Advances in handheld echocardiography

Standard portable echocardiography (STAND) machines are expensive and have limited battery capacity. Handheld (HAND) battery-powered devices cost less, but have limited functionality (table 2). HAND has previously demonstrated utility for assessment of cardiac function, chamber size, pericardial effusion and congenital heart disease.32 ,33 Additionally, HAND performs well in the assessment of mitral valve function.32

In 2014, Beaton et al34 reported the first data on use of new-generation HAND (GE, VScan) for detection of RHD in LMICs. In a controlled clinical environment with an artificially high prevalence of RHD (32.8%), HAND showed 90% sensitivity (100% for definite RHD) and 93% specificity compared with STAND for RHD detection. A large school-based study of HAND followed with 1420 children undergoing blinded HAND and STAND studies. In this screening environment, the performance of HAND decreased slightly with 79% sensitivity (98% for definite RHD) and 87% specificity.29 Ten per cent of HAND studies were evaluated for intraobserver and interobserver reliability in this study; self-agreement ranged from 71% to 94% (k=0.70–0.84) and inter-reviewer agreement ranged from 67% to 83% (k=0.34–0.46). There is need for a consensus around what would be an acceptable sensitivity threshold for the use of HAND. Examples of definite and borderline RHD comparing HAND with STAND are shown in figure 1.

Despite early promise, there remain technical challenges with HAND. The battery life is short, necessitating carriage of extra units and/or access to electricity during screening. The devices were designed for intermittent use; therefore, with continuous use they overheat, requiring cooling before further use. Patient information can be entered only through voice recording. Image storage and sharing is in a non-digital imaging and communications in medicine (DICOM) format and measurements on the device are cumbersome. Moreover, the colour Doppler settings cannot be adjusted and tend to result in overestimation of the presence and length of valvular regurgitation.29 ,34

Next-generation devices will likely have growing functionality—potentially including spectral Doppler, frequency shifting, longer battery life and text-entry of patient information. Integration of applications (APPs), remote wireless uploads and multiple export formats are also possible. Ultimately, HAND functionality could be available in standard tablets and smart-phones with easy to connect transducers, further reducing costs.

Advancements in the development of simplified screening protocols

Implementing the WHF criteria requires a high degree of time and training.24 Use of these criteria may not be practical for field application where large volume screening and skilled practitioners are required.22 ,35 Additionally, screening differs from diagnosis in that high sensitivity (to avoid missed cases and opportunity for early intervention) is of relatively greater importance than high specificity (false-positive screens), though this must be carefully weighed in the context of limited subspecialty resources for follow-up evaluations and the stigma of false-positive exams.22

In 2012, Mirabel et al36 raised the possibility that a single echocardiographic criterion might suffice for RHD screening. The group compared mitral regurgitation (MR) ≥2 cm with reference criteria (similar to the WHF criteria, not yet published) in a retrospective look at 2370 STAND screening echocardiograms from Mozambique. The positive predictive value of the simplified approach was 92%. While not statistically significant, 4 of 15 RHD cases (25%) were missed using the simplified approach,36 indicating further refinements might improve results.

Lu et al then evaluated single and combined simplified criteria through a retrospective review of Ugandan HAND screening studies.29 ,37 MR ≥1.5 cm and/or any aortic insufficiency best balanced sensitivity (73.3%) and specificity (82.4%), with near perfect sensitivity for definite RHD (97.9%).37 Morphological criteria were neither sensitive nor specific,37 and have shown poor inter-reviewer reliability compared with functional assessment.29 ,34 Similar results were found during phase 1 of a non-physician HAND study in New Caledonia that demonstrated optimal performance for RHD detection when employing the simplified criteria of MR ≥2.0 cm and/or any aortic insufficiency.30

Further studies of task shifting and field-testing of these criteria are needed (see next section). It will also be imperative to determine the ideal age(s) and frequency of screening. While most of the data on screening is focused on children, two studies highlight a greater prevalence in older children and young adults, making the case for screening adult populations as well.9 ,38 ,39 Additionally, a single set of simplified criteria may not be ideal for all situations. To date, RHD screening has relied on a two-stage process, with confirmatory diagnosis and initiation of secondary prophylaxis following expert review. When thinking about scaling up in LMICs, this approach could overwhelm the most resource-limited environments. Further investigations into a single-stage screening and secondary prophylaxis programmes, best integrated into a primary healthcare system, would improve feasibility, but would require additional consideration of acceptable sensitivity and specificity.

Advancement in task-shifting RHD screening to non-experts

The shortage of physicians in LMICs remains a barrier to echocardiographic screening. Task-shifting, or reallocation of clinical tasks to non-physician health workers could increase access, decrease cost, and free higher-level providers to engage in more complex tasks.40 Brief and focused echocardiographic training, use of simplified diagnostic algorithms and follow-up expert confirmatory echocardiography for positive screens have been central to pilot programmes testing this approach.

In 2013, Colquhoun et al conducted a pilot study using focused training in STAND and implementation of a simplified screening algorithm (MR ≥1.5 cm). Two nurses showed 83%–100% sensitivity and 67%–79% specificity for RHD detection in a blinded cohort of 50 children.41 In 2015, Mirabel et al30 conducted a field test, combining focused training on HAND and a simplified screening protocol (MR ≥2 cm and/or any aortic insufficiency) and found that two non-physicians had a 77.6%–83.7% sensitivity and a 90.9%–92% specificity for RHD detection.. In this study, the concordance for any RHD and definite RHD was 91.4% (k=0.57, CI 0.50 to 0.65) and 91.8% (k=0.53 CI 0.44 to 0.61), respectively. Similarly, Ploutz et al42 evaluated the performance of Ugandan nurses using HAND and found a sensitivity of 74.4% (CI 58.8 to 86.5%), improving to 90.9% (CI 58.7 to 98.5%) for definite RHD.

In order to replicate early successes, standardised training protocols and competency testing for non-experts will be needed. Recently, Engelman et al35 published their experience with an 8-week non-physician training programme combining didactic lessons and hands-on experience. Following completion, all seven participants could obtain high-quality imaging, and no cases of RHD were missed in the practical exam. Excitingly, this experience has been used to create a series of freely available, interactive, web-based modules (http://www.wiredhealthresources.net/mod-rheumatic-heart-disease.html). Larger-scale, multisite testing of training, echocardiography protocols are needed to ensure generalisability.

Additionally, if and when it becomes appropriate to scale-up non-physician screening, it will be important to remember that a single strategy may not be appropriate for all settings. It will be important to consider local needs and workforce restrictions, in particular settings where non-physicians cannot, by law, interpret echocardiograms. In these settings, a strategy of non-physician image obtainment with expert interpretation may be employed. A second consideration will be choosing the most appropriate simplified protocol. This too may vary by setting, as there are differences among endemic areas on rates of early aortic valve involvement and mitral stenosis. Finally, rigorous evaluation of different strategies and screening protocols to weigh sensitivity versus specificity and cost-effectiveness in different environments will be needed to ensure responsible investment of limited healthcare dollars.

Remaining questions to be answered before broadly recommending echocardiographic RHD screening in endemic areas

Advanced RHD exerts an enormous toll on the people and healthcare budgets of LMICs.31 Early detection of RHD allows for the initiation of inexpensive penicillin injections that could prevent disease progression in the majority of children with latent RHD. However, healthcare budgets in RHD endemic nations are small with many competing priorities. Important questions remain before universal echocardiographic screening can be advocated in LMICs, including determining the clinical significance of latent RHD, examining the impact of RHD screening on children and communities, and evaluating the cost-effectiveness of large-scale echocardiographic screening (box 1).

Determining the clinical significance of latent RHD

RHD fulfils most requirements for appropriateness of population-based screening as it is a condition of significant magnitude, is detectable by echocardiography and is treatable by long-acting penicillin.22 ,43 However, doubts remain that screening, diagnosis and intervention in early stages will provide a better prognosis than intervention when clinical symptoms develop. The natural history of patients with clinical ARF and RHD is well established. The earlier in the disease process and the less severe the valvular involvement, the more likely it is that valvular changes will stabilise or regress with secondary prophylaxis.44 The natural history of children with echocardiographic features of RHD, but absent history of ARF, is less certain.22 ,43

Longitudinal follow-up of children with latent RHD is beginning to provide answers. Cohorts from Nicaragua, India, Uganda and South-Pacific New Caledonia have reported outcomes ranging from 5 to 43 months after initial diagnosis.9 ,21 ,28 ,45 ,46 Children with borderline or possible RHD (differences in diagnostic criteria) have shown that 49%–69% remained stable and 21%–42% showed disease regression.9 ,21 ,45 Importantly, 4%–12% showed clinical progression during this time21 ,28 ,46 and several developed recurrent ARF.21 Children initially diagnosed with latent, but definite RHD were at higher risk of disease persistence with 75% remaining definite, 25% regressing to borderline and none showing complete resolution.21 Risk factors for disease persistence or progression include younger age (p=0.05), higher antistreptolysin O titres at diagnosis (p=0.05) and more morphological valve abnormalities (p=0.013).21 Importantly, in all but one study,45 children with borderline/possible RHD were not receiving secondary prophylaxis, further emphasising the likely benign nature of the earliest of valvular changes, at least with short-term follow-up.

Contrasting this, a recent prospective follow-up study of Australian children demonstrated a less favourable prognosis for those diagnosed with borderline RHD.47 This study was the first to use a case–control design, the 2012 WHF criteria from onset, and to report medium-term outcomes (2.5–5 years). Children with borderline RHD (n=55) were compared with matched normal controls (n=104). One-third of children with borderline RHD had been prescribed secondary prophylaxis, while the remaining two-thirds were being followed clinically. Children with borderline RHD were at a significantly greater risk of ARF, progression of valvular disease and development of definite RHD compared with matched controls.47 Conversely, Mirabel et al28 found no significant risk in terms of incidence of ARF between RHD and non-RHD groups after a medium follow-up of 2.6 years.

It will take time to fully understand the natural history and the prognostic value of latent RHD. What is clear is that children with latent, definite RHD are at high risk of disease persistence and should be placed on secondary prophylaxis. While not reaching the large numbers of children with borderline RHD, the prevalence of definite RHD has consistently been ≥1% in screening studies, and likely justifies RHD screening in its own right. There is also growing evidence that at least some children with borderline RHD are at high risk of recurrent ARF and disease progression. Given differences in clinical practice regarding antibiotic prophylaxis in this population, large-scale, prospective, registry-based data collection will likely be the best way to increase our understanding of risk in these children.

Evaluating the impact of RHD screening on children and communities

There remains little investigation on the impact of echocardiographic RHD screening on asymptomatic children and communities. Screening children for other cardiovascular disease has shown potential harmful effects. A historical study 50 years ago showed that diagnosis of a cardiac abnormality resulted in 40% of primary caregivers placing significant restrictions on children's activity, even when these children were later identified as normal.48 Additionally, these same children showed negative cognitive effects, with significantly decreased scores on the verbal and performance scores of the Wechsler Intelligence Scale for Children.49 These findings highlight the potential negative impact of misdiagnosis, and the importance of parental and child perception on health related quality of life (QOL). A single pilot study from Australia has looked at the impact of RHD diagnosis on QOL. Caregivers reported lower QOL scores both for themselves and for the screened child raising concerns regarding the potential negative impact of echocardiography-based RHD screening.50 Wark et al51 also reported lower child and caregiver QOL scores in a pilot RHD screening programme. Conversely, a survey of New Zealand caregivers reported universally positive impressions of participation in RHD screening.52 Larger studies and longer follow-up are needed to determine the significance of these early findings.

Calculating the cost-effectiveness of echocardiographic RHD screening

Given limited resources in RHD endemic settings, public health strategies targeting RHD should be practical and affordable. Watkins et al53 reported that a 10-year ARF-RHD control programme in Cuba, which included primary and secondary prevention strategies, reduced morbidity and mortality and was also cost-effective. Data on the cost-effectiveness of echocardiographic screening is encouraging, but limited to two studies using Markov modelling.54 ,55

In the first, Manji et al compared three prevention strategies to the cost of ‘doing nothing’: (1) throat swab and short course of antibiotic treatment, (2) antibiotic prophylaxis for all children in the age range of 5–21 years, (3) echocardiographic screening and prolonged antibiotic prophylaxis of those with evidence of early RHD. Using direct and indirect cost estimates from sub-Saharan Africa, all three strategies were considered cost-effective, but echocardiographic screening emerged with the most favourable cost-effectiveness ratio and incremental cost-effectiveness ratio (ICER).54 In the second, Zachariah and Sammaliev compared Quality-adjusted life year and society costs of no-screening with a two-stage echo-screening programme (step 1: screening by technicians; step 2: full echo by cardiologists for positive screens) using cost estimates from the Northern Territory of Australia.55 The two-step echocardiographic strategy emerged dominant, with lower total cost and ICER. No comparison with primary prevention strategies was made.

There is more work to be done before echocardiographic screening can be considered cost-effective. Modelling is an important first step, but direct cost data from functional programmes must be analysed, and different programmatic strategies (type and number of personnel, type and number of equipment, optimal screening age and interval, combination of primary and secondary programmes) will need to be considered. It is clear from the sensitivity analysis preformed by Manji et al that cost-effectiveness can vary greatly with factors including direct and indirect cost of RHD, the cost of antibiotics, compliance with prescribed antibiotics and number of screened persons per technician per year, and that given changes to these, other strategies can emerge dominant.54 Further research into optimal screening strategies and the natural history of latent RHD as well as further infrastructure developments to ensure the facilities, personnel and resources needed to care for patient with RHD will increase the accuracy of these measures.