additional accuracy in the evaluation of LVH remains poorly characterized.

A relevant degree of variability in LV mass determination could be attributed to online measurement inaccuracies due to lower imaging resolution of older equipment, which could reach 10% of parietal thickness. Nowadays, variability due to on-line or off-line analysis of digitalized images calculations is of considerably smaller magnitude 8.

We will focus our discussion on M-mode estimation of LV mass, since most epidemiological reports use this imaging modality. Preference for M-mode is based on its technical feasibility and availability at the time when most studies were performed. However, despite adequate correlation with two-dimensional measurements 9, M-mode has been suggested to averagely underestimate LV mass in about 20 g 10. Theoretically, two-dimensional imaging would be more adequate in samples of patients with cardiovascular disease, where LV shape assumptions play a critical role in LV mass estimation.

Determination of left ventricular volumes is accomplished using formulas that fit ventricular shape to primary geometric figures. Ellipse, cylinder, cone, and truncated polyhedrons have been employed and validated in normally shaped ventricles, although most studies have limited sample sizes 11. Some geometrical assumptions are best fit using two-dimensional images while others can be performed assuming geometric forms from M-mode imaging.

If a two-dimensional approach is used, both area-length and truncated-ellipsoid models are feasible and reasonably accurate, with validated formulas 6. Among other formulas for volume calculation, the modified Simpson's Discs Rule, which is greatly facilitated nowadays by built-in software, offers flexibility that allows accurate estimation of left ventricular volumes even in greatly distorted ventricles. Even though volumes determined by Simpson's Rule are frequently used for ventricular function determination, they are not routinely used to calculate LV mass, probably due to limitations resulting from poor epicardial delimitation in some patients. Other geometrical models have been proposed in the past 1213, with variable use in clinical practice.

The method of cubed formulas, which incorporate only one dimension of the left ventricular cavity and assume an ellipsoid geometry, is the most widely used to calculate left ventricular volumes and mass in noninvasive laboratories worldwide. The broad acceptance of M-mode derived estimation relies on the fact that it was the first method that was validated and also on its technical simplicity. The one-dimensional approach, however, imposes more strict geometrical assumptions and amplifies the risk of inaccuracy, as measurement errors are cubed.

Ultrasound signals are reinforced where surfaces change density, allowing definition of limits between surface layers. The inclusion or exclusion of these echoes from interfaces of the left ventricular cavity or myocardial wall can cause significant discrepancies in the overall measurements 14. Initial M-mode standard recommended inclusion of the edges as part of interventricular septum thickness, but exclusion of the posterior wall epicardial edge 15. Investigators of the University of Pennsylvania developed a criteria (The Penn Convention) in which all edges are not included in parietal thickness measurements, but are considered as part of the ventricle cavity 16. This approach underestimate LV mass when compared to the M-mode convention, proposed by the American Society of Echocardiography (ASE). This latter convention (ASE) is the most accepted border definition criteria, becoming the standard recommendation for M-mode estimations, and uses the leading edge of each layer 17(Figure 1). Employing Penn and ASE convention with the same volume formulas may originate LV mass discrepancies in the range of 15% in men and 18% in women 18. Measurement convention must be acknowledged and adequately corrected for in comparisons of clinical studies of LVH.

In the past few years, spatial resolution of transthoracic echocardiography has greatly enhanced, leading to major improvements in image quality. Most of this progress can be attributed to second harmonic imaging, that significantly increases signal-to-noise ratio by receiving only harmonic frequencies. Although this technology is now widely available in most ultrasound equipment, its potential accuracy to evaluate LVH remains poorly characterized. A recent study suggests that LV mass estimations using second harmonic imaging can cause as much as a 26% increment in mean LV mass index corrected for body surface area, when compared to standard fundamental imaging using similar formulas and conventions 19. This is probably due to an increase in border refringency.

The most commonly used formulas to estimate LV mass are all variations of the same mathematical principle, based in the volume formulas stated above. Original calculations from Troy and coworkers were the first to be recommended as standard to estimate LV mass from M-mode measurements (Formula 1) 15.

Formula 1: LV mass(Troy) = 1.05 ([LVIDD + PWTD + IVSTD]³- [LVIDD]³) g.

Where: LVIDD = Left Ventricular Internal Diameter in Diastole

       PWTD = Posterior Wall Thickness in Diastole

       IVSTD = Interventricular Septum Thickness in Diastole

Subsequently, Devereux and colleagues suggested a slightly modified regression equation, using the Penn convention as the border definition criteria (Formula 2). Their prediction equation in this pivotal study was derived from necropsy findings of 34 patients 16.

Formula 2: LV mass(Penn) = 1.04 ([LVIDD + PWTD + IVSTD]³- [LVIDD]³) -13,6 g.

As depicted above, each regression equation was derived based on a specific border limits convention, an issue that is source of great confusion when interpreting different studies 20. As expected, LV mass calculations derived from both formulas are linearly correlated, but final crude estimations may differ by more than 20%. Devereux and colleagues proposed a new adjusted equation, validated on necropsy findings of 52 individuals 21, using the ASE convention and accounting for this discrepancy (Formula 3).

Formula3: LVmass(ASE): 0.8 (1.04 ([LVIDD + PWTD + IVSTD]³- [LVIDD]³))+ 0,6 g.

Some critical aspects must be acknowledged regarding LV mass formulas. First, all necropsy validation studies have limited sample sizes and evaluate heterogeneous ventricular configurations. Second, these formulas may not perform adequately in distorted ventricles, where a two-dimensional approach is preferred. Different formulas may yield distinct cut point values, as demonstrated by Levy and coworkers in the Framingham cohort 18. Finally, other post-mortem study showed only moderate correlation between echocardiographic and autopsy LV mass estimations (correlation coefficients ranging from 0.58 to 0.67) 22.

Both body size and body habitus are clearly associated with LV dimensions and mass. Diverse normalization and indexes were created and tested to adjust for three different sources of physiologic variation in LV mass: lean body mass, obesity, and gender. However, the interdependence of such associations should be carefully understood to allow an adequate correction of LV mass without distorting its association with cardiovascular disease.

Several indexes for body size correction have been proposed, such as height, diverse allometric height adjustments, weight, body surface area, body mass index, and free-fat mass. The best way for normalization of LV mass is still controversial and another source of confusion. Different body-size adjustment criteria and their standard cut points result in different prevalence of patients with LVH 18. Not surprisingly, those with higher LV mass are more frequently classified as hypertrophic by different classifications simultaneously 23.

The body surface area correction, using the Dubois formula 24, reduces variability due to body size and gender 25, but this index underestimates LV mass in the upper range of the body surface area distribution 26. A correction based on height alone would allow evaluation of the separate role of obesity in LVH as proposed by Levy and coworkers 18.

Adjustment of LV mass with body surface area would imply that obese patients are expected to have higher LV mass estimations per se. In this scenario, height-based adjustments can more accurately estimate LV mass and the resulting cardiovascular risk associated with LVH in the obese. This turns to be particularly relevant in risk stratification due to the frequent clustering of hypertension, obesity and LHV. Different allometric height-based adjustments have been used. Height^2.7, derived from regression models in normal samples from De Simone and coworkers 26, appears to offer the most accurate estimation of LV hypertrophy and risk factors for pathologic changes in the heart structure, particularly in obese subjects. Zoccali and colleagues found LVH indexed by height^2.7 to be a better predictor of cardiovascular events than LVH indexed using body surface area in a group of patients under dialysis 27. Liao and colleagues 28 studied 988 patients and identified progressive increments in death rates with both body surface area and with height^2.7 indexing criteria. Subjects simultaneously classified as LVH with body surface area and height^2.7 criteria had increased average LV mass and a 3-fold increase in death rates, while those classified as LVH only when indexed by height indexes had no increase in future cardiovascular events. In summary, it appears prudent to favor indexes that do not adjust for obesity, such as height, and height^2.7, particularly in studies in which the independent impact of obesity is in question. Body surface area indexing permits adequate classification of most of patients in clinical practice, incorporating in LVH determination some of the risk associated with obesity.

Finally, men have increased LV mass and at least part of this effect can be attributed to body size differences. Gender differences in LV mass are first noticed around puberty and can be minimized although not eliminated by adequate indexing of body size 29. Due to this difference in LV mass, some criteria for body size adjustments use gender-specific cut points for normality as will be seen below.

The determination of cut points in biological variables to define abnormality is frequently a source of controversy, and can be driven by different strategies. The definition of what constitutes an abnormal LV mass is no exception to this rule.

Parietal thickness and its relation to LV chamber size have been recognized as measures of hypertrophy for more than 30 years 41. Relative wall thickness (RWT) is measured in clinical studies both as: 2 * posterior wall thickness divided by LV diastolic diameter or, septal wall thickness + posterior wall thickness divided by LV diastolic diameter. Even thought these measures have been used interchangeably by some investigators, septal asymmetry (IVSTD/PWTD > 1.3) was present in about 5% of Framingham subjects 42 and can lead to an underestimation of relative thickness when only posterior wall thickness is used. The reference cut point value for increased relative wall thickness derived from upper limits of normal samples is usually 0.44 43 or 0.45 42, irrespective of which formula is used. RWT provides information regarding LV geometry independent of other calculations 44, precluding the requirement of most corrections. Nevertheless, significant LVH can occur without major changes in RWT, particularly when simultaneous pressure and volume overload are present.

Attempts have been made to evaluate separately adaptive responses in parietal thickness increase and in dilation. Initially, Savage and coworkers 42 stratified Framingham patients with LVH in subgroups as: disproportionate septal LVH; concentric LVH; eccentric-dilated LVH, and eccentric non-dilated LVH. They identified in the 3 last categories increasing levels of systolic blood pressure, utilizing retrospective blood pressure data from 30-years of the cohort follow-up, suggesting a progressive character of adaptive mechanisms. A later approach defined 4 distinct geometric patterns: normal geometry, concentric remodeling, concentric hypertrophy and eccentric hypertrophy (Figure 2). Ganau and coworkers 43, using echocardiographic hemodynamic estimates, reinforced the impression that the geometric patterns parallels progressive hemodynamic changes. Hypertensive target organ disease measured by fundoscopic alterations are also more frequent in hypertrophic geometric patterns 30.

Koren and coworkers 45 used cut points of 125 g/m² for LVH and 0.45 for RWT in a sample of hypertensive patients and found a 10-year incidence of cardiovascular events of 31% in those with concentric hypertrophy compared to 11% in those with normal geometry. In 1995, two cohorts studies were simultaneously published evaluating geometric patterns impact in the incidence of cardiovascular events. Verdecchia and colleagues 46 studying 694 patients with body surface area indexed LV mass lower than 125 g/m², without additional adjustment for obesity and other metabolic risk factors, found a relative risk of 2.6 in the 272 patients with concentric remodeling compared to normal geometry patients. Krumholz and coworkers 47, studied 3209 from The Framingham study, indexed LVM by height using cut points of 143 g/m in men and 102 g/m in women and adjusted the models for obesity and other relevant covariates. Their analysis showed a relative risk of 2.1 for all cause mortality with concentric hypertrophy, but not additional risk in those classified as concentric remodeling. Relative risk became nonsignificant when a correction for LV mass was included in the models. Verdecchia and colleagues 48, afterwards could not demonstrate additional risk associated with increased relative wall thickness in those classified as hypertrophic. These data may suggest a smaller independent risk associated with increased wall thickness in hypertensive patients without LVH criteria.

Even though the additional prognostic role of geometric patterns over LVH may be lesser than initially supposed, this classification permits identification of determined adaptive processes. Concentric remodeling may be related to specific pathophysiological adaptations, particularly related to glucose and insulin metabolism 495051 and studies in contemporary cohorts have also shown an association of concentric forms with diabetes 5253.

We believe that geometric classification, with adequate body size indexing and clearly defined standardization, may be an alternative and informative strategy to evaluate adaptive responses, providing information beyond that provided by classification with respect to left ventricular hypertrophy.

Differences in LV mass due to gender, independent of questions related to body size, may have pathophysiological implications. Women have been shown to have an increased parietal hypertrophic responses to pressure overload 5455., even after body size correction. This adaptive pattern was demonstrated also in animal models 56. The unfavorable prognostic implications of this hypertrophic response are suggested by the findings of Liao and coworkers of a 5-fold greater risk of death associated with LV hypertrophy indexed by BSA in woman compared to the risk associated to LVH in men. However, despite using gender specific cut-offs for LVH, additional adjustment for obesity was not performed. Employment of height^2.7 indexing allowed to use a unique cut-point of 51 g/m^2.7 for both genders 26, reducing the impact of gender in LVH inference, at least in African-Americans 57.

Although the best strategy to adjust LVM for obesity is a matter of debate, obesity is increasingly recognized as an independent predictor of cardiovascular morbidity and mortality 5859. The increase in LV mass related to obesity is probably more than a mere physiologic adaptation.

Obesity has been shown to be independently associated to LVH 60, particularly in populations with a high prevalence of hypertension and other metabolic risk factors 6162. Despite this association, the impact of obesity on LVH may be less than expected 63, as Iacobellis and colleagues 64 have demonstrated that "uncomplicated obesity" was not a risk factor for LVH when indexed by either body surface area or height ^2.7. As obesity, however, causes complications, it is frequently accompanied by additional risk factors. Adjusting by height ^2.7 minimizes the interference of obesity in LV mass estimates (See above: Body-size Indexing).

LV mass progressively increases during aging 65, particularly parietal thickness 4, which was seen in both normotensive and hypertensive patients 66. Heart size increases during infancy and adolescence due to body size enlargement and, at this stage, the gender differences become prominent 67. The rate of LV mass increase due to age changes in magnitude 29, weakening its independent role at older individuals, when other risk factors play a greater role 63. Dannenbeg and coworkers 68 demonstrated that LV mass did not increase with age in a healthy sub-sample of The Framingham study, suggesting that most of the supposed physiological increase is caused by other determinants. These results are reinforced by studies in younger subjects where the age-associated increase in LV mass is partially explained by body size and blood pressure changes 67. Nevertheless, it appears prudent to adjust for age in epidemiological investigations related to LV mass and hypertrophy.

LVH is particularly prevalent in African-Americans 56269707172. In these analyses, two particular aspects deserve consideration. An increased crude prevalence of LVH in African-Americans and Hispanics is more evident using height-indexed LV mass than with body surface area-indexed LV mass 71, suggesting that obesity may partially explain the reported ethnic differences. Furthermore, adaptive response to hypertension may differ across ethnic groups. Hypertensive African-Americans, in comparison with hypertensive whites, have increased relative wall thickness, resulting in an increased frequency of concentric remodeling, given equivalent LV mass estimates 7273. However, Afro-American ancestry has been identified as an independent risk factor for LVH 74.

Numerous population based studies have unequivocally shown an association between hypertension and LVH 456575. Other reports usually stratify their analysis by or restrict to those with hypertension to allow better evaluation of additional risk factors 45737677. It is interesting that even within the normal range, increases in blood pressure is related to an increased LV mass 67. This increment may be attributed to the classical pathophysiological concept of hypertrophic response to increased overload, although neuro-humoral and genetic factors have been also implicated 78. LVH association with hypertension is so evident that it is recognized as target organ damage in hypertensive disease by several clinical practice guidelines, representing an intermediate unfavorable prognostic marker 7980.

Together with obesity and hypertension, diabetes has been implicated as an important determinant of left ventricular mass in most population-based studies 552628182. Myocardial and systemic mechanisms, as an increased extra-cellular matrix, vascular hypertrophy and vasoconstriction 83, have been attributed to this hypertrophic response.

An adaptive response has been shown to diverse degrees of altered carbohydrate metabolism, as in Cardiovascular Health Study 82 and in The Strong Heart Study cohort, where diabetes 52, impaired glucose tolerance 84 and insulin levels85 where associated with increased LV mass. Although associated with an increase in left ventricular mass, hyperinsulinemia 49 and insulin resistance 50 show a stronger association with concentric remodeling. Concentric hypertrophy is more pronounced in diabetes presenting with microalbuminuria 5186., which could imply a progressive adaptive process.

A gender difference in the left ventricular response to diabetes, with an increase in parietal thickening, rather than hypertrophy, being prominent in women has been suggested 818587.

LV mass increase is also seen in individuals with other known risk factors, as in those linked to the metabolic syndrome 6288., where pathophysiological aspects related to this syndrome may directly affect ventricular adaptive mechanisms.

A multitude of other factors have been shown to be independently related to LV mass. It should be emphasized that estimates of the relative magnitude of these factors varies according to the degree of adjustment for other known risk factors in statistical modeling. Primary valvular and myocardial disease are clearly related to LV mass increase but will not be subject of our review.

Environmental exposures such as alcohol consumption 89, salt intake 90, smoking 489. and increased leisure-time physical activity in men 91 have been associated to increased LV mass. Other factors such as blood lipids, pulmonary function, the heart rate and hematocrit have also been implicated but with some inconsistency among different studies 4759293. Also, low weight at 1 year-old has been suggested as LV hypertrophy risk factor, concordant with Barker's Theory of the fetal and early life origin of chronic disease94.

Clinical validity and impact of such factors is controversial, but it may be important to consider them as relevant potential confounders in epidemiological studies investigating the role of novel risk factors in LVH and the role of LVH in disease prediction.