Latent Rheumatic Heart Disease: Identifying the Children at Highest Risk of Unfavorable Outcome

Abstract

Background

Screening echocardiography (echo) has emerged as a potentially powerful tool for early diagnosis of rheumatic heart disease (RHD). The utility of screening echo hinges on the rate of RHD progression and the ability of penicillin prophylaxis to improve outcome. We report the longitudinal outcomes of a cohort of children with latent RHD and identify risk factors for unfavorable outcomes.

Methods

This was a prospective natural history study conducted under the Ugandan RHD registry. Children with latent RHD and ≥1 year of follow-up were included. All echos were re-reviewed by experts (2012 WHF criteria) for inclusion and evidence of change. Bi- and multi- variable logistic regression, Kaplan-Meier analysis, as well as Cox proportional hazard models were developed to search for risk factors for unfavorable outcome and compare progression-free survival between those treated and not treated with penicillin. Propensity and other matching methods with sensitivity analysis were implemented for the evaluation of the penicillin effect.

Results

Blinded review confirmed 227 cases of latent RHD: 164 borderline and 63 definite (42 mild, 21 moderate/severe). Median age at diagnosis was 12 years and median follow-up was 2.3 years (IQR 2.0–2.9). Penicillin prophylaxis was prescribed in 49.3% with overall adherence of 84.7%. Of children with moderate/severe definite RHD, 47.6% had echo progression (including 2 deaths), and 9.5% echo regression. Children with mild definite and borderline RHD showed 26% and 9.8% echo progression and 45.2% and 46.3% echo improvement respectively. Of those with mild definite RHD or borderline RHD, more advanced disease category, younger age, and morphological mitral valve features were risk factors for an unfavorable outcome.

Conclusions

Latent RHD is a heterogeneous diagnosis with variable disease outcomes. Children with moderate to severe latent RHD have poor outcomes. Children with both borderline and mild definite RHD are at substantial risk of progression. While long-term outcome remains unclear, the initial change in latent RHD may be evident during the first 1–2 years following diagnosis. Natural history data is inherently limited and a randomized clinical trial is needed to definitively determine the impact of penicillin prophylaxis on the trajectory of latent RHD.

INTRODUCTION

RHD is the most common acquired cardiovascular disease among children and young adults. In 2015, there were an estimated 33.4 million prevalent clinical cases with 319,400 deaths and 10.5 million disability adjusted life years¹. The global distribution of RHD is uneven and strongly associated with conditions of social deprivation. The highest age-standardized RHD death and prevalence rates are found in South Asia, Oceania, and Central Sub-Saharan African populations¹ though the disease remains endemic in most low and middle income countries (LIMC)².

A recent multinational registry (REMEDY) in LIMC showed most RHD patients have advanced RHD (63.9%) and complications at time of diagnosis³. However, RHD is a cumulative process and opportunities exist for early intervention. While the initial episode(s) of acute rheumatic fever (ARF)/RHD almost exclusively occur in childhood, RHD most commonly presents in adolescents and young adults^{4, 5}. The latent period between ARF and clinically apparent RHD presents an opportunity for early intervention.

Screening echocardiography (echo) has emerged as a potentially powerful tool for early RHD detection^6–8. RHD diagnosed through screening echo has been termed latent RHD, and classified as borderline or definite⁹. In screening over 15,000 school children in Uganda, our team has detected a 3% prevalence of latent RHD^{10, 11}. These findings have been replicated in other parts of Africa^{12, 13}, Asia^{14, 15}, South and Central America^{8, 16} and the Pacific^17–20 – uncovering a large population that may benefit from early detection.

The utility of screening echo in RHD endemic populations hinges on two critical parameters: (1) the rate of RHD progression and (2) the ability of Benzathine penicillin G (BPG) prophylaxis to improve outcome, a critical requirement of evaluating the suitability of any condition for screening²¹. Policy decisions on the role of screening for RHD have stalled because of the lack of high quality data to guide management of latent RHD.

Natural history studies have the potential to estimate the rate of latent RHD progression. In addition, in the absence of a randomized controlled trial, natural history studies may provide information on the ability of penicillin to modify disease course. There have been eight previous published natural history studies of latent RHD^{8, 22–28}. Three of these were published prior to the publication of standardized criteria for latent RHD diagnosis^{8, 23, 26}, making them difficult to compare. The remaining 5 studies^{22, 24, 25, 27, 28} report on relatively small cohorts, ranging from 25–55 children limiting the power to detect features that predict progression or regression of disease.

We have studied the largest cohort of children with latent RHD to date, with the hypothesis that risk factors for unfavorable outcome can be found through study of natural history data. The included children are part of the Ugandan National RHD registry (The Registry), enrolled during previous research studies on RHD prevalence¹⁰ and use of handheld echo^{11, 29, 30}, and during RHD outreach and an epidemiological study (unpublished) in Uganda.

METHODS

Setting

This study took place in Uganda under the auspices of the Ugandan National RHD Registry (The Registry), which is IRB approved by Makerere University, Kampala Uganda, Case Western Reserve University, Cleveland Ohio, and Children’s National Health System, Washington DC. The Registry was established in 2010 to capture the burden of clinical RHD presenting to the Ugandan Heart Institute, the single tertiary cardiology center in Uganda, and to support children with latent RHD detected thorough school-based screening. In 2015 The Registry expanded to the North (Gulu) where children with latent RHD detected through school screening were added. The majority of children in this study come from two large cohorts – children screened as part of the 2011 epidemiological study in Kampala, the Ugandan capitol and largest city¹⁰ (previous 2-year outcomes of this cohort reported²²) and children screened as part of ongoing collaborative research and clinical outreach in Gulu (2013-present)^{11, 30} (Figure 1).

Figure 1.

Map of Uganda showing the two districts of enrollment.

Participants

Any child with latent RHD (borderline or definite RHD according to the 2012 WHF criteria⁹) enrolled in The Registry and with at least 1 year of follow-up data was eligible for study participation. In addition, a 3-month effort was made to contact patients who had been lost to clinical follow-up (no follow-up in last 6 months for definite RHD and no follow-up in last 12 months for borderline RHD) and bring them back for regular clinical care and registry enrollment (if not previously enrolled). Written informed consent and/or assent (≥ 8 years) for this project was obtained from all children and families during registry enrollment.

Follow-up Assessment

The Registry captures demographic, socioeconomic, and clinical information, including cardiovascular symptoms, physical exam, medication prescription and adherence at enrollment. Historical data are captured when available and prospective data are gathered at follow-up visits. Registry staff routinely contacts patients who have lapsed from clinical follow-up, and if reached, the reason for absence is noted, including death. For this study, the following variables were abstracted from the first follow-up after screening: age, sex, registry site, and length of follow-up (days). For study purposes, prescription of every-28-day intramuscular Benzathine penicillin G BPG, left at the discretion of the treating provider (Ugandan cardiologist), was determined at initial follow-up visit. Adherence was based on days of coverage (each injection providing 28 days of coverage) when possible and when not possible, as a percentage of prescribed injections received.

Echocardiographic Reporting

A pediatric cardiologist with expertise in RHD (AB, CS) blindly re-reported on the original first post-screening follow-up echos to determine study eligibility. These studies and all follow-up studies were performed on a high-end echo machine (General Electric VIVID Q, Milwaukee, WI or Philips iE33, Best, Holland); results are stored on the Children’s National Health System core echo laboratory PACS system (Philips Xcelera, Best Holland). Diagnosis was made strictly according to the 2012 WHF criteria⁹ and reported as definite RHD, borderline RHD, congenital or other acquired heart disease, or normal. Definite and borderline RHD were further classified into WHF sub-categories (A–C borderline, A–D definite)⁹. Mitral and aortic regurgitation and mitral stenosis were classified as trivial (only regurgitation), mild, moderate, or severe according to published standards^{31, 32}. Children with definite RHD were classified by severity of regurgitation at the most affected valve (mild, moderate, severe), or severe if any mitral stenosis was present at baseline. Children with non-specific valvular abnormalities not meeting criteria for RHD by the 2012 WHF criteria⁹ were excluded, as outcomes for these children have been previously reported as benign^{22, 24}.

All available longitudinal echos (typically every 6 months in definite RHD and every 12 months in borderline RHD) were re-interpreted in series (AB, CS). Serial reporting was chosen, as it is the standard method in clinical practice to determine an individual’s longitudinal outcome. Cases were assessed for echocardiographic progression or regression and the time to event (days) was noted.

Echocardiographic progression was defined as:

• A worsening in diagnostic category (borderline to definite)

• A worsening in the severity of regurgitation at the mitral or aortic valve (none or physiologic, mild, moderate, severe)

• Development of new mitral stenosis or an increase in grade of mitral stenosis (mild, moderate, severe)

• Death due to complication of RHD

Echocardiographic regression was defined as:

• An improvement in diagnostic category (definite, borderline, normal)

• An improvement in severity of regurgitation (severe, moderate, mild, physiologic, none)

Reasons for echocardiographic progression and regression were noted and time to first echo noting change recorded. Cases that did not meet these definitions for progression or regression were considered stable. If a child showed a non-linear outcome (example: initial worsening and then return to baseline), the findings of the last available echo were utilized to determine final outcome.

Since echocardiograms reviews in series were unblinded, two measures of agreement were undertaken to ensure that this procedure did not bias results. First, a subset of randomly selected echos was re-reviewed by a blinded outside cardiologist (MM) with expertise in the 2012 WHF criteria and in latent RHD to calculate inter-rater agreement. Second, a subset of final echos were re-reviewed blindly by the original readers to calculate intra-reviewer agreement on final diagnosis.

Statistical Analysis

Propensity score and other bias-reducing analyses were implemented in Stata (Version 14); all other analyses were performed using SAS (Version 9.3). Descriptive statistics were used to summarize disease progression vs. regression by overall category and sub-category. Cohen’s κappa statistic was used to evaluate the inter-rater agreement of diagnosis. Moderate and severe definite RHD, including all patients with mitral stenosis, were considered “missed clinical” RHD cases rather than true latent RHD. These “missed clinical” cases were excluded from risk factor analysis.

Borderline and mild definite RHD group were combined for risk factor analysis utilizing Fisher’s exact test for categorical variables and the Wilcoxon-Mann-Whitney test for continuous, non-normally distributed, variables. The outcome categories stable, progression, regression were dichotomized in 2 ways in subsequent analyses (Table 1). Outcome 1 compared progression vs. the collapsed favorable outcomes stable or regression. The definition of outcome 2 varied by category of RHD. For mild definite cases, progression or stable was compared to regression (with the rationale that stable definite RHD has clinical significance). For borderline cases, progression was compared to the collapsed favorable outcome stable or regression (with the rationale that stable borderline RHD is of little immediate consequence). Bivariate logistic regression was used to identify individual risk factors for progression, and factors that were significant at p<0.10 were included as covariables in multivariable models. Kaplan Meier curves and log-rank test were utilized to compare the progression free survival distributions of borderline and mild definite RHD.

Table 1.

Binomial Classifications by Outcome.

	Unfavorable	Favorable
Outcome 1	Progression	Stable or Regression
Outcome 2	Mild Definite: Stable or Progression	Mild Definite: Regression
	Borderline: Progression	Borderline: Stable or Regression

Following logistic modeling, multivariable Cox proportional hazards modeling was conducted for outcome 1, adjusting for covariables, to determine risk factors for progression over the time period studied. To substantially reduce differences between the groups receiving and not receiving penicillin prophylaxis at baseline, propensity score matched as well as other forms of matched analysis were implemented. The other forms included regression-based adjustment for confounders, inverse probability weighting and nearest-neighbor matching. Sensitivity analyses were conducted to evaluate the likely impact of uncontrolled differences between the groups receiving and not receiving penicillin.

RESULTS

Two hundred and eighty-two cases of latent RHD (> 1 year of follow-up) were identified in The Registry, with 1,062 corresponding echos. On blinded re-review of the first follow-up echos (after screening), 50 were determined to be non-specific valve disease not meeting the 2012 WHF criteria for RHD, 5 as other heart disease (4 mitral valve prolapse, 1 sub-aortic membrane), and 227 as latent RHD: 63 definite RHD and 164 borderline RHD. Of those with definite RHD, 42 were classified as mild and 21 as moderate-to-severe (Figure 2). Details of echo diagnosis at baseline are shown in Table 2. Inter- and intra-reviewer agreement based on the trinomial classification (normal/other, borderline, definite) was substantial (Kappa (inter)=0.72, 95%CI 0.52–0.93, Kappa (intra)=0.78, 95% CI 0.67–0.88 95%)³³.

Figure 2.

Flowchart of study entry and classification according to CONSORT recommendations.

TABLE 2.

Characteristics of 227 Children with Latent RHD

Demographics
Median Age at Screening (IQR), years		12 (10–14)
Gender – n (%)	Female	137 (60.4)
Enrollment Site – n (%)	UHI	47 (20.7)
	Gulu	180 (79.3)
Median Duration of Follow-up (IQR), years		2.3 (2.0–2.9)
Mean Echocardiograms/Patient (SD)		3.8 (+/ 1.5)
Echocardiographic Characteristics
Definite RHD – n (%)		63 (27.8)
WHF Subcategory of Definite RHD – n (%)
A: Pathological MR + morphological features of the MV		47 (74.6)
B: Mitral stenosis		3 (4.8)
C: Pathological AR + morphological features of the AV		6 (9.5)
D: Borderline disease of the MV and AV		7 (11.1)
Severity of Definite RHD – n (%)	Mild	42 (66.7)
	Moderate	15 (23.8)
	Severe	6 (9.5)
Borderline RHD – n (%)		164 (72.2)
WHF Subcategory of Borderline RHD – n (%)
A: Morphological features of the mitral valve		9 (5.5)
B: Pathological MR		135 (82.3)
C: Pathological AR		20 (12.2)
Severity of Borderline RHD– n (%)	Mild	163 (99.4)
	Moderate	1 (0.6)
	Severe	0
Secondary Penicillin Prophylaxis
Prescribed BPG – n (%)		108 (47.5)
Data Type – n (%)	Continuous (days of coverage)	83 (76.9)
	Categorical (# injections received)	25 (23.1)
Mean Adherence (SD)		84.7 (+/− 0.15)
Adherence ≥ 80% – n (%)		89 (82.4)

The characteristics of the 227 children with confirmed latent RHD are detailed in Table 2. Median age at diagnosis was 12 years, and 60.4% were female. The median duration of follow-up was 2.4 years (range 1.1 to 5.9 years), with 26 children having greater than 5 years of follow-up. BPG was prescribed in 108 children (49.3%), including 59 of 63 children with definite RHD (93.6%) and 49 of 164 children with borderline RHD (29.9%). Of those prescribed penicillin, the overall adherence (≥80% days covered or ≥ 80% injection received) was quite good at 84.7%, with 89 children (82.4%) individually meeting the ≥ 80% benchmark. There were no cases of ARF captured in the registry during the follow-up period.

RHD Status

The outcomes of children with moderate and severe RHD were poor (Figure 3). Out of 21 children, only 2 (9.5%) showed echocardiographic regression during the study period. Of the other 19 children, 9 (42.9%) remained with moderate/severe RHD, 8 (38.1%) progressed and 2 (9.5%) died during the follow-up period (Figure 2). No child underwent a valve intervention during the study period.

Figure 3.

RHD Trajectory by Initial Diagnostic Classification (Favorable and Unfavorable outcomes are shown above the bars in solid (favorable) and dashed (unfavorable) bars for Outcome 1 (O1) and Outcome 2 (O2).)

Among children with mild definite RHD (n=42), 11 (26.2%) children showed progression, 12 (28.6%) children remained stable, and 19 (45.2%) children showed echocardiographic regression. Penicillin was prescribed to 39 out of 42 children with mild definite RHD (92.9%) and 27 of 39 (69.2%) had adherence ≥ 80%, reflecting effective prophylaxis in 64.3% of children with mild definite RHD (27/42). The most common reasons for progression included worsening of mitral regurgitation (54.5%) and worsening of aortic regurgitation (36.4%). The most common reasons for regression included improvement in mitral regurgitation (65.0%), and improvement of mitral valve morphology (35.0%). Full details on progression and regression are provided in Table 3.

Table 3.

Breakdown of Echocardiographic Changes in Cases of Progression and Regression (mild definite and borderline RHD)

Reasons for Progression (n=27 children*)
New pathologic or worsening grade of mitral regurgitation	13 (48.1%)
New pathologic or worsening grade of aortic regurgitation	6 (22.2%)
Newly met criteria for abnormal MV morphology†	15 (55.6%)
Newly met criteria for abnormal AV morphology†	2 (7.4%)
Reasons for Regression (n=95 children*)
Improvement in grade or disappearance of pathologic mitral regurgitation	72 (75.0%)
Improvement in grade or disappearance of pathologic aortic regurgitation	16 (16.7%)
Resolution of Improvement in MV morphological findings‡	7 (7.3%)
Resolution of AV morphological findings‡	5 (5.2%)

^*Some children showed more than one reason for progression or regression;

^†At least 2 morphological features at the mitral valve including thickened anterior mitral leaflet (≥3mm), restricted leaflet motion, chordal thickening, excessive anterior mitral leaflet motion OR at least 2 morphological features at the aortic valve including thickening, prolapse, non-coaptation, or restricted motion;

^‡Less than 2 morphological features of the mitral valve or aortic valve.

Among children with borderline RHD (n= 164), 16 (9.8%) showed progression, 72 (43.9%) remained stable, and 76 (46.3%) showed echocardiographic regression. Penicillin was prescribed to 49 children (29.9%) with borderline RHD, and 44 (90%) had adherence ≥ 80%, reflecting effective prophylaxis in 26.8% of these children (44/164). The most common reasons for progression included worsening of mitral regurgitation (56.3%) and worsening of aortic regurgitation (36.4%). The most common reasons for regression included resolution of abnormal mitral valve morphological findings (35.0%), and improvement of mitral regurgitation (65.0%). The majority of initial progression occurred during the first 2 years of follow-up. Children with mild definite RHD were more likely to show progression than children with borderline RHD (p<0.01,Figure 4). Full details on progression and regression are provided in Table 3.

Figure 4.

Progression free Survival Stratified by Disease Status

There were 5 children who demonstrated a more complex pattern of change. Of these, 4 waivered between normal and borderline RHD based solely on small change in MR jet length (above and below the 2cm cutoff). In these cases, there was no clinically meaningful change, but only variation in diagnostic category due to strict application of the 2012 WHF criteria. In the final case, an initial echo showed definite RHD (thickened anterior mitral leaflet and excessive anterior mitral leaflet motion as well as mild MR). On second echo, the excessive motion had resolved (then borderline because <2 morphological criteria were met), and then on echos 3–5, the posterior mitral leaflet became restricted. It is likely in this child, that this pattern of change simply represents natural RHD progression.

Risk Factors

As penicillin prophylaxis was prescribed to only one-third of children with borderline RHD, before undertaking risk factor analysis, we evaluated whether there were any echocardiographic or clinical differences between children with borderline RHD who were or were not initially prescribed prophylaxis. Length of mitral regurgitation was found to be associated with penicillin prescription (p=0.03), while year of diagnosis, sub-category of borderline RHD (A–C), presence of aortic insufficiency, and abnormal mitral valve morphology were not associated (respectively p =0.19, 0.08, 0.65, 0.72).

Table 4 shows the results of the bivariate analysis for risk factors. In Outcome 1, category of mild definite RHD, morphologically abnormal mitral or aortic valves, pathological aortic insufficiency, and penicillin prescription were more likely to progress. In Outcome 2, the same risk factors emerged, except for the addition of mitral valve excessive anterior leaflet motion as an additional risk factor (Table 4).

Table 4.

Clinical risk factors for unfavorable outcome, bivariate logistic regression analysis

	Outcome 1*		Outcome 2*
Variable	Odds Ratio (95% CI)	P-value†	Odds Ratio (95% CI)	P-value†
Disease Category (mild definite)	3.28 (1.39–7.76)	0.007	10.17 (4.59–22.54)	<0.001
Length of Follow-up (per year)	1.00 (1.00–1.00)	0.648	1.00 (1.00–1.00)	0.353
Age at Diagnosis (per year)	0.88 (0.75–1.03)	0.099	0.88 (0.77–1.01)	0.073
Gender (female)	1.04 (0.45–2.41)	0.925	0.80 (0.39–1.65)	0.551
Mitral Valve Morphological Features
Anterior leaflet ≥ 3mm	2.84 (1.23–6.57)	0.015	3.09 (1.43–6.42)	0.004
Chordal thickening	4.27 (1.16–15.73)	0.029	4.09 (1.18–14.20)	0.027
Restrictive leaflet motion	3.98 (1.58–10.03)	0.003	5.95 (2.55–13.87)	<0.001
Excessive anterior leaflet motion	1.56 (0.48–5.00)	0.459	2.93 (1/13–7.61)	0.027
Aortic Valve Morphological Features
Any WHF morphological abnormality	2.86 (1.12–7.27)	0.028	3.66 (1.65–8.13)	0.001
Functional Valve Abnormalities
Length of MR (per 1 cm)	0.72 (0.45–1.15)	0.165	1.05 (0.68–1.62)	0.815
Length of MR	‡	0.502	-------------	‡
Pathological AR	3.67 (1.54–8.76)	0.003	3.66 (1.65–8.13)	0.001
Benzathine Penicillin G
Penicillin prescription	2.90 (1.24–6.82)	0.014	3.99 (1.85–8.59)	<0.001
Adherence (≥ 80%)	0.94 (0.53–1.67)	0.844	0.73 (0.25–2.09)	0.556

CI: Confidence Interval, WHF: World Heart Federation, MR: Mitral regurgitation, AR: Aortic regurgitation.

^*Outcome 1: Progression vs. Stable + Regression (ALL); Outcome 2: Mild definite RHD (Progression + Stable vs. Regression) and Borderline RHD (Progression vs. Stable + Regression);

^†Type 3 p-value;

^‡Odds ratios for this categorical based on <2cm compared to 2cm, 3cm, and 4cm – none were significant.

In a multivariable model, children who had pathologic aortic insufficiency were more likely to progress (Outcome 1) than those without (OR 5.37, 95% CI 1.54–18.69). In outcome 2, pathological aortic regurgitation remained an independent risk factor in addition to the mitral valve morphological features restricted leaflet motion and excessive anterior leaflet motion (OR 6.09, 95% CI 1.81–20.44, OR 5.29, 95% CI 1.22–22.92, OR 7.31, 95% CI 1.13–46.94, respectively). Increasing age at diagnosis decreased the odds of progression in both models (OR 0.83, 95% CI 0.70–0.99 and 0.84, 95% CI 0.71–0.99) for every year increased age for Outcome 1 and 2 respectively) (Table 5).

Table 5.

Clinical risk factors for RHD progression, multivariable logistic regression analysis

	Outcome 1*		Outcome 2*
Variable	Odds Ratio (95% CI)	P-value†	Odds Ratio (95% CI)	P-value†
Disease Category (borderline/mild definite)	0.58 (0.14–2.37)	0.125	2.24 (0.63–7.97)	0.212
Age at Diagnosis (continuous)	0.83 (0.70–0.99)	0.035	0.84 (0.71–0.99)	0.039
Mitral Valve Morphological Features
Anterior leaflet ≥ 3mm	3.03 (0.29–31.84)	0.355	10.57 (0.69–161.33)	0.090
Chordal thickening	2.55 (0.51–12.86)	0.255	1.56 (0.33–7.43)	0.577
Restrictive leaflet motion	3.39 (0.89–12.83)	0.072	5.29 (1.22–22.92)	0.026
Excessive anterior leaflet motion	‡	‡	7.31 (1.13–46.94)	0.036
Aortic Valve Morphological Features
Any WHF morphological abnormality	2.27 (0.69–7.48)	0.178	1.28 (0.39–4.20)	0.682
Functional Valve Abnormalities
Pathological AR present	5.37 (1.54–18.69)	0.008	6.09 (1.81–20.44)	0.004
Benzathine Penicillin G
Penicillin prescription (yes/no)	2.37 (0.79–7.12)	0.125	1.99 (0.69–5.76)	0.203

CI: Confidence Interval, WHF: World Heart Federation, AR: Aortic regurgitation.

^*Outcome 1: Progression vs. Stable + Regression (ALL); Outcome 2: Mild definite RHD (Progression + Stable vs. Regression) and Borderline RHD (Progression vs. Stable + Regression);

^†Type 3 p-value.

^‡Not included in this model.

Results of Cox proportional hazards regression for Outcome 1 were similar to those obtained using logistic regression. Risk of progression over time was increased for children with pathologic aortic insufficiency (HR = 5.69, 95% CI 2.01–16.14), with hazard ratios for other risk factors similar to the previously reported ORs (data not shown).

After adjusting for differences in sex, age, presence of aortic regurgitation as well as presence and degree of mitral regurgitation, penicillin prophylaxis in children with borderline RHD was associated with increased odds of progression (outcome 2; OR=2.57, 95% CI 0.81–8.19). As indicated in Table 6, propensity score matched and other forms of analyses designed to remove confounding did not alter the above result. All of these analyses produced results consistent with more than a 2-fold increase in the odds of progression (9–10% absolute risk difference) in penicillin users. However, none of the results achieved better than borderline statistical significance (p=0.086 to p=0.098). In addition, sensitivity analysis strongly indicated that accounting for any uncontrolled confounding would likely lead to a further increase in the greater risk of progression associated with penicillin prophylaxis in participants affected by borderline RHD.

Table 6.

Results of Analyses to Remove Extraneous Differences between Penicillin-Treated and Untreated

Method to Remove Extraneous Treatment Group Differences	Average %Risk of Progression		p-value
	Untreated	Difference Treated - Untreated
Regression Modeling	5.2	8.8	0.097
Inverse Prob Weighting	5.2	8.8	0.093
Nearest Neighbor Matching		8.9	0.098
Propensity Score Matching		9.9	0.086

DISCUSSION

A significant knowledge gap remains surrounding natural history of latent RHD found by echo screening. Understanding the natural history of latent, echo-detected, RHD remains a critical component of evaluating the utility of more widespread implementation of echocardiographic screening programs. Here, we report the natural history of the largest longitudinal cohort of children with latent RHD to date looking specifically at risk factors for unfavorable outcomes. Four major findings emerge from our study.

First, not all screen-detected RHD should be considered latent RHD as some children were found to have moderate-to-severe RHD on screening echo and they demonstrated poor outcomes. In our cohort, only 2 such children (9.5%) showed improvement over the study period, and there were 2 deaths (9.5%). Similar findings were recently reported from Fiji, where more than 80% of children with moderate-to-severe RHD at time of screening showed persistence or progression²⁴. When screening reveals moderate to severe RHD, children should be considered “missed clinical” cases, and treated as such, with referral to a specialist provider, given consideration of BPG prescription, and valvular intervention as appropriate and available.

Second, children with mild definite RHD at time of screening echo show more progression than children with borderline RHD at screening. In our cohort, 25% of children with mild definite RHD showed progression compared to only 10% of those with borderline RHD. Progression is challenging to directly compare across cohorts, even in those using the 2012 WHF criteria, due to inconsistent definitions of progression, use of different binomial outcomes (stable + progression), and inclusion of children with advanced RHD. However the ratio of progression among those with mild definite RHD compared to those with borderline in our cohort (2.5:1) is similar to reports from Fiji (2.5:1)²⁴. The progression seen among children with borderline RHD in our study (10%) is also similar to smaller recent cohorts from New Caledonia (5%)²⁸ and Fiji (12%)²⁴, although lower than reported from South Africa (20%)²⁷ and Australia (24%)²⁵.

However, category of RHD itself may not be as great a risk as specific echocardiographic features. Our third major finding was the identification of risk factors for unfavorable outcome based on multiple variable model. While initial RHD category (borderline vs. mild definite RHD) was a highly significant risk factor for unfavorable outcome in our bivariate model, it did not retain statistical significance in the multiple variable model. For outcome 1 (progression vs. stable + regression), only presence of pathological aortic insufficiency emerged as a significant risk factor, while older age at diagnosis was protective from progression. In outcome 2 (mild definite RHD progression + stable vs. regression or borderline RHD progression vs. stable + regression), additional mitral morphological features including restrictive leaflet motion and excessive anterior leaflet motion, were also risk factors.

There are two aspects of progression that are of primary interest in terms of prevention of sequelae, these are first evidence of (initial) progression and rate of progression over time. Our Kaplan-Meier analysis provided important evidence that initial echocardiographic progression may be detected during the first 2 years of clinical follow-up. There is also evidence that RHD progression may not be uniform. The rate of RHD progression is affected by the ongoing inflammatory response from the first insult and repeat insults in the form of repeated GAS exposure and ARF recurrences, providing reasons to suspect non-uniform progression. Data from clinical ARF and RHD are clear that progression of RHD continues over time.^{34, 35} In our study, we are referencing only initial progression. Our data suggests that initial evidence of progression may emerge early in follow-up, which has implications for use of secondary prophylaxis and for retention in care. Longer natural history studies are needed to determine if early evidence of progression is, in fact, predictive of longer-term outcomes including need for future cardiac medications, interventions and/or premature death and whether these key indicators of progression occur at differential rates over time.

Length of follow-up did not emerge as a risk factor in our bivariate analysis. It may be that we were underpowered to find differences, however the previous report had similarly small numbers (19 children with ≥ 5 years follow-up²⁴) and was underpowered to conduct a multivariate analysis.. It is also important to remember that in our study, we are referencing only initial progression. Data from clinical ARF and RHD is clear that progression of RHD continues over time^{34, 35}. However our data suggests that disease trajectory may emerge early in follow-up, which has implications for use of secondary prophylaxis and for retention in care. Longer natural history studies are needed to determine if early trajectory is, in fact, predictive of longer-term outcomes including need for future cardiac medications, interventions and/or premature death.

Fourth, the impact of penicillin on improving the trajectory of latent RHD remains unclear. BPG prophylaxis emerged as a risk factor for progression, a finding also reported from Australia²⁵. This result failed to reach statistical significance but persisted unabated despite extensive efforts, including propensity score matching with sensitivity analyses, to remove extraneous differences between penicillin treated and untreated children. This counterintuitive finding has important implications for prevention of sequelae in persons with latent disease and thus needs to be further investigated in a randomized clinical trial.

As more data emerges on the natural history of latent RHD, it is important to consider standardized reporting of outcomes³⁶. The 2012 WHF criteria represent the best available evidence for the echocardiographic diagnosis and classification of latent RHD, and have been widely adopted facilitating reproducible diagnostic categories⁹. Defining latent RHD trajectory has been less uniform both in what constitutes change and in how trinomial outcomes are grouped into binomial classifications. Here, as reported by several recent studies^{24, 25}, we use an expanded definition of change including but not limited to change in WHF category. Expert agreement on these definitions would facilitate better comparison across populations, but until this agreement, clear reporting of definitions and provision of granular details on reasons for change is essential³⁶. Grouping of disease trajectory into clinically meaningful outcomes is also important. In this study we chose to report outcomes in 2 ways, first, with only progression vs. stable and regression and second, in the way we found most clinically relevant, grouping progression and stable definite RHD as the worse outcome, while only considering progression of borderline RHD as the worse of two outcomes. Interestingly, these groupings did not affect our multivariate analysis greatly, with only additional morphological features emerging in Outcome 2.

When possible, it is also important to include multiple variable analysis to determine which risk factors remain important. In our cohort, many seemingly key variables in bivariate analysis, including disease category and prescription of BPG prophylaxis, were not independent risk factors in our multivariate analysis. In part, this reflects the number of correlated confounders likely to surface in natural history studies. As many longitudinal RHD cohorts remain small, combining data for a larger multicenter analysis remains a top priority.

Limitations

The main limitation in our study was the inability to determine the impact of BPG prophylaxis on the trajectory of latent RHD. This limitation is inherent to all natural history cohorts, reinforcing the need for a well-designed randomized controlled trial in this population. Based on our data, we would advocate the exclusion of children with moderate-to-severe definite RHD at time of screening but believe that equipoise exists to justify inclusion of both mild definite and borderline RHD in such a trial. Similarly, we did not collect data on streptococcal exposure in this cohort and cannot determine contribution of repeated GAS exposure to latent disease trajectory. A second limitation is the relatively small number of mild definite RHD cases included in our cohort (n=42). While comparatively large relative to previous studies^{8, 22–28}, this is still a small number of patients and may be underpowered to detect important risk factors for unfavorable outcome. Thirdly, there is selection bias of our included cases: school-going children at time of screening and only those who were enrolled in the national RHD registry. There are no data to suggest children who do not attend school are at greater risk of latent RHD or of progression, however this is possible considering the lower overall socioeconomic status in this group, a known driver of RHD. Fourth, significant changes in systemic blood pressure can change the appearance and grade of mitral and aortic regurgitation. We did not capture blood pressure at each echo, however of 223 patients with data on age, height, and BP at registry enrollment, 221 were normotensive, suggesting this is likely not a driver of change in disease categorization in our population. Additionally, not all children with screen-detected RHD returned for initial follow-up and/or were enrolled in the national registry. We cannot determine if or how these losses affected our results. Finally, while initial echos were re-reviewed blindly for disease classification, we intentionally chose to review serial echos in an un-blinded fashion (able to look forward and back within a case). This study is among the first to review and report on serial echos during the follow-up period (as compared to only first and last), which has provided important information on timing of initial change. We also believe that side-by-side comparison strengthens the ability of a reviewer to see meaningful differences over time, as is standard clinical practice. Additionally, we invited an expert cardiologist outside our group to review a randomly selected batch of echos, and found substantial agreement in category of diagnosis (normal/other, borderline, definite), indicating no significant bias was introduced through this method.

Conclusions

Latent RHD is a heterogeneous diagnosis with variable disease outcomes. Children with moderate-to-severe latent RHD found by echo screening should be considered missed clinical cases and treated as such, as their outcomes are poor. Children with both borderline and mild definite RHD are at substantial risk of progression, and at a minimum should be enrolled in close clinical follow-up. While long-term outcomes remain unclear, the worsening of latent RHD may be evident through echo assessment during the first 1–2 years following diagnosis. Natural history data is inherently limited by confounding variables and a randomized controlled clinical trial is needed to definitively determine the impact of BPG prophylaxis on the trajectory of latent RHD.

CLINICAL PERSPECTIVE.

• What is New?

  • ◦
    Rheumatic heart disease (RHD) remains the most common cardiovascular disease among the world’s youth.
  • ◦
    Echocardiographic screening provides a promising tool for early detection but the natural history of screen-detected RHD (latent RHD) remains unclear.
  • ◦
    In this large series of patients, we show that children with borderline and definite RHD by echo screening are at substantial risk of progression, 9% and 26% respectively at a median follow-up of 2.6 years.
  • ◦
    Further, we identify younger age at diagnosis, and presence of morphological mitral valve features at diagnosis as independent risk factors for unfavorable outcomes.
• What are the Clinical Implications?

  • ◦
    This study adds to the growing knowledge of latent RHD and the role of echocardiographic screening as a tool to reduce the global burden of RHD.
  • ◦
    Children with screen-detected RHD remain at substantial clinical risk.
  • ◦
    Children with moderate to severe RHD at screening should be considered for treatment as clinically diagnosed RHD.
  • ◦
    Children with borderline or mild definite RHD at screening should, at a minimum, be maintained in close clinical follow-up.
  • ◦
    Additionally, our data highlight the need for a randomized clinical trial in this population to determine the effect of penicillin prophylaxis on the clinical course of RHD.