Prognosis associated with geometric patterns of left ventricular remodeling: systematic review and network meta-analysis [version 1; peer review: 1 approved with reservations, 1 not approved]

Background: There are four geometric patterns (normal geometry, concentric remodeling, concentric and eccentric hypertrophy) used to describe cardiac remodeling. Although left ventricular hypertrophy (LVH) is associated with adverse prognosis, the incremental prognostic value of geometric patterns is less certain. We examined characteristics and prognosis associated with the four conventional patterns of left ventricle (LV) remodeling. Methods: A comprehensive literature search was performed on MEDLINE/PubMed, Embase and the Cochrane Library until January 2019. Network meta-analysis was used to pool data from direct and indirect prognostic comparisons of the four geometric patterns. All-cause mortality was defined as the study outcome. Results: A total of 22 echocardiographic studies (76,142 individuals; 50.1% males; 64.4±7.9 years) of diverse cardiovascular diseases were included. Concentric LVH was associated with the highest prevalence of cardiovascular risk factors and diseases; and eccentric hypertrophy


Introduction
The left ventricle (LV) remodels as a response to cardiovascular disease and myocardial injury. Characterized by an increase in LV myocardial mass, left ventricular hypertrophy (LVH) is an established predictor of poorer cardiovascular outcomes 1 .
Four classical geometric patterns of LV remodeling have been defined based on LV mass and relative wall thickness: normal, concentric remodeling, concentric and eccentric hypertrophy 2 . This convenient approach of characterizing LV remodeling has been studied across various patient populations, including patients with coronary artery disease, aortic stenosis, hypertensive heart disease and community-based general populations 3-6 . Whilst some studies demonstrated prognostic associations with these patterns of LV remodeling, others have not. Knowledge of remodeling patterns (concentric and eccentric hypertrophy) provided particularly limited incremental prognostic information beyond LVH 1,[7][8][9][10] .
In this study, we aim to conduct a comprehensive systematic review and network meta-analysis to examine the characteristics and prognosis associated with the four conventional geometric patterns of LV remodeling.

Methods
Literature search and eligibility criteria A comprehensive literature search was performed on MEDLINE/ PubMed (1946 onwards), Embase (1974 onwards) and the Cochrane Library (1996 onwards) until January 2019. Full text publications evaluating the four conventional LV geometry patterns (normal geometry, concentric remodeling, concentric and eccentric hypertrophy) and prognosis were included. The basic search protocol and specific terms used in the search strategy are available as Extended data 11 . We conducted the literature search using Medical Subject Headings or Emtree, and free text terms. There were no restrictions on language.
Two investigators (Q.Z. and G.L.) independently searched for eligible studies based on the pre-defined eligibility criteria. Fulltext studies that compared the prognosis of the four conventional LV geometry patterns (i.e. normal geometry, concentric remodeling, concentric and eccentric hypertrophy) were included. We excluded publications in non-adult populations, case reports, commentaries, abstracts, letters-to-editors and review articles. The bibliography in the identified publications and review articles were also reviewed.

Data extraction and quality assessment
The following data were extracted in duplicates by the two investigators (Q.Z. and G.L.) from the included studies: (1) study characteristics (publication year and patient population); (2) baseline characteristics (mean age, sex distribution, and proportion of patients with hypertension, coronary artery disease, diabetes and other significant risk factors); (3) the four LV remodeling patterns; and (4) adverse prognosis defined as all-cause mortality. Eligible studies that did not report allcause mortality as an end-point were still included to examine clinical characteristics associated with geometric patterns of LV remodeling. Any disagreements were resolved by discussion with a third investigator (C.W.L.C.). In publications with survival curves, the cumulative survival rates were estimated by digitizing the plots (WebPlotDigitizer version 3.9, Austin, Texas, USA).
Two investigators (C.W.L.C. and Z.Q.) independently appraised the quality of each study using the Quality In Prognosis Studies tool 12 . Six domains (study participation, study attrition, prognostic factor measurement, outcome measurement, study confounding; and statistical analysis and reporting) were evaluated to assess the risk of bias in the prognostic studies. In each of the six domains, the risk of bias was classified as "low", "moderate" or "high".

Statistical analysis
A network geometry of the four LV remodeling patterns was constructed. Each node represented a remodeling pattern and its size was weighted by the number of individuals in that group. The connecting line between two nodes denoted direct comparison and its thickness reflected the number of studies included.
The random-effects meta-regression models were used to measure the impact of baseline characteristics on the effect size of the outcome. The risk ratio (RR) of each LV remodeling group was estimated and reported in the study. To rank the prognosis of all the geometric patterns, we used surface under the cumulative ranking (SUCRA) values 13 . Rank probabilities of all the groups were first estimated, then followed by a step function to summarize the cumulative ranking for estimating the SUCRA values of each group, ranging from 0 to 100%. Larger SUCRA values indicated better prognosis.
Both node-splitting and inconsistency modeling were used to test the consistency assumption. The former method involved fitting a series of node-splitting models, one model for each group pair in which there was direct and indirect comparisons. In the latter method, an inconsistency model was fitted and the global Wald test would determine if significant inconsistency was present 14 . Statistical analyses were performed using Stata/ MP Version 13 (StataCorp., College Station, Texas, USA), with the network and network graphs package.

Studies and participants
From an initial 257 publications, 22 echocardiographic studies of diverse cardiovascular diseases satisfied inclusion/exclusion criteria and were included in this study ( Figure 1) 6-10,15-31 . The thresholds used to define LVH and increased concentricity were heterogeneous across the studies (Table 1).
Of the 76,142 individuals pooled from the 22 studies (50.1% males; 64.4±7.9 years), 49.7% had normal geometry; and 31.1%, 10.5% and 8.7% had concentric remodeling, concentric and eccentric hypertrophy, respectively. The proportion of females with concentric and eccentric hypertrophy was high (40-45%). Compared to the other geometric patterns, concentric hypertrophy was associated with the highest prevalence of cardio-metabolic risk factors and cardiovascular diseases. Eccentric hypertrophy was associated with a high prevalence of atrial fibrillation and low LV ejection fraction (Table 2).

Adverse prognosis associated with geometric patterns of left ventricular remodeling
Most of the studies demonstrated low risk of bias in the six domains examined (Table 3). The network geometry of LV remodeling patterns was constructed in Figure 2. Concentric remodeling was associated with higher all-cause mortality compared to normal geometry ( . Based on the SUCRA values, the geometric patterns ranked from best to worst prognosis were: normal geometry, concentric remodeling, eccentric hypertrophy and concentric hypertrophy ( Figure 3).
Results from both node-splitting method and inconsistency model showed no evidence on the violation of consistency assumption between direct and indirect comparisons. Specifically, the pooled estimates between models of consistency (red diamonds) and inconsistency (green diamonds) were identical because all the studies included the four remodeling patterns (Figure 4).

Discussion
In this systematic review and network meta-analysis of 22 echocardiographic publications (n=76,133 individuals), we report the characteristics and prognosis associated with the * Two populations were studied in the same publication.
RWT regional wall thickness; HFpEF heart failure preserved ejection fraction; EF ejection fraction; TAVI transcatheter aortic valve implantation     The pathophysiology of LVH has been well described and studied for the past 50 years 32,33 . Cardiac hypertrophy is initially an adaptive response to the wall stress according to the Law  of LaPlace. Ultimately, cardiac decompensation occurs as a consequence of myocyte death and myocardial fibrosis 34,35 . Whilst geometric patterns of LV remodeling are clinically meaningful to describe the hypertrophic response due to mechanical stress from either pressure (concentric hypertrophy) or volume overload (eccentric hypertrophy), it may not adequately identify the transition point where adaptive hypertrophy decompensates ( Figure 5). This transition point before cardiac decompensation occurs is an important potential risk marker to target more intensive management and closer surveillance. In this study, we have demonstrated that both concentric and eccentric hypertrophy were associated with similar risks of increased all-cause mortality. These observations may suggest that the risk of adverse prognosis is increased once LVH develops, regardless of geometric patterns. It may also suggest that some patients with concentric or eccentric LVH may be in the compensated phase and begets the question of whether there are other strategies to identify high-risk LVH phenotypes.
To address the complex interaction between LV dilatation and myocardial thickening in the pathophysiology of LVH, several studies have recently examined an expanded four-group LVH classification: dilated/non-dilated concentric hypertrophy and dilated/non-dilated eccentric hypertrophy [36][37][38][39][40] . In this proposed four-group LVH classification, dilated concentric hypertrophy was associated with the worst prognosis and non-dilated eccentric hypertrophy had the most favourable profile 36-39 . However, more guidance is needed before this complex classification can be integrated into routine clinical practice. Recently, we have developed the remodeling index (RI), based on a biophysical model of Laplace's Law. The RI integrates LV volume and myocardial thickening into a single measurement 41 . We further demonstrated that hypertensive LVH patients with abnormally low RI (suggestive of excessive myocardial thickening relative to LV dilatation) had increased myocardial fibrosis, elevated circulating markers of myocardial injury and wall stress; and in a small number of patients with dilated cardiomyopathy, an abnormally high RI (suggestive of excessive LV dilatation relative to myocardial thickening) was associated with adverse cardiovascular events 41 . The prognostic value and clinical utility of this index are currently being examined in a large cohort of hypertensive patients (ClinicalTrials.gov identifier: NCT02670031).
These emerging data support the notion that cardiac remodeling in hypertrophy is heterogeneous and complex; and the conventional geometric patterns of LV remodeling is not adequate to risk-stratify patients with LVH.

Study limitations
The study populations included in the meta-analysis were heterogeneous. It is possible that the conventional remodeling patterns has incremental prognostic value in certain cardiac conditions. Unfortunately, the limited number of studies precluded stratified analyses to examine the prognostic value of LV geometric patterns in the different cardiovascular conditions. The definitions used for classifying geometric patterns were not consistent across the different studies. This is concerning and reinforces the necessity to apply consensus definitions in future studies 2 .

Conclusions
Concentric and eccentric hypertrophy are associated with increased and similar all-cause mortality. Possible explanations for these observations include the heterogeneous populations, inconsistent definitions used in the classification and the inherent limitations of the conventional patterns of LV geometry to adequately risk stratify LVH. Well-validated novel approaches to improve risk stratification of LVH should be explored in future research.

Data availability
Underlying data All data underlying the results are available as part of the article and no additional source data are required.

Grant information
The author(s) declared that no grants were involved in funding this work.

Tee Joo Yeo
National University Heart Centre Singapore, National University Health System, Singapore, Singapore Zheng et al performed a systematic review and meta-analysis of echocardiographic studies that included prognostic characteristics of 4 geometric patterns of cardiac remodeling.
Through the combined total of 22 studies comprising diverse populations, including hypertension, aortic stenosis, heart failure with preserved ejection fraction, ischemic strokes, chronic kidney disease etc, the authors conclude that the presence of concentric LVH and eccentric hypertrophy was associated with a higher risk of all-cause mortality.
The overall aim of this study is commendable. Nonetheless, there are a few areas of note: The study includes populations both with and without pre-existing cardiovascular disease; in such a situation, the baseline risk profiles differ significantly (eg. an apparently healthy individual with newly diagnosed hypertension has a substantially lower risk profile compared to an individual with pre-existing or longstanding aortic stenosis although both may develop LVH eventually).

○
A suggested approach would be to stratify the risk of LVH based on aetiology (e.g. hypertension-related, aortic stenosis related, multifactorial etc.) which would make the meta-analysis more applicable to the above sub-groups of patients rather than providing a single risk ratio for an entire phenotypic category of cardiac remodeling.

○
The authors may consider excluding the studies involving the general population to limit the study population to those with pre-existing disease, as the absolute number of Framingham subjects with LVH was actually small. ○ Importantly, the classification of LVH and RWT differed in certain studies, so applicability is affected. Although the authors listed this as a limitation, studies with substantial differences in quantifying LVH might need to be excluded in order to keep the study population ○ homogenous.
In Table 2, regional (?or relative, as stated by the American Society of Echocardiography and other international echocardiography societies) wall thickness for all 4 categories of cardiac remodeling ranged between 0.95 to 1.29. Do check if this is correct, as that would classify all subjects in the concentric geometry.
○ For the conclusion, the authors are right to say that the study populations were heterogeneous. As such, it might be more appropriate to generate different risks for the larger subsets of patients (eg. LVH from hypertension vs LVH from aortic stenosis vs LVH from combined causes) rather than to generalise risk for all patients regardless of aetiology.
○ Are the rationale for, and objectives of, the Systematic Review clearly stated? Yes

Are sufficient details of the methods and analysis provided to allow replication by others? Yes
Is the statistical analysis and its interpretation appropriate? I cannot comment. A qualified statistician is required.

Are the conclusions drawn adequately supported by the results presented in the review?
Partly