OUP user menu

Prognostic value of myocardial 123I-metaiodobenzylguanidine (MIBG) parameters in patients with heart failure: a systematic review

Hein J. Verberne, Lizzy M. Brewster, G. Aernout Somsen, Berthe L.F. van Eck-Smit
DOI: http://dx.doi.org/10.1093/eurheartj/ehn113 1147-1159 First published online: 17 March 2008

Abstract

Aims To derive a more precise estimate of the prognostic significance of myocardial 123I-metaiodobenzylguanidine (MIBG) parameters [early heart mediastinal ratio (H/M), late H/M, and myocardial washout] in heart failure (HF).

Methods and results Eighteen studies with a total of 1755 patients, stratifying survival, and cardiac events in patients with HF by MIBG, were eligible for analysis. The pooled hazard ratio (HR) estimates for cardiac death and cardiac events associated with washout showed no significant heterogeneity and were 1.72 [95%CI (confidence interval), 1.72–2.52; P = 0.006] and 1.08 (95%CI: 1.03–1.12; P < 0.001), respectively. The pooled HR estimates for cardiac death and cardiac events associated with early H/M and late H/M showed significant heterogeneity (I2 ≥ 75%). Limiting the pooling to the qualitative best three studies rendered I2 insignificant (I2 = 0) and resulted in a pooled HR of late H/M for cardiac death of 1.82 (95%CI: 0.80–4.12; P = 0.15) and for cardiac events of 1.98 (95%CI: 1.57–2.50; P < 0.001).

Conclusion Our results indicate that patients with HF and decreased late H/M or increased myocardial MIBG washout have a worse prognosis compared with those with normal semi-quantitative myocardial MIBG parameters.

Keywords
  • Heart failure
  • Meta-analysis
  • Prognosis
  • MIBG

Introduction

Cardiac sympathetic neuronal activity can non-invasively be assessed by the use of 123I-metaiodobenzylguanidine (123I-MIBG), an analogue of norepinephrine.1 In patients with chronic heart failure (HF), angiotensin-converting enzyme inhibitors,2 β-receptor antagonists,3,4 spironolactone5,6 and chronic cardiac resynchronization therapy7 have been shown to ameliorate functional capacity and prognosis. Using semi-quantitative analysis (i.e. early heart-to-mediastinum (H/M) ratio, late H/M and myocardial washout) these beneficial effects were associated with an increase in 123I-MIBG uptake and a reduced washout. There is no consensus, however, on the prognostic value of these semi-quantitative parameters of myocardial 123I-MIBG uptake in patients with HF. The lack of consensus is reflected in the absence of 123I-MIBG in any of the current guidelines regarding either HF or myocardial scintigraphic imaging.812

The purpose of this systematic review was to critically assess existing evidence on the prognostic value of semi-quantitative parameters of myocardial 123I-MIBG uptake in patients with HF.

Methods

This section describes the essentials of the materials and methods. See the Appendix for a more detailed description of materials and methods.

Eligibility criteria

Published studies were eligible if survival was analysed in patients with HF stratified by semi-quantitative 123I-MIBG myocardial parameters (i.e. early H/M, late H/M, and myocardial washout). The primary outcomes of interest were cardiac death and cardiac events (i.e. combination of cardiac death, myocardial infarction, myocardial transplantation, and hospitalization due to progression of HF).

Search strategy

A computer-assisted search was performed of the medical databases MEDLINE (January 1980 to January 2006), PubMed (January 1980 to January 2006), EMBASE (January 1980 to January 2006), the Cochrane Controlled Trial Register and the Cochrane Database of Systematic Reviews (from their inception to January 2006). A highly sensitive search strategy described by Haynes et al.13 was adapted to our specific requirements.

Selection procedure

All studies matching the eligibility criteria were retrieved. In case of overlapping and duplicated data sets, care was taken to include only the most recent or most complete data set.

Methods of the review

One author (H.J.V.) assessed articles for possible inclusion in the review by checking titles and abstracts. The subsequent selected articles were independently assessed by two authors (H.J.V. and B.L.F.v.E.S.). A standard data abstract form was developed for systematic collection of data on key characteristics (cardiac events), methodological quality, participants, follow-up, and parameters related to image acquisition and calculation of semi-quantitative 123I-MIBG myocardial parameters (i.e. early H/M, late H/M, and myocardial washout). Disagreement in acquired data was resolved through final discussion. Investigators were contacted twice to obtain missing information.

Methodological quality assessment

As there are no widely agreed quality criteria for assessing prognostic studies, criteria were formulated based on suggestions made by Altman.14 Internal validity criteria were assigned a score for presence (adequate methods; score = 1) or absence (inadequate methods, potential for bias; score = 0). Quality scores were expressed as a percentage of the maximum score (9). No criteria for external validity (evaluation of generalizability) were formulated.

Statistical analysis

The association between 123I-MIBG parameters of myocardial sympathetic activity and cardiac death, and the association between 123I-MIBG parameters of myocardial sympathetic activity and cardiac events were derived as a weighted average of study-specific estimates of the hazard ratio (HR), using inverse variance weights.15,16

Summary data from published studies were pooled using the random-effects models as a primary analysis.17 The percentage variability of the pooled HR attributable to heterogeneity among studies was quantified using the I2 statistic.18 If heterogeneity was present sources of heterogeneity were explored and a decision was made whether to aggregate the studies or not. Sensitivity analysis was performed by reanalysing the data using the fixed-effects model. Evidence of publication bias was examined by constructing funnel plots.19 Statistical analyses were performed using Review Manager (RevMan) [Computer program], version 4.2 for Windows, Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2003.

Results

Literature search

Full reports or abstracts from 107 references of papers yielded 18 studies that fulfilled the inclusion criteria of our systematic review.2037 Figure 1 shows the progress through the selection of studies eligible for the systematic review.

Figure 1

Selection of studies eligible for systematic review. Eighty nine studies were excluded: 13 reviews and case reports, 14 studies on neuroblastomas and pheochromocytomas, 19 studies on the effect of medication, 2 studies on cardiomyopathy in children and a group of 39 studies with miscellaneous subjects ranging from hypertension, diabetes mellitus, internal conversion devices to myocardial infarction. Furthermore, two studies with overlapping and/or duplicated data sets were excluded. Non-electronic search (contact with authors and hand searching) did not result in additional (unpublished) studies that fulfilled the eligibility criteria.

Methodological quality of included studies

The quality scores ranged from 4 to 9 (median 6) (Table 1). Only two studies achieved the maximum score of 9 points and one study lost 1 point because of the lack of blind assessment of outcome, i.e. three studies were in the upper fifth of the scoring scale (ranging from 0 to 9).2022 The majority of studies achieved a suboptimal scoring due to the lack of blind assessment and interpretation of the semi-quantitative 123I-MIBG myocardial parameters and due to the lack of blind assessment of outcome. Nine studies had a quality score in the 2nd fifth of the scoring scale2331 and the remaining six studies had scores in the 3rd fifth of the scoring scale.3237

View this table:
Table 1

Quality assessment of included studies: quality criteria for validity were scored on presence (1) or absence (0) and comprised of items related to patient sample (fully defined inclusion criteria; fully described clinical and demographic characteristics), study design (only a prospective design was defined as adequate) definition and assessment of outcome (fully defined outcome; objective and unbiased assessment of outcome), duration of follow-up (follow-up of at least 3 months), semi-quantitative 123I-metaiodobenzylguanidine (123I-MIBG) myocardial parameters (fully defined and described; blind assessment) and avoidance of verification bias (assessment of outcome blind to 123I-MIBG result)

Study (Reference)Methodological quality criteriaTotal scorePercentage of maximum score
Patient sampleStudy designOutcome measureFollow-upMIBG parametersVerification bias
Inclusion criteria definedFully describedProspectiveDefinedBlind assessment≥3 monthsDefinedBlind assessmentAbsent
Cohen-Solal et al.201111111119100
Yamada et al.211111111119100
Ogita et al.2211111111889
Anastasiou-Nana et al.231111111778
Merlet et al.241111111778
Merlet et al.251111111778
Nakata et al.261111111778
Nakata et al.271111111778
Wakabayashi et al.281111111778
de Milliano et al.29111111667
Gerson et al.3011111667
Kyuma et al.31111111667
Arimoto et al.3211111556
Imamura et al.3311111556
Fujimoto et al.341111444
Fujimoto et al.351111444
Matsui et al.361111444
Momose et al.371111444

Study characteristics

Baseline characteristics of all 18 eligible studies are summarized in Table 2. The majority of patients were male (1261/1643: 77%, Merlet et al.25 did not specify the gender of 112 patients) and of middle age (>40 years). The NYHA classification differed between the contributing studies. Some studies included the whole spectrum of HF from patients with only mild symptoms (class I) to patients with severe symptoms (class IV). Other studies were more focused on a smaller patient cohort.29 Left ventricular function, expressed as left ventricular ejection fraction (LVEF), was in the majority (14/18) of the contributing studies well defined and in accordance with a depressed left ventricular function (LVEF < 40%).2030,33,35,36 Two studies included patients with LVEF > 50%31,34 and two studies did not make any reference to left ventricular function.32,37 As with NYHA classification, etiology of HF differed between studies. Eleven studies reported on combined ischaemic and idiopathic HF,2024,26,27,29,3133 6 reported on idiopathic HF alone.25,30,3437 In one study ischaemic and idiopathic HF patients were separately analysed and compared.28 Of this study no overall means were provided. Therefore, in our analysis ischaemic (A) and non-ischaemic (B) were separately reviewed. None of the contributing studies, however, reported on ischaemic HF alone.

View this table:
Table 2

Study characteristics [order of presentation according to decreasing study quality (Table 1)]

Study (Reference)Patients
DescriptivesHeart failureEventsMIBG parameters
NM/FAge (mean ± SD)LVEF (%)NYHAEtiology (IS/ID)DeathMICardiac TxHospFollow-up (months)Early H/MLate H/MWashout
Cohen-Solal et al.209388/555 ± 1025 ± 10>I24/69102310 ± 81.39 ± 0.211.31 ± 0.2035 ± 6
Yamada et al.216551/1463 ± 1228 ± 8I–III41/2431034 ± 19nsansansa
Ogita et al.227964/1564 ± ns<40ns45/3413111–52nsbnsbnsb
Anastasiou-Nana et al.235246/656 ± 1231 ± 12ns27/2514241.47 ± 0.151.35 ± 0.16ns
Merlet et al.249076/1452 ± 722 ± 8>I24/6622101–27ns1.22 ± 0.15ns
Merlet et al.25112ns50 ± 1021 ± 9>I0/112251927 ± 20ns1.23 ± 0.19ns
Nakata et al.26205146/5961 ± 1434 ± 6I–IV81/1243835 ± 151.98 ± 0.451.80 ± 0.3735 ± 15
Nakata et al.27167130/3719–85<40I–IV66/1014243nsansansa
Wakabayashi et al.28 (A)c7662/1468 ± 1232 ± 11I–IV76/02854 ± 311.95 ± 0.411.79 ± 0.3834 ± 13
Wakabayashi et al.28 (B)c5643/1360 ± 1330 ± 11I-IV0/561955 ± 291.91 ± 0.431.71 ± 0.4142 ± 13
de Milliano et al.295839/1965 ± 1027 ± 7II–III31/271611–58ns1.8 ± 0.4ns
Gerson et al.303727/1048 ± 927 ± 10>I0/378449 ± 91.54 ± 0.2nsns
Kyuma et al.31158110/4864 ± 1341 ± 17I–IV45/1131716ns1.73 ± 0.3736 ± 14
Arimoto et al.327647/2963 ± 15nsI–III12/64146–301.79 ± 0.28d1.69 ± 0.30d25 ± 7d
Imamura et al.33171125/4663 ± 1127 ± 10I–IV75/96111627 ± 111.78 ± 0.271.64 ± 0.2846 ± 12
Fujimoto et al.347457/1757 ± 1244 ± 16I–III0/7441225 ± 151.82 ± 0.281.70 ± 0.3142 ± 11
Fujimoto et al.355343/1057 ± 1134 ± 10I–III0/532744 ± 33nsns41 ± 11
Matsui et al.367455/1955 ± 131 ± 1>I0/74121124 ± 2ns1.89 ± 0.0335 ± 1
Momose et al.375952/752 ± 15nsns0/591625 ± 13nsansansa
  • ns, not specified in publication; –, not reported in publication; IS, ischaemic; ID, idiopathic.

  • aData were organized according patients with and patients without events. Overall means were not provided.

  • bPatients were divided into two groups according washout (cutoff 27%). Overall means were not provided.

  • cData were organized according to ischaemic (A) and non-ischaemic (B) cardiomyopathy. Overall means were not provided. For our analysis ischaemic (A) and non-ischaemic (B) were separately reviewed.

  • dAnalysis was performed in 64 patients in whom follow-up was completed. Baseline characteristics of these patients were not provided.

The combination of all three semi-quantitative parameters (early H/M, late H/M, and myocardial washout) was reported in six studies.20,23,26,33,34,37 Three studies reported only late H/M24,25,29 and two studies reported only early H/M30 or myocardial washout,35 respectively. In four studies no specific semi-quantitative baseline parameter was reported.21,22,27,37 One study reported a combination of early H/M and late H/M without washout23 and two a combination of late H/M and myocardial washout without early H/M.31,36

Majority of the studies (17/18) reported on cardiac death and of these six focused on cardiac death alone.23,2628,31,37 On the other hand only one study restricted the analysis to hospitalization due to progression of HF.32

Aspects of 123I-metaiodobenzylguanidine scintigraphy

Acquisition parameters and calculation of semi-quantitative parameters are shown in Table 3. The acquisition of early H/M was in nine studies2022,30,3337 performed at 15–20 min and ranged in the remaining six studies from 30 to 60 min.23,2628,32,33 The time of late H/M acquisition was more uniform; 17 studies acquired late H/M at 240 min after injection. Type of collimator used for acquisition was not reported in six studies. Low energy collimators were most widely used and only one study used a medium energy collimator.29

View this table:
Table 3

Acquisition and calculation of semi-quantitative parameters [order of presentation according to decreasing study quality (Table 1)]

Study (Reference)Acquisition parameters
Time after injection (planar and or SPECT)Calculation of MIBG parameters
Early H/MLate H/MCollimatorEarly H/MLate H/MWashout
Cohen-Solal et al.2020240nsH/MH/M(Early H− Late H)/Early Ha
Yamada et al.2120200LEHRHM/MHM/M{[Early (HM)/Early M] − [Late (HM)/Late M]}/[Early (HM)/Early M]a
Ogita et al.2220200LEHRHM/MHM/M{[Early (HM)/Early M] − [Late (HM)/Late M]}/[Early (HM)/Early M]a
Anastasiou-Nana et al.2360240LEAPH/MH/M
Merlet et al.24240nsH/M
Merlet et al.25240nsH/M
Nakata et al.2630240LEAPH/MH/MPolar maps on pixel base (SPECT)b
Nakata et al.2730240LEAPH/MH/MPolar maps on pixel base (SPECT)b
Wakabayashi et al.2830240LEAPH/MH/MPolar maps on pixel base (SPECT)b
de Milliano et al.29240MEH/M
Gerson et al.3015240nsHM/MHM/M{[Early (HM)/Early M] − [Late (HM)/Late M]}/[Early (HM)/Early M]b
Kyuma et al.3130240LEAPH/MH/MPolar maps on pixel base (SPECT)b
Arimoto et al.3230240LEHRH/MH/M[(Early H − Early M) − (Late H − Late M)]/(Early H − Early M)b
Imamura et al.3315240nsH/MH/M[(Early H − Early M) − (Late H − Late M)]/(Early H − Early M)b
Fujimoto et al.3420240LEAPH/MH/MPolar maps on pixel base (SPECT)b
Fujimoto et al.3520240LEAPnsnsPolar maps on pixel base (SPECT)b
Matsui et al.3615180LEH/MH/M[(Early H − Early M) − (Late H − Late M)]/(Early H − Early M)b
Momose et al.3715240nsH/MH/M[(Early H − Early M) − (Late H − Late M)]/(Early H − Early M)a
  • ns, not specified in publication; –, not reported in publication; LEHR, low energy high resolution; LEAP, low energy all purpose; ME, medium energy; LE, low energy.

  • aValues corrected for decay.

  • bCorrection for decay not reported in publication.

Calculation of early and late H/M was more or less uniform. Three studies used a small variation and corrected myocardial count density for background by subtracting mediastinum count density.21,22,30 A large variation in calculation of myocardial washout was reported ranging from relatively simple formulas (with or without correction for decay) to polar map pixel based subtraction techniques.

Retrieval of study specific estimates of hazard ratio

All 19 survival and/or cardiac event estimates from 18 studies were eligible for pooling, 12 of these 19 provided study specific estimates of HR associated with 123I-MIBG semi-quantitative myocardial parameters and/or its associated 95%CI (confidence interval).20,21,23,2629,31,3335 In the remainder HR estimates were calculated from information presented.22,24,25,30,32,36,37

Relationship between 123I-metaiodobenzylguanidine and survival and/or cardiac events

Figure 2 shows a forest plot of study specific estimates of HR and late H/M (14 studies reported on death and two on cardiac events). The pooled HR for death and late H/M and the pooled HR for a cardiac event and late H/M showed evidence of heterogeneity (I2 = 87.6% and I2 = 88.1%, respectively). Exploring heterogeneity revealed that when only the three qualitative best studies (upper fifth, see Table 1) were selected, heterogeneity disappeared (I2 = 0). Figure 3 shows the pooled HR for death and late H/M in these selected studies [1.82 (95%CI: 0.80–4.12; I2 = 0%; test for overall effect: Z = 1.44, P = 0.15)] and the HR for a cardiac event and late H/M was 1.98 (95%CI: 1.57–2.50; I2 = 0%; test for overall effect: Z = 5.79, P < 0.001. No other common denominator to combine studies (i.e. patient characteristics, NYHA classification, etiology of HF, events, duration of follow-up or technical aspects) reduced heterogeneity to an acceptable level.

Figure 2

Forest plots of hazard ratios for late H/M for all eligible studies. Forest plots of hazard ratios for late H/M in relation to cardiac death (upper panel) and cardiac events (lower panel) for eligible studies. Random effects model: squares are mean weighted hazard ratios (logarithmic scale). The size of the squares represents study weight and horizontal lines represent 95% CI (confidence interval). Arrowheads depict data outside scale. Order of study presentation based on decreasing variance of the logHR (i.e. increasing weight in the pooled HR estimates).

Figure 3

Forest plots of hazard ratios for late H/M for qualitative best studies. Forest plots of hazard ratios for late H/M in relation to cardiac death (upper panel) and cardiac events (lower panel) for the qualitative best studies. Random effects model: black diamonds are pooled estimates.

Figure 4 shows a forest plot of study specific estimates of HR and early H/M (six studies reported on death and one on cardiac events). The pooled HR for death and early H/M and the pooled HR for a cardiac event and early H/M showed evidence of heterogeneity (I2 = 85.2% and I2 = 85.6%, respectively). Exploring the heterogeneity no combination of studies reduced I2 to an acceptable level. Again a selection of studies based on the qualitative best studies (upper fifth, Table 1) was applied to reduce heterogeneity. Of these, however, only the study from Cohen-Solal et al.20 reported on early H/M. In this study the HR for death and early H/M was 2.27 (95%CI: 0.66–7.82; test for effect: Z = 1.29, P = 0.20) and the HR for a cardiac event and early H/M was 1.59 (95%CI: 0.80–3.15; test for effect: Z = 1.33, P = 0.18).

Figure 4

Forest plots of hazard ratios for early H/M for all eligible studies. Forest plots of hazard ratios for early H/M in relation to cardiac death (upper panel) and cardiac events (lower panel) for all eligible studies. Random effects model.

Figure 5 shows a forest plot of study specific estimates of HR and myocardial washout (six studies reported on death and three on cardiac events). The pooled HR for death and washout was 1.72 (95%CI: 1.72–2.52), with an I2 = 62.7% and a significant overall effect (P = 0.006). The pooled HR for a cardiac event and washout was 1.08 (95%CI: 1.03–1.12), with an I2 = 68.0% and a significant overall effect (P < 0.001). Limiting the pooling only to the three qualitative best studies (upper fifth, see Table 1) improved heterogeneity. Of these, however, only the study from Cohen-Solal et al.20 reported on washout and death. In this study the HR for death and washout was 1.04 (95%CI: 0.30–3.61; test for effect: Z = 0.07, P = 0.94). Figure 6 shows the pooled HR for a cardiac event and washout in the two remaining best quality studies [1.08 (95%CI: 1.01–1.14; I2 = 26.7%; test for overall effect: Z = 2.31, P = 0.02)]. The reduction in heterogeneity, however, did not result in a significant change of the pooled HR. Furthermore, no other combination of studies showed a reduction of I2.

Figure 5

Forest plots of hazard ratios for myocardial washout for all eligible studies. Forest plots of hazard ratios for myocardial washout in relation to cardiac death (upper panel) and cardiac events (lower panel) for all eligible studies. Random effects model: black diamonds are pooled estimates.

Figure 6

Forest plots of hazard ratios for myocardial washout in relation to cardiac events for qualitative best studies. Random effects model: squares are mean weighted hazard ratios (logarithmic scale). The size of the squares represents study weight and horizontal lines represent 95% CI. Black diamond is pooled estimate.

Sensitivity analysis, using the fixed-effects model instead of the random-effects model, resulted in similar magnitudes of the pooled HR estimates and conclusions about heterogeneity. Furthermore, no evidence for publication bias was found.

Discussion

Data from our analysis suggest that patients with HF and decreased late H/M or increased myocardial MIBG washout have a worse prognosis compared with those with normal semi-quantitative myocardial MIBG parameters. Furthermore, decreased late H/M in patients with HF is not associated with cardiac death.

The notion that semi-quantitative myocardial parameters of 123I-MIBG are determinants of prognosis in patients with HF is plausible. In patients with chronic HF, angiotensin-converting enzyme inhibitors,2 β-receptor antagonists,3,4 spironolactone5,6 and chronic cardiac resynchronization therapy7 have shown to ameliorate functional capacity and prognosis. Using semi-quantitative analysis, these beneficial effects were associated with an increase in 123I-MIBG uptake and a reduced washout. However, one of the major problems with these studies is that many have limited power, analysing only relatively small numbers of patients. To address the issue of power and to derive more robust estimates of prognosis associated with semi-quantitative myocardial parameters of 123I-MIBG we pooled published studies. A systematic review process was adopted in ascertaining studies, thereby avoiding selection bias.

In patients with chronic HF, sympathetic activity is initially increased as a compensatory mechanism. However, chronically elevated stimulation of the adrenergic system is associated with sustaining the process of remodelling. Cardiac sympathetic neuronal activity can non-invasively be assessed by the use of 123I-MIBG, a radio-labelled analogue of norepinephrine.1 After intravenous injection, 123I-MIBG is internalized by presynaptic nerve endings of postganglionic neuronal cells through the energy-dependent uptake-1 system. Anterior planar scintigraphic images are obtained 15 min (early) and 4 h (late) after injection. The commonly used myocardial 123I-MIBG indices are the heart-to-mediastinum ratio (H/M ratio) and myocardial washout. On anterior planar images, regions of interest (ROIs) are drawn over the heart (H) and the mediastinum (M). The mean count-density in each ROI is obtained and the H/M ratio (specific activity/non-specific activity) is calculated. Myocardial 123I-MIBG washout is calculated as the difference between the early and late H/M and expressed as a percentage of the early H/M. The early H/M probably reflects the integrity of presynaptic nerve terminals and uptake-1 function. The late H/M combines information on neuronal function from uptake to release through the storage vesicle at the nerve terminals. Myocardial 123I-MIBG washout is an index of the degree of sympathetic drive. This implies that increased adrenergic drive is associated with high myocardial 123I-MIBG washout and low myocardial 123I-MIBG delayed uptake.

Publication bias is a major concern in all forms of pooled analyses, as studies reporting significant findings are more likely to be published than those reporting non-significant results. Indeed, it is not unusual for small-sized early studies to report a positive relationship that subsequent larger studies fail to replicate. In the present study, there was no evidence for publication bias.

As with any systematic review, further sources of bias may have affected these results. Another potential source of bias is whether all relevant studies have been identified. While a small number of part-published studies may have been omitted, it is likely that all key published analyses have been identified. Furthermore, bias in the selection of studies was avoided by adopting rigid inclusion criteria. A potential source of bias specific to this study is that of overlapping data sets. Although a number of such data sets were identified, this bias was minimized by excluding such data sets, replacing these by only the most recent or complete study. Heterogeneity between studies may represent a further potential source of bias. Indeed for late H/M and early H/M I2 was equal or more than 75%. However, after limiting the analysis to the qualitative best studies heterogeneity was reduced to an acceptable level. Another common denominator (i.e. patient characteristics, NYHA classification, etiology of HF, events, duration of follow-up or technical aspects) in these three ‘high quality’ studies that explained for the reduction in heterogeneity could not be found.

However, due to heterogeneity the results were obtained from a relatively small number of high quality studies. This relatively small number of high quality studies emphasizes the need to improve the quality of future research. In the light of the findings of our study special attention should be given to parameters that improve the internal validity (adequate methods). In addition, the scorings system developed for the current analysis may be of use in the design of future studies.

Only five studies performed and analysed all the three semi-quantitative myocardial MIBG parameters.20,23,26,33,34,37 The exclusion of one or two semi-quantitative myocardial MIBG parameters in the remaining studies in which insufficient data were presented for specific data point extraction might have affected our results. The selection of studies for inclusion is a critical factor for any such analysis, and in accordance with the Cochrane collaboration principles, we have attempted to avoid selection bias by including all relevant studies.38 Analyses without these partially absent semi-quantitative myocardial MIBG parameters showed a worse prognosis for patients with HF and increased myocardial sympathetic activity. However, we cannot exclude or estimate the bias from these studies.

Therapy like medication or bi-ventricular devices may also have been a potential factor for bias. Myocardial MIBG uptake is influenced by drugs like angiotensin-converting enzyme inhibitors,2 β-receptor antagonists,3,4 and spironolactone.5,6 In most of the included studies baseline information on medication is given indicating that patients are already on medication at the time of the MIBG scintigraphy. It is to be expected that these drugs may have increased myocardial uptake and reduced myocardial washout at the baseline assessment. This effect may therefore have blunted the found results. Despite this, possible bias analyses showed a worse prognosis for patients with HF and increased myocardial sympathetic activity. However, we cannot exclude or estimate the bias from medication.

In addition to the prime emission of 159 keV photons, 123I emits high-energy photons of more than 400 keV [∼2.87%, main contributor 529 keV (1.28%)]. These high-energy photons lead to penetration of the collimator septa and cause scatter that is detected in the 159-keV energy window. Planar H/M ratios are influenced by scatter and septal penetration from increasing amounts of liver activity. These effects are less pronounced for medium-energy collimators.39 Regardless, low-energy collimators were frequently used in the selected studies. This may have biased the outcome of the analysis.

Accuracy in (semi-)quantification of scintigraphic images is important. However, when the size of the quantified region is close to the spatial resolution of the gamma-camera, the quantification becomes imprecise: the activity is systematically underestimated. This phenomenon, i.e. partial volume effect, is notably present when measuring radioactivity in relatively small structures. The partial volume effect plays a role in the assessment of regional (i.e. small volumes) myocardial distribution of 123I-MIBG as assessed with SPECT. However, the focus of this systematic review is on global (i.e. larger volumes) planar myocardial measurements. In these planar measurements the partial volume effect plays a minor role.

Furthermore it is important to realize that prognosis is conditional. In the ideal case one would like to monitor the natural history of a disease. However, in (clinical) practice patients will be treated and thereby influence the history of a disease. In that sense prognosis is conditional.

In conclusion, data from our analysis suggest that patients with HF and abnormal semi-quantitative myocardial MIBG parameters have a significantly worse prognosis compared with those with relatively normal semi-quantitative myocardial MIBG parameters. More specifically, a decreased late H/M is associated with a higher incidence of cardiac events and is, however, not associated with cardiac death. Furthermore, an increased myocardial washout is associated with both cardiac death and cardiac events. Therefore, semi-quantitative myocardial MIBG uptake and washout are promising prognostic markers in patients with HF.

Acknowledgments

Conflict of interest: none declared.

Appendix

Search strategy

A computer-assisted search was performed of the medical databases MEDLINE (January 1980 to January 2006), PubMed (January 1980 to January 2006), EMBASE (January 1980 to January 2006), the Cochrane Controlled Trial Register and the Cochrane Database of Systematic Reviews (from their inception to January 2006). A highly sensitive search strategy described by Haynes et al.13 was adapted to our specific requirements and resulted in the following search strategy: (((MIBG* [WORD] OR metaiodobenzylguanidine [WORD]) AND (heart [WORD] AND failure [WORD])) AND (incidence [MESH] OR mortality [MESH] OR follow-up studies [MESH] OR mortality [SH] OR prognos* [WORD] OR predict* [WORD] OR course [WORD])). No language restriction was used. We searched the Internet for unpublished studies and abstracts. Review articles and bibliographies of relevant studies were hand-searched to identify additional relevant publications.

Methodological quality assessment

As there are no widely agreed quality criteria for assessing prognostic studies, criteria were formulated based on suggestions made by Altman.14 Internal validity criteria were assigned a score for presence (adequate methods; score = 1) or absence (inadequate methods, potential for bias; score = 0). Quality scores were expressed as a percentage of the maximum score (9). No criteria for external validity (evaluation of generalizability) were formulated.

Criteria for internal validity comprised of items related to patient sample (fully defined inclusion criteria; fully described clinical and demographic characteristics), study design (only a prospective design was defined as adequate) definition and assessment of outcome (fully defined outcome; objective and unbiased assessment of outcome), duration of follow-up (follow-up of at least 3 months), semi-quantitative 123I-MIBG myocardial parameters (fully defined and described; blind assessment) and avoidance of verification bias (assessment of outcome blind to 123I-MIBG result).

Statistical analysis

The association between 123I-MIBG parameters of myocardial sympathetic activity and cardiac death, and the association between 123I-MIBG parameters of myocardial sympathetic activity and cardiac events were derived as a weighted average of study-specific estimates of the HR, using inverse variance weights.15 The logHR and the corresponding variance were used as data points for pooling purposes. In studies quoting the 95%CI and HR, the logHR variance was calculated by:

Embedded Image

where UL and LL are the upper and lower limits of the 95% CI for ln(HR).15 In studies not quoting the HR or CIs, these were calculated from data presented using a hierarchical series of steps as per Parmar et al.15 Data points were calculated from the following parameters: the HR point estimate, the log-rank statistic or its P-value.

In studies quoting the P-value and HR, the study variance was estimated from the cumulative distribution function of the normal distribution. To be conservative, significant values quoted in papers as less than the specified threshold were assumed to be at that threshold. If data were only presented as graphical representations of the survival distributions, survival rates were extracted at specified times in order to reconstruct the HR-estimate and its variance, under the assumption that the rate of patients censored was constant during the study follow-up.15 If censoring data were presented, censored subjects were subtracted from the denominators by counting tick marks on survival curves. Finally, if survival at a given point in time, S, defined by X/N (X being the number of patients surviving and N the number of patients followed-up), the hazard was estimated from: β= −log(S)/t, in which t is the time at which survival was measured, as per Samson et al.16 If insufficient data were presented for data point extraction studies were excluded from pooling.

HRs obtained after adjusting for other covariates were used when provided. Summary data from published studies were pooled using the random-effects models as a primary analysis. The random-effects model assumes that studies were a random sample of a hypothetical population of studies taking into account variability within and among studies.17 The percentage variability of the pooled HR attributable to heterogeneity among studies was quantified using the I2 statistic.18 Studies with a high variation across studies due to heterogeneity (I2 values ≥ 75%) were not aggregated.18 If heterogeneity was present, sources of heterogeneity were explored and a decision was made whether to aggregate the studies or not. Sensitivity analysis was performed by reanalysing the data using the fixed-effects model.

Studies were plotted in order of decreasing variance of the logHR. Horizontal lines represent 95% CIs (i.e. forest plots). Each square box represents the HR point estimate and its area is proportional to the weight of the study. The black diamond represents the overall summary estimate, with CIs given by its width. The unbroken vertical line is at the null value (HR = 1.0). Estimates from small studies that have less precision in estimating the underlying HR will therefore scatter widely. Evidence of publication bias was examined by constructing funnel plots.19 Statistical analyses were performed using Review Manager (RevMan) [Computer program], version 4.2 for Windows, Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2003.

References

View Abstract