Introduction to Results section

591 records were identified through database searching and 28 additional records were identified through other sources; 428 remained after duplicates were removed. One hundred full-text articles were assessed for eligibility and after the exclusion of 76 full-text articles, five high quality SRs and 19 full text published RCTs were included in our SR (Appendix 3 and 4). Of the included RCTs, only three were judged to be of low risk of bias. The PRISMA flowchart outlining the study selection process is presented in Appendix 2.

Updating only one SR of already available SRs was not possible due to different, and a wide range of our research questions, as well as different inclusion criteria and followed-up duration of RCTs included. If data from existing SRs or HTAs was not available we used data from the included 19 RCTs.

Five high quality SRs were found to answer some of the assessment element questions {Feltner et al, 2014; Kotb et al, 2015; Pandor et al, 2013; Inglis et al, 2011; Clark et al, 2007}, details can be found in Appendix 3. Only three RCTs on STS in chronic heart failure patients {Laramee 2003, Riegel 2002, Riegel 2006} were included in all five SRs (see Appendix 5). Becausue not all assessment element questions could be answered by the results from the five included SRs, 19 published RCTs (Appendix 4 and 6), were included in order to answer the remaining assessment element questions. Out of them, 17 RCTs were already included in one or several of the five SRs (Appendix 5).

The two most recent RCTs, published by Angermann et al 2012 { } and Krum et al 2013 { } were not included in the SR published by Kotb et al 2015 { }, and RCT published by Krum et al 2013 { } was not included in the SR published by Feltner et al 2014 { }. Out of the 19 RCTs (see Appendix 4 and 6), only three  were judged to be of low risk of bias {DeBusk 2004, GESICA 2005, Chaudhry 2010}, five as unclear risk of bias {Tsuyuki 2004, Cleland 2005, Riegel 2006, Sisk 2006, Krum 2013} and the remaining 11 were rated as high risk of bias {Gattis 1999, Rainville 1999, Barth 2001, Riegel 2002, Laramee 2003, Galbreath 2004, DeWalt 2006, Ramachandran 2007, Wakefield 2008, Mortara 2009, Angermann 2012}. The majority of RCTs (17 RCTs) had a follow-up period of 6 or 12 months or more; more specifically two RCTs had 3 months period {Barth 2001, Laramee 2003}; eight had 6 months follow-up period {Gattis 1999, Riegel 2002, Tsuyuki 2004, Riegel 2006, Ramachandran 2007, Wakefield 2008, Chaudhry 2010, Angermann 2012}; for six  RCTs follow-up period was 12 months {Rainville 1999, DeBusk 2004, DeWalt 2006, Sisk 2006, Wakefield 2008, Krum 2013}. One RCT had follow-up period of 8 months {Cleland 2005}, one of 16 months {GESICA 2005} and one RCT had 18 months {Galbreath 2004} follow-up period. Only one ongoing RCT was found in publicly available register noted that participants were not yet being recruitinig (Appendix 7).

Answers on specific assessment element questions

All five high quality systematic reviews were used to answer the question ”What is the effects of Structured telephone support (STS) on overall mortality in adults with chronic heart failure, compared to standard care without Structured telephone support (STS)?” {Feltner et al, 2014; Kotb et al, 2015; Pandor et al, 2013; Inglis et al, 2011; Clark et al, 2007}, as well as two more recent RCTs {Angremann 2012, Krum 2013}, details can be found in Appendices 3 and 4.

Data found on overal mortality in STS group compared with usual care were conflicting. One SR Feltner et al, 2014 { }, including RCT published by Angermann 2012  { } and one SR with network meta analysis published by Kotb et al, 2015 { } were found which reported significantly lower overal mortality in STS group comparing with usual care. RCT published by Cleland 2005 { } found significantly lower overal mortality in STS group (UC=20 (24%) vs NTS=27 (16%), P=0.0397) as well.

The three other SRs and two out of 19  RCTs, reported non-significant difference in overall mortality in STS group compared to usual care {Pandor 2013, Inglis 2011, Clark 2007, Angermann 2012, Cleland 2005}.

Seven studies ({Krum et al 2013, Gatiss 1999, Rainville 1999, DeWalt et al 2006 { }, Chaudhry et al 2010 { }, Angermann et al 2013, DeBusk et al 2004) assessed the composite outcome (composite end point of all-cause death or hospitalisation or combined outcome of rehospitalisation, emergency department visit, or death). Only three of them {Krum et al 2013, Gatiss 1999, Rainville 1999}, found a significant reduction on this outcome.  

All but three trials in this SR were rated at a medium risk of bias; three trials were rated at high risk of bias primarily for high risk of selection bias and measurement bias. Most trials were conducted in the United States: three in multicenter settings and all others at a single center. Three trials were conducted in multicenter settings in Europe and Canada one trial was conducted at a single center in Brazil. In this most recent SR and HTA { }, STS interventions produced a mortality benefit, with  RR (95% CI) of 0.74 (0.56–0.97), at 3-6 months, with NNT of 27 (Table 1). Specific RR with 95% CI was listed for already included RCTs in previous SRs and in our new SR {Laramee 2003, Riegel 2006, Wakefield 2008, Angermann 2012}. According the Angermann et al, 2012 { } results, mortality was significantly lower in STS group comparing with usual care (HR, 0.62; 0.40–0.96; P=0.03)

Table 1. Mortality data from six RCTs on STS compared with usual care, Feltner et al, 2014 { }

Abbreviations: NNT = number needed to treat; RR = risk ratio; SOE = strength of evidence according the Feltner et al 2014 { }

Kotb et al, 2015 { } with direct comparisons of data from 15 trials showed mortality Odds Ratio (OR) of 0.85 (0.73 to 1.0), compared to usual care. In Network meta-analysis (NMA), compared to usual care, structured telephone support significantly reduced the odds of mortality (Odds Ratio 0.80; 95% Credible Intervals CrI [0.66 to 0.96]).

Pandor et al, 2013 { } included 11 studies evaluating STS [10 used standard telephone equipment using human to human (HH) support and one provided support via an automated telephone interactive response system (HM) with an alert system]; 1 study assessed both STS and telemonitoring compared with usual care. The duration of follow-up ranged from 6 months to 18 months.

Compared with usual care, STS HH did not show statistically significant benefit in reducing all-cause mortality [hazard ratio (HR) 0.77, 95% CrI 0.55 to 1.08]. No favourable effect on mortality was observed with STS HM.

In SR published by Inglis et al, 2011 { }, out of the 25 full text peer-reviewed studies included in the meta-analysis, 16 evaluated structured telephone support (5613 participants), 11 evaluated telemonitoring (2710 participants), and two tested both interventions. Structured telephone support demonstrating a non-significant positive effect on all-cause mortality (RR 0.88, 95% CI 0.76 to 1.01, P = 0.08) (Box 1). In sensitivity analysis done on 9 RCT with a follow-up period longer than 6 months, also included in our new SR, difference between two groups was not statistically significant  {Rainville 1999, DeBusk 2004, Galbreath 2004, Cleland 2005, GESICA 2005, DeWalt 2006, Sisk 2006, Wakefield 2008, Mortara 2009}.

Box 1. All-cause mortality, structured telephone support in comparison with usual care, in SR published by  Inglis et al, 2011 { }

All-cause mortality

STS vs UC RR 0.88 (95% CI 0.76 to 1.01, P = 0.08) (15 published trials, n= 5563)

Sensitivity analysis: Follow-up period (>6 months), 9 published trials, n=4292, RR 0.87 [0.74, 1.02]:

Cleland et al, 2005 (Struct Tele): 0.66 [0.40, 1.11]

DeBusket al, 2004: 0.74 [0.44, 1.26]

DeWalt et al, 2006: 0.79 [0.18, 3.37]

Galbreath et al, 2004: 0.70 [0.47, 1.04]

GESICA 2005 (DIAL): 0.88 [0.77, 1.00]

Mortara et al, 2009 (Struct Tele): 1.51 [0.62, 3.68]

Rainville et al, 1999: 0.25 [0.03, 2.04]

Sisk et al, 2006: 1.00 [0.57, 1.75]

Wakefield  et al, 2008: 1.12 [0.60, 2.09]

Adding two studies on structured telephone support (Angermann 2007; Krum 2009 (CHAT) which were not full peer-reviewed publications to the meta-analysis increased the effect of structured telephone support (RR 0.85, 95% CI 0.75 to 0.97, P = 0.02, I² 0%).

Abbreviations: RR = risk ratio; STS: structured telephone support; UC: usual care

telephone support, based on 10 RCTs results (also included in our SR) {Cleland 2005, Rainville 1999, Gattis 1999, Barth 2001, Riegel 2002, Laramee 2003, DeBusk 2004, Tsuyuki 2004, GESICA 2005, Riegel 2006} not statistically significant reduced all-cause mortality by 20% (8% to 31%); RR 0.85 (0.72 to 1.01, P=0.06, based on 482 deaths in 3542 patients)(Box 2).

Box 2. All-cause mortality, structured telephone support in comparison with usual care, in SR published by  Clark et al, 2007 { }

All-cause mortality

Structured telephone support RR 0.85 (0.72 to 1.01, P=0.06), based on 482 deaths in 3542 patients;

Cleland et al, 2005: RR 0.61 (0.40 to 0.94)

Gattis et al, 1999: RR 0.61 (0.15 to 2.46)

Rainville et al, 1999: RR 0.25 (0.03 to 2.01)

Barth et al, 2001: RR not estimated

Riegel et al, 2002: RR 0.88 (0.50 to 1.54)

Laramee et al, 2003: RR 0.90 (0.44 to 1.82)

DeBusk et al, 2004: RR 0.74 (0.44 to 1.26)

Tsuyuki et al, 2004: RR 1.30 (0.64 to 2.64)

GESICA Investigators 2005: RR 0.95 (0.75 to 1.20)

Riegel et al, 2006: RR 0.71 (0.26 to 1.93)

Angermann et al, 2012 { } in open, randomized, 2-armed, parallel-group, multicenter 6 months trial, aimed to clarify the mode of action of the program and individual patient requirements, thus providing a rational basis for more targeted health care strategies in heart failure, showed overall mortality significantly lower in intervention group in comparison to usual care group. Overall, 32 (9%) patients in the intervention group and 52 (14%) patients in the usual care group died (HR, 0.62; 0.40–0.96; P=0.03). Five of these deaths occurred after dropout (HNC: n=4, UC: n=1).

Krum et al, 2013 { } aimed at determining whether an automated telephone support system would improve quality of life and reduce death and hospital admissions for rural and remote heart failure patients in Australia, at 12 months follow up. This RCT reported that 16 out of 209 patients in the usual care (UC) group and 17 out of 170 patients in UC + Intervention (I) group died during the study. This resulted in an non statisticaly significant unadjusted hazard ratio (HR) for  all-cause death of 1.3 (range 0.65–2.77, P = 0.43) and an adjusted hazard ratio of 1.36 (0.63–2.93, P = 0.439).

Despite the fact that study Krum et al, 2013 { }  found no change in the primary endpoint (Packer clinical composite score, a clinical composite consisting of mortality, overnight hospitalization for worsened HF and NYHA class global self-assessment), the intervention did lead to a significant reduction in the risk of the composite of all-cause death or hospitalization, as well as all-cause hospitalization alone. 124 of 209 UC and 86 of 170 UC + I patients reached the endpoint of all-cause death or all-cause hospitalization. This resulted in an unadjusted hazard ratio of 0.75 (range 0.57–0.99, P = 0.045) and an adjusted hazard ratio of 0.70 (range 0.53–0.92, P = 0.011). The same is true for other RCTs that looked at composite end point of all cause mortality and rehospitalization for heart failure or all cause hospitalization {Gatiss 1999, Rainville 1999}. DeWalt et al 2006 { }, Chaudhry et al 2010 { }, and Angermann et al 2013 did not find any significant reduction in the risk of the composite of all-cause death or hospitalization or in the combined outcome of rehospitalization, emergency department visit, or death, DeBusk et al 2004 { }  .

 Details can be found in Appendix 3 and 4.