Article Text
PDF
Abstract
Objectives To assess how much competitive sport contributes to sudden cardiac death (SCD) in young athletes and the impact on population health if this group were to be screened in the UK.
Methods Using reported and imputed incidence rates of SCD in athletes and non-athletes and false-negative and false-positive test rates reported in three key Italian screening studies, the authors calculated the population and attributable risk fractions of SCD in young athletes and the total population (athletes and non-athletes) aged 12–35 years before and after screening; the number of athletes needed to screen (NNS) to prevent one SCD and the sensitivity and the specificity of screening with electrocardiogram. Using these parameters, the authors developed a decision tree model based on the UK population aged 12–35 years to estimate the annual number of SCDs, the expected number of screening and diagnostic tests and the number of athletes disqualified from competitive sport per SCD prevented.
Results Participation in competitive athletics contributes to 81.9% (62.4% to 91.6%) of SCD in athletes but only 26.6% (−20.3% to 55.8%) in the total population. After screening, the contribution in the total population falls to 7.2% (−10.7% to 22.4%). The NNS is 38 151 (20 534 to 267 380). A UK screening programme would result in 1 520 021 young athletes being screened, with 140 361 referred for diagnosis. Of an expected 196 SCDs per year, 40 (6 to 74) would be prevented. For every life saved, 791 athletes would be disqualified.
Conclusions The impact of screening on reducing SCD in young athletes is only modest and would be achieved with significant harms to population health.
Statistics from Altmetric.com
National or state programmes to screen young athletes for signs of cardiovascular disease (CVD) have been in place for more than 20 years in Italy and the USA. Their aim is to prevent sudden cardiac death (SCD) by identifying those with disease and disqualifying them from competitive athletics, an activity reported to double the risk of SCD.1 In the UK, there is no national programme, with current screening practice being ad hoc and primarily undertaken in the private and voluntary sectors.2
In recent years, screening of athletes has gained support from the European Society of Cardiology, international sports associations,3 national voluntary organisations and the media.4,–,8 Leading medical journals have published articles supporting screening,4 5 9 and UK parliaments have debated the issue.10 11
However, support is not universal, with those against screening arguing that there is inadequate evidence of effectiveness and evaluation of value for money.5 12 Recent reviews have not identified any randomised controlled trials of screening, probably because it would be ethically and technically challenging to do them.5 Some reasonable quality observational evidence is available from Italy where there has been a state-sponsored, national preparticipation screening programme since 1982 (see box 1).5 13
The Italian screening programme13
▶ Condition: SCD (unexpected death of natural causes with loss of all functions instantaneously or within 1 h of the onset of collapse symptoms) confirmed by postmortem and pathology.
▶ Test: Annual physical examination, personal and family history and 12-lead ECG.
▶ Intervention: disqualification from competitive sport using more restrictive criteria than in the Bethesda Conferences.
▶ Population: Athletes aged 12–35 years participating in competitive sport.
▶ Comparator: Non-athletes aged 12–35 years.
▶ Exposure: Two-year rate (person-years) calculated using the Sports Medicine Database which records annually the athletes participating in official regional competitions.
▶ Outcome: Deaths from the Registry on Juvenile Sudden Death prospectively collected in Veneto since 1979. SCD rates in non-athletes did not include former athletes who had been screened previously or disqualified.
Before embarking on any screening programme, clear evidence is needed that it will provide more benefit than harm to population health. Italian studies that provide the best available evidence5 have emphasised the benefit of screening in athletes (a tenfold fall in incidence of SCD at 15 years is claimed) but less so the harms, as evaluation of the screening test's performance (sensitivity and specificity) has been only partial.13 14 Thus, the level of unnecessary anxiety, invasive testing and disqualification from competitive sport that would arise because of false-positive tests has not been established nor has the degree of false reassurance been generated by false negatives.
This paper assesses the implications of establishing a UK programme to screen young, competitive athletes in terms of benefits and harms to population health by combining data from three key Italian studies.13,–,15 In addition to the contribution competitive athletics make to the incidence of SCD in young athletes and the total population (athletes and non-athletes), we estimate the annual number of SCDs, the expected number of screening and diagnostic tests, the number of athletes disqualified from competitive sport, the number of SCDs prevented and the ratio of disqualifications per SCD prevented. The purpose was to facilitate a more balanced debate on whether to screen young athletes in the UK.
Methods
First, we describe how we calculated the contribution of competitive sport to the incidence of SCD, the effectiveness of screening and the screening test performance. Then, we describe how our model of screening in the UK combined these parameters with other assumptions to provide our outputs.
Contribution of competitive sport to SCD and the impact of screening
To calculate how much competitive sport contributes to the incidence of SCD in young athletes and people aged 12–35 years before and after screening, we used data from the study of Corrado et al13 that provided the total number of deaths and incidence of SCD in athletes and non-athletes during the prescreening (1979–1981), early screening (1982–1992) and late screening (1993–2004) periods in Italy. We imputed the person-years exposure for athletes and non-athletes for each screening period when not reported by dividing the number of SCDs by the reported incidence. Then, we calculated the total exposure in athlete and non-athlete person-years for each period, plus the postscreening period (early and late screening periods combined). We used the summed deaths and imputed exposures to calculate the incidence of SCD in the postscreening period in athletes, non-athletes and the total population (athletes and non-athletes). The incidence of SCD in athletes only (to avoid confounding) was then used to calculate the attributable risk (AR) and AR fraction (ARF); and in non-athletes and the total population, to estimate the population AR (PAR) and population ARF (PARF). The AR was used to calculate the number needed to screen (NNS) to prevent one SCD in athletes for each period (see box 2).1
Box 2 Explanation of epidemiological terms1
Performance of the screening test and prevalence of CVD
To calculate the sensitivity and specificity of the Italian screening test and the prevalence of CVD (sufficient for disqualification) in young athletes, we used data from two studies.13 14 Both of these had similar proportions of males (80.4%13 vs 74.0%14) though with a slightly different age profile (mean [SD] age 18.9 [6] years13 vs 24 [6] years).14 The number of false negative tests in Veneto was estimated by applying the proportion of false-negative tests reported in the study of Pelliccia et al of 4450 elite athletes followed up at a specialist centre in Rome (using criterion standard diagnostic techniques) to the number of negative tests reported in Veneto.14 This was combined with the number of true positives to calculate the screening test's sensitivity.
Similarly, the number of false positives from the study of Corrado et al of 3914 athletes with positive screening results referred for specialist evaluation was combined with the proportion of true negatives (estimated using the true-negative rates from the study of Pelliccia et al) applied to the number of negative tests in Veneto to calculate the specificity. The prevalence of CVD in athletes was estimated by dividing the number of true positives and false negatives by the total number screened.
Modelling of screening young athletes in the UK
We developed a simple decision tree model that applied the proportion of athletes in the population, the prevalence of CVD in athletes, the sensitivity and the specificity of the screening test and NNS (assuming 100% test uptake rate) to the UK population of 12–35-year-olds.16 This calculated the number of positive and negative screening and diagnostic tests and the number of SCDs prevented. The expected number of SCDs in athletes and non-athletes was calculated by applying the prescreening incidences to our estimated UK populations for these groups.
We also calculated the number of SCDs due to competitive sport that would not be prevented by screening. First, we assumed that athletes would have the same risk of SCD as non-athletes if they did not participate in competitive sport. Using the postscreening incidence in non-athletes, the known ratio of males to females and the relative risk of SCD in male to female non-athletes,1 we estimated the incidence of SCD in male and female non-athletes. We then calculated the incidence in all the athletes by summing these rates, weighted by the proportion of males to females in the athletic population (80:20), and multiplied by the estimated number of athletes that would be screened in the UK. Finally, this figure was subtracted from our estimate of the total number of SCDs prevented by screening.
To assess the balance between potential benefits and harms of screening, we calculated the number of athletes who would be disqualified (the major harm) to prevent one SCD. The sensitivity of this calculation and the number of SCDs prevented was tested over a range of CVD and athlete prevalences and test uptake rates.
Statistics
Confidence intervals (95% CIs) around our estimated incidence and rate differences were calculated using Stata V.8. We used the normal distribution to test for differences between rates using StatsDirect V.2.6.6. A two-tailed p<0.05 was considered statistically significant.
Results
Table 1 presents the reported and estimated incidence of SCD in athletes, non-athletes and the prescreening and postscreening periods and their 95% CIs. The accuracy of the exposure estimates (used for calculating incidence rates) was also corroborated by our estimate of the total exposure in athletes (2 894 737) and non-athletes (33 544 304) over the study period (1979–2004), which only differed from that reported (2 938 730 and 33 205 370) by 1.5% and 1%, respectively.13
Estimated exposure and incidence of SCD for athletes and non-athletes and the population in the prescreening and postscreening periods
From the Italian studies, we estimated that the incidence of SCD for the postscreening period was 1.57 (95% CI 1.13 to 2.13) per 100 000 person-years, statistically lower than that reported in the prescreening period (4.19; 95% CI 2.29 to 7.03) per 100 000 person-years. It was higher than the postscreening incidence in non-athletes (0.80; 95% CI 0.70 to 0.91). In non-athletes, there was no difference between the prescreening and postscreening incidences statistically. Regarding the impact of screening on the population incidence of SCD, although the rate of SCD fell from 1.05 (95% CI 0.77 to 1.41) to 0.86 (95% CI 0.76 to 0.97) per 100 000 person-years, this difference was not statistically significant. This may be because of lack of power, as the prescreening period was relatively short and the number of SCDs was small.
Table 2 summarises the epidemiological measures used to assess the impact of competitive sport and screening in athletes, non-athletes and the total population (athletes and non-athletes) over the prescreening and postscreening periods. It shows that athletes' risk of SCD is 3.42 (95% CI 1.21 to 5.63) times greater than that of non-athletes and that training and participation in competition accounted for four fifths (81.6%; 95% CI 62.4% to 91.6%) of this increased rate. After 23 years of screening in Italy, this elevated rate had fallen by 62.6% (25.6% to 80.0%), whereas the incidence of SCD in athletes had fallen on average by 2.62 (4.87 to 0.37) per 100 000 person-years, giving an NNS of 38 151 (20 534 to 267 380). The contribution of competitive sports to the rate of SCD in the total population aged 12–35 years fell from 26.6% (−20.3% to 55.8%) to 7.2% (−10.7% to 22.4%). Some of this difference may be explained by confounding, as the rate of SCD in non-athletes over the screening period increased slightly.
Summary of epidemiological measures used to assess the impact of competitive sport and preparticipation screening in athletes and the total population (athletes and non-athletes) in the prescreening and postscreening period
Table 3 shows our estimates of the preparticipation screening test's accuracy plus other derived parameters used to assess the implication of screening in the UK. The screening test (as used in Italy) was moderately sensitive (88.7%; 95% CI 86.6% to 90.5%), detecting nearly nine of 10 cases with CVD. The specificity was also good at 92.6% (95% CI 92.3% to 92.9%), misclassifying only one in 14 athletes without disease. The positive predictive value was low at 22.5% (95% CI 21.2% to 23.8%), with only one in five positive tests correctly identifying someone with CVD. The negative predictive value was relatively high at 99.7% (95% CI 99.6% to 99.8%), only mis-categorising one negative test result in 342. The estimated prevalence of CVD in young athletes was 2.3% (2.2% to 2.5%). The estimated rate of SCD in athletes' after screening if they were not to participate in sport was 1.02 per 100 000 person-years. This was higher than the rate seen in non-athletes (0.8 per 100 000 person-years) because of the greater proportion of males with a higher risk of SCD.1
Parameters used to model the implications of screening young athletes in the UK, derived from the Italian programme
Figure 1 shows a hypothetical screening pathway based on an estimated UK population of 18 695 074 aged 12–35 years in 2007. Without screening, 196 (95% CI 144 to 264) SCDs per year would be expected in the UK, 64 in athletes. The introduction of screening might reduce this by 40 (95% CI 6 to 74) deaths per year on average (over a 23-year period). As the study of Corrado et al reported no deaths in diagnostically positive cases, these could be assumed to occur in negative diagnostic tests and false-negative screening tests. As 15 SCDs in athletes would be expected if all screened athletes were to desist from competitive sport, we estimate that nine SCDs, due to competitive sport, would not be prevented by screening. These are most likely to be in athletes with false-negative tests, as criterion standard diagnostic testing would have ruled out most cardiac risks. We also estimated one SCD in false positives and none in true positives by multiplying the incidence of SCD in athletes assuming non-participation in competitive sport by the number of tests taken in these groups.
Screening pathway for young people aged 12–5 years in the UK.
Figure 2 shows how the harm-to-benefit ratio of screening varies with prevalence. The ratio of screening harm (disqualifications) to SCDs prevented increases linearly with the prevalence of CVD in the athletic population. However, the proportion of false- to true-positive screening tests declines with CVD prevalence non-linearly. This suggests that although the major harm-to-benefit ratio decreases with prevalence, the relative proportion of minor harms caused by false-positive tests increases disproportionately, especially at prevalences below 1%.
Harm-to-benefit ratio of screening and number of false positives per true positive by prevalence of CVD in athletes (sufficient for disqualification).
Figure 3 shows how the number of SCDs in athletes after screening and the number of SCDs due to sport not prevented by screening fall as prevalence reduces. It also shows that these two measures of SCD are affected differentially by test uptake rates. At an uptake of 60% to 80%, the number of overall SCDs prevented becomes less than those not prevented.
Reduction in SCDs in athletes after screening and SCDs due to sport not prevented by screening by proportion of athletes in the population and test uptake rate.
Discussion
The Italian findings show that screening reduced the rate of SCD in athletes by 2.62 per 100 000 and though contributing to a 62.6% fall in incidence, the rate of SCD in athletes still remained elevated at twice that of non-athletes. In terms of the impact on population health, competitive sport accounted for 26.6% of all SCDs before screening, falling to 7.2% after screening. The fall in incidence of SCD in the population after screening was not statistically significant. To prevent one SCD, we estimate that 38 151 athletes would need to be screened.
Assessing the impact in the UK, screening would prevent up to 40 of an estimated 196 SCDs annually if the prevalence of CVD and rate of participation in competitive athletics were similar to Italy and screening uptake was 100%. Twenty-four SCDs in athletes would not be prevented, although only nine of these could be attributed to competitive sport. Nearly all of these SCDs would probably occur in athletes falsely reassured by negative screening test results. To achieve these benefits, over 1.5 million athletes would need to be screened, with 140 000 being referred for diagnostic testing, leading to 31 522 athletes being disqualified. Thus, in preventing one SCD, nearly 800 athletes would be excluded from competition. This ratio of harm (true-positive tests) to benefit was sensitive to the prevalence, reducing in populations where CVD is less prevalent. However, the total number of minor harms (false-positive tests) remained over 100 000. It was not sensitive to the proportion of competitive athletes or the screening uptake rate. The efficiency of screening was influenced by the uptake rate, with the number of SCDs not prevented exceeding those prevented at around 70%.
Strengths and weaknesses of study
This is the first study to quantify the impact of establishing a screening programme for competitive athletes aged 12–35 years in the UK. We used data from large, long-running, cohort studies in Italy, widely considered to be the most robust data available, to calculate our outcomes. Outcomes were estimated over the first 23 years of screening to provide a more realistic picture of the potential benefits and harms, as this included the period when materials, qualifications, facilities and administrative mechanisms were being developed and the full potential of screening is unlikely to have been realised.
We assessed how much competitive athletics contribute to SCD in athletes and the wider population and how much this is reduced by screening and the NNS. To our knowledge, these epidemiological measures are not reported elsewhere. We explored the sensitivity of the results to parameters in the model where uncertainty was high.
Our outcomes were based on secondary data, with some calculations relying on imputed estimates of exposure that were inaccurate by 1% to 1.5%. However, we estimate that this would have only influenced the values of AR, ARF, PAR and PARF by 1% to 2%.
The data came from two different cohorts. Therefore, our estimates of the effectiveness of screening and the prevalence of CVD, sensitivity and specificity of the screening test may have been biased. For example, the true NNS may be larger than our estimate, as the original study did not adjust for changes in athlete diet, training regimens and consumption of performance-enhancing drugs over time, which may have also contributed to the reduction in SCD rates.
To assess the performance of the screening test and the prevalence of CVD, we used data from a study of elite athletes. We could not determine whether this cohort differed in age and sex from regular athletes, as this data was only provided for disqualified elite athletes (who were slightly older with a lesser proportion of males).1 14 Nevertheless, our estimate of false negatives was conservative, as it did not include the 41 athletes with physiological left ventricular hypertrophy. Therefore, we may have slightly over-estimated the sensitivity and underestimated the specificity and prevalence of CVD. This would concord more with the results of a USA study of high-school athletes screened using ECG.17
Policy implications
The validity of our calculations rests on the circumstances, programme and outcomes in Italy being replicated in the UK. This cannot be assumed. Studies from Japan and the USA show significant variation in postscreening incidence of SCD, from 50% higher to nearly 50% lower than in Italy.18 Such differences may be explained not only by different profiles of underlying causes of SCD but also the effectiveness of a country's screening strategy.
Evidence also suggests that the prevalence of CVD in athletes varies by country. In the UK, the study of Wilson et al of junior athletes and physically active school children found much lower prevalences than in Italy (0.46% and 0.24%, respectively19), similar to that reported in the USA (0.4%).17 Although this would significantly improve the ratio of disqualifications per SCD prevented (169:1), assuming the effectiveness of screening remained the same, it would make little impact on the 100 000 or so false positives. These minor harms have not been studied in young, screened athletes,5 but research on breast and prostate cancer screening show that false-positive screening tests elevate anxiety.20 21
What is already known on this topic
▶ Athletes involved in competitive sport are at increased risk of SCD.
▶ After 23 years of screening young Italian athletes involved in competitive sport, their incidence of SCD fell to that of non-athletes.
What this study adds
▶ Estimates how much participation in competitive sport contributes to SCD in athletes and the total population.
▶ Calculates the NNS to prevent one SCD.
▶ Estimates the sensitivity and the specificity of the screening test used in Italy.
▶ Estimates the harm-to-benefit ratio of screening.
The proportion of 12–35-year-olds engaged in competitive athletics in the UK is also not clear from surveys.22 23 Taking these differences into account, we believe screening would, in practice, prevent significantly less SCDs than we have estimated, especially if it was not mandatory.
Screening seems to be less effective than other UK screening programmes. For example, the NNS for colon and breast cancers (not including 40–49-year-olds) over 5 years is 1374 (95% CI 955 to 2802) to 2451 (95% CI 1576 to 7651).24 To prevent one SCD, 7630 (38 151/5) athletes would need to be screened annually over 5 years. Cost-effectiveness studies of screening give estimates from €14 220 to $44 000 per year of life saved in Italy25 and the USA (using ECG), respectively.26 However, these analyses do not cost the major or minor harms of screening, biasing their estimates downwards.
Establishing a UK screening programme would require significant investment in organisational systems and human capacity if it were to deliver an equitable and acceptable, high-quality service. However, the (cost)-effectiveness is uncertain, with existing evidence suggesting that the small impact on population health would come at a high price for many athletes. Advocating a national screening programme at this time would, therefore, be premature.
Conclusions
The impact of screening on reducing SCD in athletes would be moderate and on population health would be small and would be achieved at a significant cost to a large number of athletes. Given our findings and the uncertainties of generalising the Italian evidence to the UK, support for a national programme cannot be justified.
Acknowledgments
The authors thank Rod Taylor and other colleagues at PenTAG for their comments on the earlier draft.
References
Footnotes
-
Competing interests None.
-
Provenance Not commissioned; externally peer reviewed.