2. What is the diagnostic value (i.e. sensitivity and/or specificity) of biomarker ANP, BNP, NT-pro-BNP, troponin-T, and troponin-I to detect asymptomatic cardiac systolic dysfunction as measured by echocardiography in childhood and adult cancer survivors?
First Author Year	Study Design Treatment era Years of follow-up	Participants	Diagnostic tests	Main outcomes~	Addt’l remarks
Krawczuk-Rybak41 2011	Single-center cohort study (Poland). Treatment era: Nm. Years of follow-up after treatment completion: mean 5.91 years (range 1.6 to 13.8).	44 childhood cancer survivors treated with anthracyclines (doxorubicin, daunorubicin) for ALL (n=37) or Hodgkin lymphoma (n=7). 30 males/ 14 females; mean age at diagnosis nm; mean age at study 14.7 years (range 6 to 23). Treatment: Cumulative anthracycline dose for ALL 180 to 540 mg/m2; for Hodgkin lymphoma 120 to 240 mg/m2; patients with Hodgkin lymphoma received 15 Gy of radiotherapy to the upper mediastinum (no information on number of fractions).	Doppler and colour flow visualization echocardiography; M-mode for heart structures and Teicholz method for contractility and LVEF (number of observers nm); an abnormal test result was defined as indexed stroke volume < 40 ml/m2 (n=16; prevalence 36.4%). NT-pro-BNP; an abnormal test result was defined as > 115 ng/ml (n=6; prevalence 13.6%). Time between tests: nm.	When the echocardiographic result is used as the reference standard^: Sensitivity: 12.5% (95% CI 2.3 to 27.9) Specificity: 85.7% (95% CI 79.9 to 94.5) Positive predictive value: 33.3% (95% CI 6.1 to 74.4) Negative predictive value: 63.2% (95% CI 58.9 to 69.6) Agreement between tests (i.e. either both abnormal or both normal): 26/44 (59.1%).	Patients had no history of heart disease and no signs of cardiac failure. The risk of selection bias is unclear: not stated if all eligible patients or a random sample thereof were included. The risk of detection bias is unclear; nm if outcome assessors were blinded. Low risk of outcome/attrition bias: all 44 patients had both tests.
Brouwer22 2011	Single-center cross-sectional study (the Netherlands). Treatment era: between 1976 and 1999; current tests between August 2004 and April 2007. Years of follow-up post-treatment: median 18.2 years (range 5.4 to 30.8).	277 childhood cancer survivors ≥ 18 years treated with potential cardiotoxic therapy (i.e. anthracyclines, platinum analogues or radiotherapy on mediastinum (including mantle field, spine or total body) for leukemia (n=113), malignant lymphoma (n=56), sarcoma (n=48), brain tumor (n=32), nephro/neuroblastoma (n=23) or germ cell tumor (n=5) and surviving at least 5 years after diagnosis. 155 males/122 females; median age at diagnosis 8.8 years (range 0 to 20.1); median age at cardiac evaluation 27.5 years (range 18.1 to 48.2). Treatment: Median cumulative anthracycline dose (doxorubicin, daunorubicin) 183 mg/m2 (range 50–600); median dose of mediastinal radiotherapy 25 Gy (no information on number of fractions); no further information on treatment doses provided; all patients received anthracyclines, platinum analogues or radiotherapy as described above.	2D echocardiography, colour flow mapping 2D guided M-mode blood pool and tissue velocity imaging (performed by a single skilled technician masked to treatment versus control group to exclude interobserver variability); an abnormal test result was defined as LVSF < 29% (n=97; prevalence 37%) or WMSI > 1.00 (n=38; prevalence 14.5%). NT-pro-BNP; an abnormal test result was defined as > 125 ng/ml (n=32; prevalence 12.2%). Time between tests: nm.	When the echocardiographic result of the LVSF is used as the reference standard^: Sensitivity: 16.5% (95% CI 10.9 to 22.1) Specificity: 90.3% (95% CI 87.0 to 93.6) Positive predictive value: 50% (95% CI 33.1 to 66.8) Negative predictive value: 64.8% (95% CI 62.4 to 67.1) Agreement between tests (i.e. either both abnormal or both normal): 165/262 (63.0%). When the echocardiographic result of the WMSI is used as the reference standard^: Sensitivity: 31.6% (95% CI 19.2 to 45.1) Specificity: 91.1% (95% CI 89.0 to 93.4) Positive predictive value: 37.5% (95% CI 22.7 to 53.6) Negative predictive value: 88.7% (95% CI 86.6 to 90.9) Agreement between tests (i.e. either both abnormal or both normal): 216/262 (82.4%).	Patients with current treatment for a relapse or secondary malignant disease or with mental incapacity were excluded. At time of study 263 out of 274 patients had NYHA class I and 11 out of 274 NYHA class II; for 3 patients no data mentioned. 17 out of 275 patients used cardioactive medications (ACE-inhibitor, β-blocker or diuretic); for 2 patients this was unknown; nm if all patients receiving medication did for cardiac causes. Selection bias cannot be ruled out (277 out of 401 eligible patients (69%) participated in this study). The risk of detection bias is low; the echocardiographic outcome assessor was blinded. Outcome/attrition bias cannot be ruled out (only for 262 out of 277 patients (95%) both test were available). The authors stated that the high prevalence of abnormal LVSF in apparently healthy sibling controls suggests (22%) the possibility of false-positive findings and challenges the appropriateness of LVSF as a reliable marker of systolic function in adults.
Mavinkurve-Groothuis42 2009	Single-center cohort study (the Netherlands). Treatment era: Nm (current study executed between May 2006 and October 2007). Median years of follow-up: 13.8 years (range 5 to 28.7).	122 long-term survivors of childhood cancer treated with anthracyclines for ALL (n=38), AML (n=8), ependymoma (n=1), Ewing sarcoma (n=6), hepatoblastoma (n=3), Hodgkin lymphoma (n=13), neuroblastoma (n=6), NHL (n=30), osteosarcoma (n=3), rhabdomyosarcoma (n=4) or Wilms tumor (n=10). 62 males/60 females; median age at diagnosis 5.7 years (range 0.03 to 14.4); median age at study 21 years (range 5 to 39.4 years). Treatment: Median cumulative anthracycline dose (doxorubicin and/or daunorubicin) 180 mg/m2 (range 50–542); 7 patients also received mediastinal irradiation (no further information provided).	Transthoracic M-mode echocardiography (performed by experienced echocardiographic technicians and supervised by 2 (pediatric) cardiologists who were unaware of the cumulative chemotherapy dose and levels of NT-pro-BNP); an abnormal test result was defined as LVEF < 55% (n=9; prevalence 7.4%). NT-pro-BNP; an abnormal test result was defined as males <10 pmol/L, females <18 pmol/L and for children age dependent reference values by Albers et al (n=16; prevalence 13.1%). Both tests were performed at the same time.	When the echo result is used as the reference standard^: Sensitivity: 22.2% (95% CI 4.0 to 57.0) Specificity: 87.6% (95% CI 86.2 to 90.4) Positive predictive value: 12.5% (95% CI 2.3 to 32.1) Negative predictive value: 93.4% (95% CI 91.8 to 96.3) Agreement between tests (i.e. either both abnormal or both normal): 101/122 (82.8%).	At time of testing none of the patients had symptomatic cardiac disease (defined as < NYHA class II) or a history of cardiovascular disease or chronic renal insufficiency. The risk of selection bias is unclear: all consecutive patients who visited the Late Effects Clinic during the study period were included, but it is not stated if those patients represented a random sample of the complete cohort of survivors. The risk of detection bias is low; echocardiographic outcome assessors were blinded. Low risk of outcome/attrition bias: all 122 patients had both tests.
Mavinkurve-Groothuis42 2009	Single-center cohort study (the Netherlands). Treatment era: nm (current study executed between May 2006 and October 2007). Median years of follow-up: 13.8 years (range 5 to 28.7).	122 long-term survivors of childhood cancer treated with anthracyclines for ALL (n=38), AML (n=8), ependymoma (n=1), Ewing sarcoma (n=6), hepatoblastoma (n=3), Hodgkin lymphoma (n=13), neuroblastoma (n=6), NHL (n=30), osteosarcoma (n=3), rhabdomyosarcoma (n=4) or Wilms tumor (n=10). 62 males/60 females; median age at diagnosis 5.7 years (range 0.03 to 14.4); median age at study 21 years (range 5 to 39.4 years). Treatment: Median cumulative anthracycline dose (doxorubicin and/or daunorubicin) 180 mg/m2 (range 50–542); 7 patients also received mediastinal irradiation (no further information provided).	Transthoracic M-mode echocardiography (performed by experienced echocardiographic technicians and supervised by 2 (pediatric) cardiologists who were unaware of the cumulative chemotherapy dose and levels of cardiac troponin T); an abnormal test result was defined as LVEF < 55% (n=9; prevalence 7.4%) or as LVSF < 29% (n=4; prevalence 3.3%). Cardiac troponin T; an abnormal test result was defined as ≥ 0.010 ng/ml (n=0%; prevalence 0%) Both tests were performed at the same time.	When the echocardiographic result of the LVEF is used as the reference standard^: Sensitivity: 0% (95% CI 0 to 0) Specificity: 100% (95% CI 100 to 100) Positive predictive value: NaN Negative predictive value: 92.6% (95% CI 92.6 to 92.6) Agreement between tests (i.e. either both abnormal or both normal): 113/122 (92.6%). When the echocardiographic result of the LVSF is used as the reference standard^: Sensitivity: 0% (95% CI 0 to 0) Specificity: 100% (95% CI 100 to 100) Positive predictive value: NaN Negative predictive value: 96.7% (95% CI 96.7 to 96.7) Agreement between tests (i.e. either both abnormal or both normal): 118/122 (96.7%).	At time of testing none of the patients had symptomatic cardiac disease (defined as < NYHA class II) or a history of cardiovascular disease or chronic renal insufficiency. The risk of selection bias is unclear: all consecutive patients who visited the Late Effects Clinic during the study period were included, but it is not stated if those patients represented a random sample of the complete cohort of survivors. The risk of detection bias is low; echocardiographic outcome assessors were blinded. Low risk of outcome/attrition bias: all 122 patients had both tests.
Sherief44 2012	Single-center cohort study (Egypt). Treatment era: nm. Mean years of follow-up: not completely clear from manuscript, but most likely 3.75 years (range 1.5 to 6).	50 survivors of childhood acute leukemia (n=39 ALL; n=11 AML) treated with anthracyclines. 30 males/20 females; mean age at diagnosis 8.4 years (range 3 to 15); mean age at evaluation 11.63 years (range 8 to 16). Treatment: n=18 cumulative anthracycline dose <150–300 mg/m2; n=32 cumulative anthracycline dose > 300 mg/m2 (but elsewhere in the manuscript n=19 < 300mg/m2 and n=31 > 300 mg/m2 was mentioned).	Conventional echocardiography (no further information provided; number of observers nm); an abnormal test result was defined as LVEF < 55% or a LVSF < 29% (n=8 subclinical cardiotoxicity in the form of increase of left ventricular dimension and EF; prevalence 16%). Cardiac troponin T; an abnormal test result was defined as > 0.010 ng/ml (n=0; prevalence 0%). Time between tests: Nm.	When the echocardiographic result is used as the reference standard^: Sensitivity: 0% (95% CI 0 to 0) Specificity: 100% (95% CI 100 to 100) Positive predictive value: NaN Negative predictive value: 84% (95% CI 84 to 84) Agreement between tests (i.e. either both abnormal or both normal): 42/50 (84%).	At time of testing all survivors were asymptomatic (i.e. no signs and symptoms of cardiac impairment); patients with renal or hepatic impairment were excluded as were patients with a history of cardiac disease and hypertension. The risk of selection bias is unclear; not clear if these 50 patients were all eligible patients or a random sample thereof. The risk of detection bias is unclear; nm if outcome assessors were blinded. Low risk of outcome/attrition bias: all 50 patients had both tests.
Kismet45 2004	Multi-center cohort study (Turkey). Treatment era: June 1982 to August 2000. Median time from last doxorubicin dose: 12 months (range 1 to 168).	24 childhood cancer patients who received doxorubicin for treatment of Hodgkin disease (n=4), rhabdomyosarcoma (n=4), Ewing sarcoma (n=3), osteosarcoma (n=3), malignant mesenchymal tumor (n=3). Wilms tumor (n=2), neuroblastoma (n=1), hepatoblastoma (n=1), clear cell sarcoma (n=1), malignant mesothelioma (n=1) and primitive neuroectodermal tumor (n=1). 14 males/10 females; median age at diagnosis nm; median age at study 14 years (range 3–31). Treatment: Median cumulative doxorubicin dose 480 mg/m2 (range 400 to 840); 4 patients also received mediastinal irradiation (no further information provided).	Two-dimensional, M-mode and Doppler echocardiography performed by pediatric cardiologists (number of observers nm); an abnormal test result was defined as LVEF < 55% and LVSF < 29% (n=2; prevalence 8.3%). Cardiac troponin T; an abnormal test result was defined as ≥ 0.010 ng/ml (n=3; prevalence 12.5%). Time between tests: within 24 hours.	When the echocardiographic result is used as the reference standard^: Sensitivity: 50% (95% CI 2.7 to 97.2) Specificity: 90.9% (95% CI 86.6 to 95.2) Positive predictive value: 33.3% (95% CI 1.8 to 64.8) Negative predictive value: 95.2% (95% CI 90.7 to 99.7) Agreement between tests (i.e. either both abnormal or both normal): 21/24 (87.5%).	None of the patients had clinical evidence of abnormal cardiac functions; patients with evidence of renal disease were excluded from the study. The risk of selection bias is unclear; not clear if these 24 patients were all eligible patients or a random sample thereof. The risk of detection bias is unclear; nm if outcome assessors were blinded. Low risk of outcome/attrition bias: all 24 patients had both tests.
Soker46 2005	Single-center study (Turkey). Treatment era: October 2000 and December 2004. Mean follow-up after the last anthracycline dose 9.39 months (range 1 to 42).	31 childhood cancer patients who received doxorubicin for treatment of ALL (n=27), AML (n=2), Hodgkin disease (n=1), NHL (n=1). 14 males/17 females; median age at diagnosis nm; median age at study 8.16 years (range 4 to 15). Treatment: Median cumulative doxorubicin dose 240 mg/m2 (range 30–600).	Two-dimensional, pulse-wave Doppler and M-mode echocardiography (performed by 1 experienced pediatric cardiologist); an abnormal test result was defined as LVEF < 60% and LVSF < 30% (n=4; prevalence 12.9%). Cardiac troponin I; an abnormal test result was defined as ≥ 0.50 ng/ml (n=0; prevalence 0%). Time between tests: performed simultaneously.	When the echocardiographic result is used as the reference standard^: Sensitivity: 0% (95% CI 0 to 0) Specificity: 100% (95% CI 100 to 100) Positive predictive value: NaN Negative predictive value: 87.1% (95% CI 87.1 to 87.1) Agreement between tests (i.e. either both abnormal or both normal): 27/31 (87.1%).	Two of the 4 patients with systolic dysfunction had clinical findings; patients who received mediastinal irradiation or had other illnesses such as infections were excluded. The risk of selection bias is unclear; not clear if these 31 patients were all eligible patients or a random sample thereof. The risk of detection bias is unclear; nm if outcome assessors were blinded. Low risk of outcome/attrition bias: all 31 patients had both tests.

Nm: not mentioned; ALL: acute lymphoblastic leukaemia; n: number; LVEF: left ventricular ejection fraction; CI: confidence interval; LVSF: left ventricular shortening fraction; WMSI: wall motion score index; NYHA: New York Heart Association; AML: acute myeloid leukaemia; NHL: non-Hodgkin lymphoma; NaN: not a number (data type)

^{^}Since echocardiography is most often used to assess cardiac function in clinical practice, we have chosen the echocardiographic results as reference standard

^~Calculated by the guideline developers based on information provided in the article (for the main outcomes we used the calculator on http://statpages.org/ctab2x2.html)

^*It was unclear if both or only one of the two markers should have been abnormal for this definition