Psychological interventions for parents of children and adolescents with chronic illness.

Parenting a child with a longstanding or life-threatening illness is very difficult, and can have a negative impact on many aspects of the parent's life. Parents of these children often have difficulty balancing caring for their child with other responsibilities such as work, social life, finance and other household tasks. As a result they may experience more stress, worries, sad feelings, family arguments and troubling child behaviour. Parents also have a major influence on their child's well-being and adjustment, and play an important role in how their child adapts to living with an illness. Treatments for parents of children with a longstanding illness aim to improve parent distress, parenting behaviours, family conflict, child distress, child disability and the child's medical symptoms. Thirty-five studies were found in the search, but only 26 of these had data that could be used in the analyses. We found studies for six child illnesses (painful conditions, cancer, diabetes, asthma, traumatic brain injury and eczema) and four types of psychological therapies (cognitive behavioural therapy, family therapy, problem solving therapy and multisystemic therapy). We looked at the effects of the treatments on parent distress, parenting behaviours, family conflict, child distress, child disability and symptoms of the child's illness immediately after the treatment and at the first available follow-up time point after the treatment had ended. We analysed the data in two ways; first we grouped the studies by each individual illness and then we grouped the studies by each individual psychological therapy. Psychological therapies can help reduce pain in children with painful conditions. Where there were results available from studies of different chronic illnesses, we found that cognitive behavioural therapy can improve the child's medical symptoms. Problem solving therapy can improve parent's distress and their ability to solve problems. More studies of psychological treatments for parents of children with a longstanding illness are needed.

Randomised controlled trials (RCTs) that compared parental psychological interventions with attention control, other active treatment or waiting list control were considered for this review. The parent intervention had to be primarily psychological in nature. Studies that met the inclusion criteria consisted of the following: RCT, published in full in a peer-reviewed journal; primary aim of the trial was an evaluation of a psychological intervention; involved parents of children who have an illness for three months or more ( Van der Lee 2007 ); involved parents of children adjusting to a diagnosis of cancer; had a n of 10 or more in both the treatment and control arm at end of treatment or follow-up. Parents of children who have endured a chronic illness for three months or more. Parents were regarded as the primary caregiver of a child or adolescent under the age of 19 years. Parents were defined, for the purposes of this review as any adult who adopts the responsibility for the role of parenting the child (this could include biological parent, guardian, other adult family member). There was no lower age limit for the children, however, by the definition of ‘chronic illness’, the child must be three months or more. The children must also be experiencing one (or more) of the following physical illnesses: headache; recurrent abdominal pain; back pain; idiopathic pain conditions; complex regional pain syndrome (CRPS); rheumatological conditions (e.g. arthritis and fibromyalgia); sickle cell disease; cancer; diabetes mellitus; asthma; traumatic brain injury; inflammatory bowel diseases; skin diseases (e.g. eczema); gynaecological disorders (e.g. chronic dysmenorrhoea and endometriosis). Chronic illnesses were selected from the National Survey of Children with Special Health Care Needs 2009 to 2010 ( Data Resource Center 2010 ). It was impractical to include all chronic illnesses on this list, therefore, we selected the most common. However, three illnesses (cancer, inflammatory bowel diseases and gynaecological disorders) were not included in the Current Health Conditions and Functional Difficulties but were added for the purposes of this review. Cancer has a high incidence level and it was predicted that in 2007, there were 10,400 children with cancer in the US alone under the age of 14 ( Linabery 2007 ). Studies that investigate interventions with parents of children who have ‘survived’ an illness such as childhood survivors of cancer were also eligible for inclusion. Inflammatory bowel diseases and gynaecological disorders are also common conditions in childhood and adolescence and were included because they are thought to be prevalent but under-represented in the academic literature. Studies were included if the interventions were primarily psychological, and had credible, recognisable psychological/psychotherapeutic content, and were specifically for, or included parents. Psychological interventions were defined as any psychotherapeutic treatment specifically designed to change parent cognition or behaviour, or both, and had the intention of improving parent or child outcomes. However, studies in which parents acted as ‘coaches’ were excluded from this review. The intervention had to aim to provide treatment to the parent rather than teach them to deliver an intervention to their child. Similarly, we also excluded health promotion therapies such as intervening with the parent to cease smoking to improve their child's asthma. We have excluded studies that combine psychological interventions with pharmacological interventions or are qualitative in nature as it is difficult to combine qualitative and quantitative data. Primarily, parent outcomes were the target of our review. However, if the study also reported child outcomes as stated below, we also analysed and reported these data. We analysed data at post-treatment and the first available follow-up period, where reported. Primary outcomes, depending on specific treatment, were: parent behaviour, parent mental health, child behaviour/disability, child mental health, child primary symptom, family function and adverse events. We made a judgement when studies reported multiple measures within one of the six outcome domains without defining their primary or secondary outcome measure. The rules of this judgement were to select the most generic, reliable and most frequently used measure within the field, and most appropriate for the given outcome category. When both parents and children reported on a measure, we extracted the self report item unless the non-self report measure was a more generic measure. For family functioning measures, we extracted parent data over child data as the review is focused on whether interventions can help parents of children with a chronic illness. We searched electronic databases and reference lists to identify studies matching the criteria. In addition, we also contacted experts and study authors for additional studies. We searched four databases for studies from inception to June (week 4) 2011 and again in March (week 1) 2012: The Cochrane Central Register of Controlled Trials (CENTRAL, beginning 1968); MEDLINE via Ovid (beginning 1946); EMBASE via Ovid (beginning 1974); PsyclNFO via Ovid (beginning 1806). The Cochrane Central Register of Controlled Trials (CENTRAL, beginning 1968); MEDLINE via Ovid (beginning 1946); EMBASE via Ovid (beginning 1974); PsyclNFO via Ovid (beginning 1806). We adapted the search strategies from the MEDLINE search (see Appendix 1 ) and they are included in Appendix 2 . There was no language restriction imposed and no unpublished literature or grey material was included. The search strategy included four categories of words: psychological interventions, parents, children/adolescents and chronic illnesses (as stated above), and was refined by a methodological filter used to identify RCTs according to Cochrane guidance ( Higgins 2011 ). We performed a reference list and citation search of each selected study which identified further studies meeting the inclusion criteria. We then repeated this stage for such studies. We also checked meta-analyses and systematic reviews that met the inclusion criteria for appropriate studies and included them if they met the inclusion criteria. We also contacted authors of selected studies and experts in the field for further studies that had not already been identified from the search. EF performed the searches of each database and collated results. Two review authors (EF, EL) then sifted through potential studies and identified those eligible to be included with CE acting as arbiter. No blinding of study authors' names, institutions or journals occurred during this process. We resolved any disagreements by discussion between all review authors. We made selection of abstracts using the following criteria. Participants Parents must be referred to in the title or abstract of each study The parent must be the primary caregiver of the child Children must have one or more of the chronic illnesses listed above Children must be in the age range three months to 19 years There must be 10 or more participants in each condition at the end of the treatment assessment Parents must be referred to in the title or abstract of each study The parent must be the primary caregiver of the child Children must have one or more of the chronic illnesses listed above Children must be in the age range three months to 19 years There must be 10 or more participants in each condition at the end of the treatment assessment Intervention The intervention must be primarily psychological in at least one condition Must be of RCT in design One or more parents must be treated by the intervention The parents and/or child must be measured at baseline and at a point in time during or after the intervention The intervention must be primarily psychological in at least one condition Must be of RCT in design One or more parents must be treated by the intervention The parents and/or child must be measured at baseline and at a point in time during or after the intervention Comparison groups Attention control group Active treatment group Treatment as usual group: this would consist of usual doctors' appointments and treatment without added psychological therapy Wait list control Attention control group Active treatment group Treatment as usual group: this would consist of usual doctors' appointments and treatment without added psychological therapy Wait list control Numerical outcomes presented We then obtained the selected studies meeting the criteria in full and EF and EL read and assessed them independently. Two review authors (EF and EL) carried out data extraction from studies that were identified by all review authors as appropriate for inclusion. The data extraction sheet was adapted from Eccleston 2009a and Eccleston 2009b . It included references, the diagnosis of the child's chronic illness, aspects of the intervention or therapy, characteristics of the treatment team, the setting of the intervention and outcome measures. We assessed risk of bias using the recommended Cochrane guidance ( Higgins 2011 ). Of the five suggested risk of bias categories, we judged studies on random sequence generation (selection bias), allocation concealment (selection bias), blinding of outcome assessment (detection bias), incomplete outcome data (attrition bias) and selective reporting (reporting bias). We excluded the option of ‘blinding participants and personnel’ because we deemed it redundant as neither therapists nor patients can be blinded to whether they deliver or receive treatment. Decisions about random sequence generation were based on whether authors gave a convincing method of randomisation. Allocation concealment judgements were based on whether sufficient methods were employed for random allocation to take place. Participants being stratified by age or gender did not count as bias but are noted in the tables. We judged risk of blinding of outcome assessment on whether the measures were administered and collected by an assessor who was blind to the treatment allocation. We judged high risk of attrition bias when no description of attrition was reported. We made an unclear decision when there was an adequate decision given but authors did not report whether there were significant differences between completers and non-completers. We concluded low risk of bias when authors gave both a description of attrition and stated that there were no significant differences between completers and non-completers. Third, we judged selective reporting bias in two parts. First, we judged studies on whether data were fully reported in the study or if authors later responded to data requests. Second, we rated each study on a three-point scale for concordance (two points = full concordance, one point = partial concordance, zero points = no concordance). We rated studies for concordance between study aims and measures (i.e. if aims corresponded to measures stated in methods section) and between measures and results (i.e. if all measures were reported in results, and no additional measures were added to results that were not stated in the methods section). We assessed quality of studies using the method advocated by Yates 2005 . Two authors (EF, EL) rated study quality for each study and disagreements were settled by discussion between all authors. The rating scale consists of two sections which creates an overall quality of study score of 35. The first section measures treatment quality (0 to 9) which assesses the treatment rationale, duration of treatment, manualisation, therapist training and engagement of patients. The second section measures the quality of the study design and methods used (0 to 26). This section measures the inclusion/exclusion criteria, rates of attrition, description of patient sample, steps taken to minimise bias (randomisation, allocation bias, measurement bias and treatment expectations), justification of outcomes and whether they are reliable and valid, follow-up, adequate statistical analyses (power, sufficient sample size, planned data analysis, statistical reporting and intention-to-treat analysis) and finally choice of control group. The studies are then categorised as ‘ high quality’ or ‘ low quality’ of being biased. The boundary between high and low quality was defined as the mid-point (quality of study high quality ≥ 18, low quality ≤ 17, treatment quality high quality ≥ 5, low quality ≤ 4, quality of study design and methods used high quality ≥ 15, low quality ≤ 14). We investigated four classes of psychological therapies: cognitive behavioural therapy (CBT), family therapy (FT), problem solving therapy (PST) and multisystemic therapy (MST). CBT is based on theories of behavioural analysis ( Bergin 1975 ), cognitive theory ( Beck 1979 ) and social learning theory ( Bandura 1977 ). CBT therefore includes a range of strategies with the goals of modifying social/environmental and behavioural factors that may exacerbate or cause symptoms, and modifying maladaptive thoughts, feelings and behaviours to reduce symptoms and prevent relapse. FT is based on family systems theory ( Haley 1976 ; Minuchin 1974 ), which emphasises the role of the family context in an individual's emotional functioning. FT interventions typically focus on altering patterns of interactions between family members, and include structural family therapy ( Minuchin 1974 ), strategic family therapy ( Haley 1976 ) and behavioural systems family therapy ( Robin 1989 ). PST is based on the D'Zurilla 1982 social problem solving model, which defines problem solving in terms of an individual's ability to recognise problems and use cognitive and behavioural skills to solve them. PST includes didactic instruction in problem solving skills, followed by in-session modelling, behavioural rehearsal and performance feedback, as well as homework assignments ( D'Zurilla 2007 ). Finally, MST is an intensive family and community-based intervention based on the Bronfenbrenner 1979 social ecological model and family systems theory ( Haley 1976 ; Minuchin 1974 ). MST therefore targets the patient, their family and broader systems such as the patient's school, work or medical team as needed. MST incorporates a wide range of evidence-based intervention techniques based on the individual needs of the patient and family ( Henggeler 2003 ), including cognitive-behaviour approaches, parent training and family therapies. We extracted data immediately post-treatment (i.e. immediately after the treatment programme had finished). Where data were available, we also analysed studies at follow-up, which is classed as the first available time point after post-treatment. We categorised outcomes into one of six outcome domains: parent behaviour, parent mental health, child behaviour/disability, child mental health, child symptoms and family functioning. Where studies had more than one comparator group, we chose the ‘active control group’ over ‘standard treatment’ or ‘wait list control’ groups. There are four therapies (CBT, FT, PST and MST), eight conditions (asthma, cancer, diabetes, gynaecological disorders, inflammatory bowel syndrome, painful conditions (these were grouped together due to the homogeneous nature of the trials), skin diseases and traumatic brain injury), two time points (post-treatment and follow-up) and six possible outcomes (parent behaviour, parent mental health, child behaviour/disability, child mental health, child symptoms and family functioning). There are six categories by which we analysed data. For each condition, across all types of psychological therapy, what is the effectiveness for the six outcomes immediately post-treatment? For each condition, across all types of psychological therapy, what is the effectiveness for the six outcomes at follow-up? For each psychological therapy, across all conditions, what is the effectiveness for the six outcomes immediately post-treatment? For each psychological therapy, across all conditions, what is the effectiveness for the six outcomes at follow-up? The interaction between the condition and the psychological therapy effectiveness. Investigation of characteristics of particularly effective treatments. Analyses are presented for each of the six outcomes, however, due to the heterogeneous nature of the conditions and studies, this was not always possible. We pooled data using standardised mean difference and random-effect models as studies did not consistently use the same scales when measuring the same outcomes. Cohen's d effect sizes can be interpreted as follows: 0.2 = small, 0.5 = medium, 0.8 = large ( Cohen 1992 ). Where possible, we combined data in a metaanalysis and, following Cochrane guidance ( Higgins 2011 ), presented data in the form of numbers needed to treat and numbers needed to harm. We contacted authors of studies when data were not reported fully in publications. However, when authors could not send data to the review authors or were non-responsive to emails, we excluded data. Subgroup analysis explored the possible sources of heterogeneity (see Results). Biases are reported within the results section of the review following Cochrane guidance on bias reporting ( Higgins 2011 ). When possible, we attempted to use a failsafe N to control for publication bias. When there were multi-arm trials or trials that compared more than one active treatment, we used the primary active treatment and compared with the least biased comparator (typically standard care or treatment as usual). Analyses of the following subgroups are presented where data permitted: parent-only interventions versus family-based interventions; intervention effects within specific illnesses; intervention effects across specific types of psychological interventions. We also explored heterogeneity through subgroup analysis (see ‘Results’).

See: ‘Characteristics of included studies’ and ‘Characteristics of excluded studies’. We extracted a total of 114 papers to identify whether they met the full inclusion criteria; 107 papers were found in the initial search, and a further seven studies were identified later in an updated search before publication. Of these 114 papers, 99 were found from the search of databases, six papers from the citation search, four papers from reference searches and five papers from authors of included studies. We deemed 35 studies (45 papers) to meet the inclusion criteria for the review, whilst 61 studies (69 papers) were excluded ( Aleman 1992 ; Anderson 1999 ; Betancourt 2004 ; Braga 2005 ; Bruzzese 2008 ; Burke 1997 ; Burke 2001 ; Cakan 2007 ; Canino 2008 ; Carey 2008 ; Chernoff 2002 ; Chiang 2009 ; Ellis 2007 ; Ellis 2008 ; Evans 1999 ; Field 1998 ; Forsander 1995 ; Forsander 2003 ; Garbutt 2010 ; Gerber 2010 ; Giallo 2008 ; Glang 2007 ; Gustafsson 1986 ; Harris 2001 ; Haus 1976 ; Hernandez 1998 ; Hommel 2012 ; Hovell 1994 ; Humphreys 2000 ; Ireys 1996 ; Ireys 2001 ; Jay 1990 ; Johnson 1987 ; Kamps 2008 ; Kaslow 2000 ; Kazak 1996 ; Kazak 2005 ; Ketchen 2006 ; Klinnert 2005 ; Klinnert 2007 ; Kroner-Herwig 1998 ; Kupfer 2010 ; Lasecki 2008 ; Logan 1997 ; Mendez 1997 ; Nelson 2011 ; Perez 1999 ; Rasoli 2008 ; Sanders 1989 ; Sanders 1996 ; Satin 1989 ; Scholten 2011 ; Sieberg 2011 ; Staab 2002 ; Sullivan-Bolyai 2010 ; Szczepanski 2010 ; Wade 2010 ; Walders 2006 ; Walker 1996 ; Warner 2011 ; Wysocki 1997 ). Of the 35 studies (45 papers) included in this review, 31 had two comparator arms and four studies had three comparator arms. Of the 31 studies that had two arms, 15 studies used active controls where patients had to actively engage in another type of treatment (e.g. education) whilst 19 used wait list or “treatment as usual controls”. The total number of participants at the end of treatment was 2723 (mean = 80 per study). The total number of participants entering treatment was 3214 (mean = 95 per study). Therefore, the completion rate for all studies was 85%, making the attrition percentage 15%. The proportion of completers across studies ranged from 59% to 100%. We categorised the studies by the primary illness of the children. There were 12 painful condition studies ( Allen 1998 ; Barakat 2010 ; Barry 1997 ; Connelly 2006 ; Duarte 2006 ; Hicks 2006 ; Kashikar-Zuck 2005 ; Kashikar-Zuck 2012 ; Levy 2010 ; Palermo 2009 ; Robins 2005 ; Sanders 1994 ). Six studies with the primary illness of cancer met the inclusion criteria ( Askins 2009 ; Hoekstra-Weebers 1998 ; Kazak 2004 ; Sahler 2002 ; Sahler 2005 ; Stehl 2009 ), nine diabetes studies ( Ambrosino 2008 ; Ellis 2004 ; Ellis 2005 ; Grey 2011 ; Laffel 2003 ; Lehmkuhl 2010 ; Olivares 1997 ; Wysocki 1999 ; Wysocki 2006 ), four asthma studies ( Celano 2012 ; Lask 1979 ; Ng 2008 ; Seid 2010 ), three traumatic brain injury studies ( Wade 2006 ; Wade 2006b ; Wade 2011 ) and one atopic eczema study ( Niebel 2000 ). However, no studies met the inclusion criteria for inflammatory bowel disease or gynaecological disorders. Similarly, we also categorised studies by the type of psychological therapy delivered. There were 19 studies that delivered CBT ( Allen 1998 ; Ambrosino 2008 ; Barakat 2010 ; Barry 1997 ; Connelly 2006 ; Duarte 2006 ; Grey 2011 ; Hicks 2006 ; Hoekstra-Weebers 1998 ; Kashikar-Zuck 2005 ; Kashikar-Zuck 2012 ; Laffel 2003 ; Levy 2010 ; Niebel 2000 ; Olivares 1997 ; Palermo 2009 ; Robins 2005 ; Sanders 1994 ; Stehl 2009 ), seven studies that delivered FT ( Celano 2012 ; Kazak 2004 ; Lask 1979 ; Lehmkuhl 2010 ; Ng 2008 ; Wysocki 1999 ; Wysocki 2006 ), seven studies that delivered PST ( Askins 2009 ; Sahler 2002 ; Sahler 2005 ; Seid 2010 ; Wade 2006 ; Wade 2006b ; Wade 2011 ) and two studies that delivered MST ( Ellis 2004 ; Ellis 2005 ). We were unable to extract quantitative data from nine of the 35 studies ( Barry 1997 ; Celano 2012 ; Duarte 2006 ; Grey 2011 ; Kazak 2004 ; Lask 1979 ; Lehmkuhl 2010 ; Olivares 1997 ; Robins 2005 ). These studies did not present means or standard deviations, or combined data with another study already included in the review ( Grey 2011 ). Therefore 26 studies (36 papers, 2253 participants at end of treatment) presented data that were included in at least one analysis ( Allen 1998 ; Ambrosino 2008 ; Askins 2009 ; Barakat 2010 ; Connelly 2006 ; Ellis 2004 ; Ellis 2005 ; Hicks 2006 ; Hoekstra-Weebers 1998 ; Kashikar-Zuck 2005 ; Kashikar-Zuck 2012 ; Laffel 2003 ; Levy 2010 ; Ng 2008 ; Niebel 2000 ; Palermo 2009 ; Sahler 2002 ; Sahler 2005 ; Sanders 1994 ; Seid 2010 ; Stehl 2009 ; Wade 2006 ; Wade 2006b ; Wade 2011 ; Wysocki 1999 ; Wysocki 2006 ). The proportion of therapy received by parent and child varied between studies. The majority of studies gave equal attention to both parent and child (22 studies). In seven studies only the parent received therapy, four of which studies were delivering treatment to parents whose children had been diagnosed with cancer. Four further studies spent the majority of treatment time with the child. The final two studies did not specify how much therapy the parent and child received. Twenty-eight studies treated patients in-person with the therapist, and seven studies used online programmes to deliver part or all of the therapy to patients. Twenty-five studies carried out therapy with individuals or with individual families, whilst eight studies used a group format. One further study used a combination of group and individual work. One study did not specify how treatment was carried out. A summary of the characteristics of therapy, and a narrative summary of treatment content, are presented in Table 1 and Table 2 respectively. Sixty-one studies did not meet the inclusion criteria for this study. Thirty-one studies had insufficient psychotherapeutic content, such as instruction, education, parents trained as ‘coaches’ for their children or health prevention interventions ( Aleman 1992 ; Anderson 1999 ; Braga 2005 ; Burke 1997 ; Burke 2001 ; Chernoff 2002 ; Chiang 2009 ; Evans 1999 ; Field 1998 ; Garbutt 2010 ; Giallo 2008 ; Glang 2007 ; Hovell 1994 ; Humphreys 2000 ; Ireys 1996 ; Ireys 2001 ; Johnson 1987 ; Kaslow 2000 ; Kazak 1996 ; Ketchen 2006 ; Klinnert 2005 ; Klinnert 2007 ; Kupfer 2010 ; Logan 1997 ; Mendez 1997 ; Nelson 2011 ; Perez 1999 ; Staab 2002 ; Sullivan-Bolyai 2010 ; Szczepanski 2010 ; Walders 2006 ). Sixteen studies had an aim that was irrelevant to the aim of the review such as fidelity studies, mixed illnesses or the intervention focusing on the parents communication with professionals ( Bruzzese 2008 ; Cakan 2007 ; Canino 2008 ; Carey 2008 ; Ellis 2007 ; Ellis 2008 ; Forsander 1995 ; Gerber 2010 ; Harris 2001 ; Hommel 2012 ; Jay 1990 ; Rasoli 2008 ; Scholten 2011 ; Wade 2010 ; Walker 1996 ; Wysocki 1997 ). Thirteen studies had an insufficient number of participants (n < 10) at post-treatment in or one more arms of treatment ( Forsander 2003 ; Gustafsson 1986 ; Haus 1976 ; Hernandez 1998 ; Kamps 2008 ; Kazak 2005 ; Kroner-Herwig 1998 ; Lasecki 2008 ; Sanders 1989 ; Sanders 1996 ; Satin 1989 ; Sieberg 2011 ; Warner 2011 ) and one paper recruited participants prospectively ( Betancourt 2004 ). These judgements were often difficult to make and led to extended discussion between review authors. We used five ‘Risk of bias’ categories: random sequence generation (selection bias), allocation concealment (selection bias), blinding of outcome assessment (detection bias), incomplete outcome data (attrition bias) and selective reporting (reporting bias) ( Figure 1 ; Figure 2 ). Sixteen studies described a convincing method of randomisation and we judged them to have a low risk of bias, a further 19 studies did not provide an adequate description and we judged them to be unclear. We rated no studies as high risk of bias for random allocation. There were 12 studies that described a convincing method of allocation and we judged them to have a low risk of allocation bias, a further 22 studies did not provide an adequate description and we judged them to be unclear. We rated ne study as high risk of allocation bias. Thirteen studies reported outcome assessors that were blinded to treatment allocation and we judged them to have a low risk of bias, a further 21 studies did not provide an adequate description and we judged them to be unclear, and we judged one study to have a high risk of outcome bias. Eleven studies reported attrition and found no significant differences between completers and non-completers, so we judged them to have a low risk of bias. Five studies reported attrition but did not report differences between completers and non-completers and so we judged them to be unclear and nine studies did not give an adequate description of attrition and so we judged them to be of high risk. Data could be fully extracted in 12 studies and were fully concordant between aims, measures and results and we judged them to have low risk of selective reporting bias. A further 13 studies were unclear, meaning data could not be extracted or aims, measures and results were only partially concordant. We found 10 studies to have high risk of selective reporting bias because data could not be extracted and they were only partially concordant. For the 35 studies that met the inclusion criteria, the mean overall quality of the study was 21.49 (standard deviation (SD) = 6.09, range seven to 32). This score is made up of the treatment quality score (M = 6.74, SD 2.06, range one to nine) and the quality of design and methods (M = 14.74, SD = 4.52, range three to 23). The ‘Risk of bias’ figures show the overall quality total, treatment quality and quality of design and methods. We performed a Spearman's correlation to investigate whether the total study quality, treatment quality, design quality or n at the end of treatment were correlated to the year of study. Year of publication was significantly and positively associated with total study quality (rho = 0.581, P < 0.001), design quality of the study (rho = 0.525, P < 0.01) and treatment quality of the study (rho = 0.566, P < 0.01). Treatment quality was significantly associated with design quality (rho = 0.665, P 0.05; rho = 0.066, P > 0.05; rho = 0.136, P > 0.05), respectively. When assessing all 45 analyses reported at post-treatment and follow-up, 15 showed low heterogeneity (I 2 = 25% to 50%). We analysed data in two categories. In the first, outcomes for each individual condition across all psychological therapies are analysed at post-treatment and follow-up. For the second, outcomes for each psychological therapy across all conditions at post-treatment and follow-up are presented. No analyses could be presented for gynaecological disorders or inflammatory bowel syndrome due to lack of studies meeting the inclusion criteria, and no adverse events were reported in any study reviewed. We entered two studies of children with chronic pain, containing a total of 92 participants, into an analysis of parent behaviour. The overall effect of all psychological therapies on parent behaviour was not significant (Z = 0.80, P > 0.05) (Analysis 1.1). We entered six studies of children with chronic pain, containing a total of 429 participants, into an analysis of child behaviour/disability. The overall effect of all psychological therapies on child behaviour/disability was not significant (Z = 1.39, P > 0.05) (Analysis 1.2). We entered four studies of children with chronic pain, containing a total of 356 participants, into an analysis of child mental health. The overall effect of all psychological therapies on child mental health was not significant (Z = 0.14, P > 0.05) (Analysis 1.3). We entered eight studies of children with chronic pain, containing a total of 512 participants, into an analysis of child symptoms. The overall effect of all psychological therapies on child symptoms was significant (Z = 2.23, P < 0.05) with a small effect size of standardised mean difference (SMD) -0.29 (95% confidence interval (CI) -0.55 to -0.03) (Analysis 1.4; Figure 3 ). There was only one study of children with chronic pain that could be entered into an analysis of family functioning, therefore no conclusion could be drawn. No studies presented extractable data on parent mental health. There was only one study of children with chronic pain that could be entered into an analysis of parent behaviour at follow-up, therefore no conclusions could be drawn. We entered three studies of children with chronic pain, containing a total of 289 participants, into an analysis of child behaviour/disability at follow-up. The overall effect of all psychological therapies on child behaviour/disability at follow-up was not significant (Z = 0.29, P > 0.05) (Analysis 2.1). We entered two studies of children with chronic pain, containing a total of 255 participants, into an analysis of child mental health at follow-up. The overall effect of all psychological therapies on child mental health at follow-up was not significant (Z = 0.28, P > 0.05) (Analysis 2.2). We entered six studies of children with chronic pain, containing a total of 391 participants, into an analysis of child symptoms at follow-up. The overall effect of all psychological therapies on child symptoms at follow-up was not significant (Z = 1.64, P > 0.05) (Analysis 2.3). There was only one study of children with chronic pain that could be entered into an analysis of family functioning at follow-up, therefore no conclusions could be drawn. No studies presented extractable data on parent mental health. We entered four studies of children with cancer, containing a total of 629 participants, into an analysis of parent behaviour. The overall effect of all psychological therapies on parent behaviour was not significant (Z = 1.28, P > 0.05) (Analysis 3.1). We entered five studies of children with cancer, containing a total of 706 participants, into an analysis of parent mental health. The overall effect of all psychological therapies on parent mental health was not significant (Z = 1.36, P > 0.05) (Analysis 3.2). No studies presented extractable data on child behaviour/disability, child mental health, child symptoms or family functioning. We entered four studies of children with cancer, containing a total of 597 participants, into an analysis of parent behaviour at follow-up. The overall effect of all psychological therapies on parent behaviour at follow-up was not significant (Z = 0.54, P > 0.05) (Analysis 4.1). We entered four studies of children with cancer, containing a total of 598 participants, into an analysis of parent mental health at follow-up. The overall effect of all psychological therapies on parent mental health at follow-up was not significant (Z = 1.20, P > 0.05) (Analysis 4.2). No studies presented extractable data on child behaviour/disability, child mental health, child symptoms or family functioning. There was only one study of children with diabetes that could be entered into analyses of parent mental health, therefore no conclusions could be drawn. We entered two studies of children with diabetes, containing a total of 198 participants, into an analysis of child mental health. The overall effect of all psychological therapies on child mental health was not significant (Z = 0.28, P > 0.05) (Analysis 5.1). There was only one study of children with diabetes that could be entered into analyses of child behaviour/disability, therefore no conclusions could be drawn. We entered six studies of children with diabetes, containing a total of 455 participants, into an analysis of child symptoms. The overall effect of all psychological therapies on child symptoms was not significant (Z = 1.70, P > 0.05) (Analysis 5.2). We entered four studies of children with diabetes, containing a total of 306 participants, into an analysis of family functioning. The overall effect of all psychological therapies on family functioning was not significant (Z = 0.09, P > 0.05) (Analysis 5.3). No studies presented extractable data on parent behaviour. There was only one study of children with diabetes that could be entered into an analysis of parent mental health at follow-up, therefore no conclusion could be drawn. We entered three studies of children with diabetes, containing a total of 239 participants, into an analysis of child symptoms at follow-up. The overall effect of all psychological therapies on child symptoms at follow-up was not significant (Z = 1.58, P > 0.05) (Analysis 6.1). No studies presented extractable data on parent behaviour, child behaviour/disability, child mental health or family functioning. There was only one study of children with asthma that could be entered into analyses on parent behaviour, therefore no conclusions could be drawn. We entered two studies of children with asthma, containing a total of 74 participants, into an analysis of parent mental health. The overall effect of all psychological therapies on parent mental health was not significant (Z = 0.86, P > 0.05) (Analysis 7.1). There was only one study of children with diabetes that could be entered into analyses of child behaviour/disability, therefore no conclusions could be drawn. We entered three studies of children with asthma, containing a total of 170 participants, into an analysis of child symptoms. The overall effect of all psychological therapies on child symptoms was not significant (Z = 1.51, P > 0.05) (Analysis 7.2). No studies presented extractable data on child mental health or family functioning. We entered two studies of children with asthma, containing a total of 132 participants, into an analysis of child symptoms at follow-up. The overall effect of all psychological therapies on child symptoms at follow-up was not significant (Z = 0.55, P > 0.05) (Analysis 8.1). No studies presented extractable data on parent behaviour, parent mental health, child behaviour/disability, child mental health or family functioning. We entered two studies of children with traumatic brain injury, containing a total of 72 participants, into an analysis of parent mental health. The overall effect of all psychological therapies on parent mental health was not significant (Z = 1.49, P > 0.05) (Analysis 9.1). We entered two studies of children with traumatic brain injury, containing a total of 72 participants, into an analysis of child behaviour/disability. The overall effect of all psychological therapies on child behaviour/disability was not significant (Z = 0.65, P > 0.05) (Analysis 9.2). We entered two studies of children with traumatic brain injury, containing a total of 67 participants, into an analysis of family functioning. The overall effect of all psychological therapies on family functioning was not significant (Z = 0.33, P > 0.05) (Analysis 9.3). No studies presented extractable data on parent behaviour, child mental health or child symptoms. No studies presented extractable data on parent behaviour, parent mental health, child behaviour/disability, child mental health, child symptoms or family functioning. There was only one study of children with skin diseases that could be entered into an analysis of parent behaviour, parent mental health, child behaviour and child symptoms at post-treatment, therefore no conclusions could be drawn. No studies presented extractable data on child mental health or family functioning. We entered four studies, containing a total of 166 participants, into an analysis of the effects of cognitive behavioural therapy (CBT) across all conditions on parent behaviour. The overall effect of CBT on parent behaviour was not significant (Z = 0.08, P > 0.05) (Analysis 10.1). We entered four studies, containing a total of 224 participants, into an analysis of the effects of CBT on parent mental health. The overall effect of CBT on parent mental health was not significant (Z = 1.05, P > 0.05) (Analysis 10.2). We entered seven studies, containing a total of 459 participants, into an analysis of the effects of CBT on child behaviour/disability. The overall effect of CBT on child behaviour/disability was not significant (Z = 0.84, P > 0.05) (Analysis 10.3). We entered five studies, containing a total of 439 participants, into an analysis of the effects of CBT on child mental health. The overall effect of CBT on child mental health was not significant (Z = 0.21, P > 0.05) (Analysis 10.4). We entered 11 studies, containing a total of 726 participants, into an analysis of the effects of CBT on child symptoms. The overall effect of CBT on child symptoms was significant (Z = 2.61, P < 0.05) with a small effect size of SMD -0.25 (95% CI -0.44 to -0.06) (Analysis 10.5; Figure 4 ). We entered three studies, containing a total of 211 participants, into an analysis of the effects of CBT on family functioning. The overall effect of CBT on family functioning was not significant (Z = 0.40, P > 0.05) (Analysis 10.6). We entered two studies, containing a total of 85 participants, into an analysis of the effects of CBT across all conditions on parent behaviour at follow-up. The overall effect of CBT on parent behaviour at follow-up was not significant (Z = 0.56, P > 0.05) (Analysis 11.1). We entered two studies, containing a total of 115 participants, into an analysis of the effects of CBT on parent mental health at follow-up. The overall effect of CBT on parent mental health at follow-up was not significant (Z = 1.26, P > 0.05) (Analysis 11.2). We entered three studies, containing a total of 289 participants, into an analysis of the effects of CBT on child behaviour/disability at follow-up. The overall effect of CBT on child behaviour/disability at follow-up was not significant (Z = 0.29, P > 0.05) (Analysis 11.3). We entered two studies, containing a total of 257 participants, into an analysis of the effects of CBT on child mental health at follow-up. The overall effect of CBT on child mental health at follow-up was not significant (Z = 0.27, P > 0.05) (Analysis 11.4). We entered seven studies, containing a total of 472 participants, into an analysis of the effects of CBT on child symptoms at follow-up. The overall effect of CBT on child symptoms at follow-up was not significant (Z = 1.78, P > 0.05) (Analysis 11.5). We entered two studies, containing a total of 107 participants, into an analysis of the effects of CBT on family functioning at follow-up. The overall effect of CBT on family functioning at follow-up was not significant (Z = 0.61, P > 0.05) (Analysis 11.6). There was only one study that could be entered into an analysis on the effects of family therapy (FT) across all conditions on parent behaviour, therefore no conclusions could be drawn. We entered two studies, containing a total of 74 participants, into an analysis of the effects of FT on parent mental health. The overall effect of FT on parent mental health was not significant (Z = 0.86, P > 0.05) (Analysis 12.1). We entered two studies, containing a total of 107 participants, into an analysis of the effects of FT on child behaviour/disability. The overall effect of FT on child behaviour/disability was not significant (Z = 1.44, P > 0.05) (Analysis 12.2). We entered four studies, containing a total of 202 participants, into an analysis of the effects of FT on child symptoms. The overall effect of FT on child symptoms was not significant (Z = 0.94, P > 0.05) (Analysis 12.3). We entered two studies, containing a total of 132 participants, into an analysis of the effects of FT on family functioning. The overall effect of FT on functioning was not significant (Z = 0.45, P > 0.05) (Analysis 12.4). No studies presented extractable data on child mental health. There was only one study that could be entered into an analysis on the effects of FT across all conditions on parent mental health at follow-up, therefore no conclusions could be drawn. We entered two studies, containing a total of 96 participants, into an analysis of the effects of FT on child symptoms at follow-up. The overall effect of FT on child symptoms was not significant (Z = 0.12, P > 0.05) (Analysis 13.1). No studies presented extractable data on parent behaviour, child behaviour/disability, child mental health or family functioning. We entered three studies, containing a total of 588 participants, into an analysis of the effects of problem solving therapy (PST) across all conditions on parent behaviour. The overall effect of PST on parent behaviour was significant (Z = 2.64, P < 0.05) with a small effect size of SMD -0.22 (95% CI -0.38 to -0.06) (Analysis 14.1; Figure 5 ). We entered five studies, containing a total of 660 participants, into an analysis of the effects of PST on parent mental health. The overall effect of PST on parent mental health was significant (Z = 2.14, P < 0.05) with a small effect size of SMD -0.27 (95% CI -0.53 to -0.02) (Analysis 14.2; Figure 6 ). We entered two studies, containing a total of 72 participants, into an analysis of the effects of PST on child behaviour/disability. The overall effect of PST on child behaviour/disability was not significant (Z = 0.65, P > 0.05) (Analysis 14.3). There was only one study that could be entered into an analysis on the effects of PST on child symptoms, therefore no conclusions could be drawn. We entered two studies, containing a total of 67 participants, into an analysis of the effects of PST on family functioning. The overall effect of PST on family functioning was not significant (Z = 0.33, P > 0.05) (Analysis 14.4). No studies presented extractable data on child mental health. We entered three studies, containing a total of 556 participants, into an analysis of the effects of PST on parent behaviour at follow-up. The overall effect of all psychological therapies on parent behaviour at follow-up was not significant (Z = 0.77, P > 0.05) (Analysis 15.1). We entered three studies, containing a total of 557 participants, into an analysis of the effects of PST on parent mental health at follow-up. The overall effect of all psychological therapies on parent mental health at follow-up was not significant (Z = 1.02, P > 0.05) (Analysis 15.2). There was only one study that could be entered into an analysis on the effects of PST on child symptoms at follow-up, therefore no conclusions could be drawn. No studies presented extractable data on child behaviour/disability, child mental health or family functioning. There was only one study that could be entered into an analysis on the effects of multisystemic therapy (MST) across all conditions on child mental health, therefore no conclusions could be drawn. We entered two studies, containing a total of 142 participants, into an analysis of the effects of MST on child symptoms. The overall effect of MST on child symptoms was not significant (Z = 1.81, P > 0.05) (Analysis 16.1). No studies presented extractable data on parent behaviour, parent mental health, child behaviour/disability or family functioning. There was only one study that could be entered into an analysis on the effects of MST across all conditions on child symptoms at follow-up, therefore no conclusions could be drawn. No studies presented extractable data on parent behaviour, parent mental health, child behaviour/disability, child mental health or family functioning.

More work is needed to develop and provide psychological interventions that directly target parents of children with chronic illness. Few interventions included in this review provided intensive treatment to parents that specifically targeted parent outcomes. We suggest that interventions which target specific strategies aimed at parent mental health and behaviour (e.g. problem solving skills training) are more likely to achieve those effects than interventions which include parents but do not purposefully target strategies in these outcome domains. Targeted relapse prevention strategies have not been attempted, and may be necessary to maintain treatment effects in the long term. There are relatively few studies of psychological interventions that target parents of children with a chronic illness. For example, there were no studies of children with gynaecological disorders or irritable bowel diseases that met criteria for inclusion in this review. There was also only one study of children with skin diseases that met the inclusion criteria, meaning we were not able to conduct any meta-analyses for this condition. Furthermore, studies in this area need to be conducted to a higher level of quality so that gaps in the evidence base can be filled and the effectiveness of psychological interventions for parents of children with chronic illness can be better understood. The next generation of trials should improve by taking account of the limitations identified in this review, including: larger sample sizes; following CONSORT guidelines ( Schulz 2010 ); the clearer identification of primary outcomes; designing treatment content to specifically target change in the primary outcomes; more consistency of measurement and greater consensus within the field around appropriate measure use within and across illness groups; lodging treatment manuals and data in a shared database to facilitate replication of intervention trials and re-analysis of results. This review has also highlighted several future directions for research that examines interventions targeting parents of children with chronic illness. Problem solving therapy (PST) looks particularly promising for improving parent mental health and parent behaviour. Research is needed to evaluate this intervention in populations other than cancer and traumatic brain injury (TBI), such as chronic pain. Replication studies are also needed for interventions that have been evaluated by only one research team, such as multisystemic therapy for families of children with diabetes and PST for families of children with TBI. We recognise that this goal may be difficult to achieve given the high degree of competition for funding and lack of interest among funding agencies for replication studies. Research is also needed to evaluate interaction effects such as the impact of changes in parent outcomes on child outcomes, as well as evaluation of specific treatment characteristics such as the intensity of intervention delivered to children versus parents. We also do not know anything about the effects of the interventions included in this review on fathers or siblings, which is a common critique of the field of paediatric psychology. Our final recommendation for future research in this area is in regards to duplication and piecemeal publication. Editorial policies are needed to inform authors regarding reporting standards for multiple publications from the same trial. Editors play a crucial role in creating and enforcing these policies, and need to take a pro-active approach to identifying such papers during the review process ( Committee on Publication Ethics 2011 ; World Association of Medical Editors 2012 ).

Chronic illness affects the lives of many children and their families. The prevalence of illness and disability differs by geographical and economical context. In the USA, Canada, Northern Europe, UK and Australia chronic activity-limiting conditions are reported to be frequent, with painful illness, allergy, asthma and obesity being common ( McDougall 2004 ). The changing demographic of childhood illness in economically wealthy countries has prompted a re-analysis of the role of paediatric medicine, as chronic illness becomes more prevalent than acute (e.g. Halfon 2010 ; Van Cleave 2010 ). Other parts of the world present different clinical challenges. In Africa, for example, life expectancy is 54 years and shorter in sub-Saharan Africa where almost half the population are children and the most prevalent chronic conditions are related to communicable diseases, in particular HIV-related disease, malaria and tuberculosis ( WHO 2011 ). The existing published literature shows a bias towards the medical management of chronic illness related to environment or lifestyle. Chronic pain in childhood is known to have widespread negative outcomes for children and parents ( Palermo 2000 ). Psychological intervention reviews have also been undertaken on the impact of sickle cell disease ( Anie 2012 ), recurrent abdominal pain/irritable bowel syndrome ( Huertas-Ceballos 2008 ), type 1 diabetes ( McBroom 2009 ), traumatic brain injury in children ( Soo 2007 ) and asthma ( Yorke 2009 ). The impact of childhood chronic illness on other family members, including parents, has been of growing interest for two reasons. First, it is now recognised that parents who have significant emotional distress of their own and poor family functioning can indirectly affect child outcomes ( Logan 2005 ; Palermo 2007 ). Second, it is now recognised that parents can have a positive effect on child adjustment to chronic illness ( Logan 2005 ). Addressing the mental health problems of parents, and enabling parents to be agents of change in the management of their child's chronic illness, have recently been promoted as viable treatment approaches ( Jordan 2007 ; Palermo 2009b ). Studies have focused on the education of parents about the specific condition or treatment (e.g. Savage 2011 ), whilst others concern lay- or nurse-mediated social support (e.g. Lewin 2010 ). In psychological science, specific treatment approaches have been developed that focus on reducing the emotional distress expressed by parents, or on altering parenting behaviours to promote better child outcomes, whether this be decreasing emotional distress, or improving physical symptoms or behaviour. Psychological interventions of interest are defined as any psychotherapeutic treatment specifically designed to change parent cognition or behaviour, or both, with the intention of improving child outcomes. Psychological interventions are varied in their approaches and there is still debate surrounding which treatment is most effective at improving mental health and behaviour in parents and children with chronic illnesses. Such interventions include cognitive behavioural therapy (CBT) which has been found to be effective with children with painful conditions (e.g. Eccleston 2009a ; Palermo 2009a ). Problem solving therapy (PST) has also been used with parents and children with various chronic illnesses ( D'Zurilla 1995 ; Sahler 2002 ). Other treatments have emerged from a family systems approach that focuses explicitly on the family as a unit of intervention ( Ellis 2005 ; Wysocki 2000 ) such as multisystemic therapy (MST) or family therapy (FT). There are a variety of interventions described as psychological. Cognitive and cognitive behavioural therapies dominate, but therapies with a psychodynamic or systemic tradition are also represented. Family and couple therapies have also been developed. All psychological interventions include a rationale for therapy. Common is education around illness and behaviour. Establishing the therapy and the therapist as credible is an important general stage ( Nock 2001 ). Next, a therapeutic relationship is established that will enable a confidential, non-blaming investigation of behaviour. Then, depending on the illness and behavioural presentation, specific components may include anxiety management, exposure for phobic targets, problem solving skills, cognitive therapy for depression and relationship management. Finally, most treatments will include a maintenance component that focuses on robust behavioural change within a normal home environment outside the clinic, over time. Such components can be seen in parent interventions using different therapies to improve parental functioning, child behaviour and mental health. Cognitive behavioural interventions specifically are based on a number of foundational assumptions. First, behaviour is socially and historically contingent ( Skinner 1953 ). Second, cognition is an emergent property of behavioural context ( James 1980 ). Third, behaviour is regulated by cognitive goals ( Bandura 1989 ). Fourth, emotions influence both behaviour and cognition ( Ashby 1999 ; Gilliom 2002 ). Fifth, most behaviour is deployed outside of conscious awareness or control ( Bargh 2008 ). Finally, some attempts to control cognition and behaviour can have paradoxical negative effects on desired outcomes ( Wegner 1994 ). Other interventions such as PST ( D'Zurilla 1995 ) provide a specific framework that includes positive problem orientation towards an issue. Cognitive-behavioural strategies are used in PST and include the following steps: Identify the problem, Define your options, Evaluate your options, Act, and finally See if it worked. PST has previously been effective with depression, anxiety and stress-related syndromes ( D'Zurilla 1999 ). Family and systemic therapies specifically focus on a contextual and relational view of the aetiology and maintenance of behaviour. In particular, the target of health behaviour change is typically related to family functioning, or in the cognitive representation of the family, rather than on individual attitudes, beliefs or behaviour. Typically, family or systems therapy approaches will include multiple family members and outcomes are often expressed on behalf of the family or dyad (two individuals regarded as a pair). The prevalence of childhood chronic illness has more than doubled in the last 20 years ( Perrin 2007 ). Parents provide a major influence in children's lives, influence that can have both a positive or negative effect on child outcomes. Psychological interventions are being developed that focus on helping parents to help both themselves and their children. Establishing the evidence at this stage of development will provide comment on current best practice, and serve to guide new treatment development.

There were three objectives of this review. First, results show that only problem solving therapy (PST) interventions that include parents of children with chronic conditions are effective in reducing the distress (improving mental health and behaviour) associated with parenting a child with a chronic illness. Second, cognitive behavioural therapy (CBT) is effective at reducing the primary symptom of a child experiencing chronic illness, in particular chronic pain. Third, we were unable to assess adverse events of interventions for the 14 chronic conditions. Parents are commonly included in the psychological treatment of children with chronic illness. Many psychological treatments do more than simply include parents, rather they actively focus on them, aiming to help parents improve their own coping, their ability to improve their child's coping, or both. We included 35 randomised controlled trials (RCTs) involving a total of 2723 primary trial participants. Over a third of the studies (n = 12) included in this review investigated conditions in which pain was the primary complaint. A further nine investigated diabetes, six examined cancer patients, four examined children with asthma, three trials investigated children with a traumatic brain injury (TBI) and one trial investigated eczema. There were no RCTs that met the inclusion criteria for gynaecological disorders or inflammatory bowel disease. The majority of studies could be classified within one of four broad treatment approaches: CBT, family therapy (FT), PST and multisystemic therapy (MST). The largest evidence base is of 19 studies in CBT, 18 of which had data that were available for extraction. We are currently able to draw few conclusions about PST, which had seven studies available of which six were included in our analyses. We are unable to draw any conclusions about FT and MST. FT had seven studies available, three of which were included in our analyses, and MST had two studies available, both of which were included in our analyses. Other psychotherapeutic approaches with parents and families have been discussed (e.g. Shapiro 2003 ) but we could find no studies or evaluations. There were a number of analyses which could not be run due to missing data, either because no study measured the selected outcome or because we were unable to extract the data from the study. This reflects the status of this developing field that has not yet met a consensus of agreed scales and questionnaires to measure relevant outcomes. First, we analysed data by each medical condition across all treatment classes, giving 72 possible analyses. There were no effects for follow-up data, leaving 36 possible analyses ( Table 3 ). For 22 of the 36 analyses, there were insufficient data to attempt a meta-analysis and so the findings are unknown (i.e. one or no studies available within a given outcome domain). Six analyses should be interpreted with caution because the total number of studies entering the meta-analysis was two. However, we have included these six analyses in this review for transparency; all six had no effect. Of the remaining eight analyses, there was one significant finding. Psychological therapies with a focus on parents were found to significantly improve child symptom reporting for painful conditions. There were no other effects of parent-focused treatment in any other condition for any other outcome that could be analysed. Second, we analysed data by each treatment class across all medical conditions, giving 48 possible analyses. There were no effects for follow-up data, leaving 24 possible analyses ( Table 3 ). For nine of the 24 analyses, there were insufficient data to attempt a meta-analysis and so the findings are unknown (i.e. one or no studies available within a given outcome domain). Six analyses were inconclusive because the total number of studies entering the meta-analysis was two. However, we have included these six analyses in this review for transparency; all six had no effect. Of the remaining nine analyses, there were three significant findings. CBT had a significant effect on child symptom reporting, and PST had a positive effect on parent behaviour and on parent mental health outcomes. We did not present data in the form of numbers needed to treat because of the limited number of effects identified; therefore, presenting continuous data in a categorical format would not have been useful. None of the significant effects were strong and these results could be strengthened or overturned with additional trials; therefore we did not calculate a failsafe N. Furthermore, it was not possible to conduct subgroup analyses regarding comparisons of parent-only interventions versus family-based interventions, intervention effects within specific illnesses, and intervention effects across specific types of psychological interventions due to the small number of trials. The overall study quality was adequate. However, the field continues to be hampered by the common practice of short and limited descriptions of treatment content, the insufficient reporting of results and a reliance on small samples. Analysis of this evidence presented a number of challenges. First, multiple measurement tools within a given domain are often employed in individual studies, and there is little agreement as to the preferred measurement tool across studies. In some cases measurement is relatively homogenous (e.g. pain intensity) whereas in others there is greater variety (e.g. family functioning scales in diabetes). These trials do not routinely a priori identify the primary outcome, and there is unusual variety of outcome reporting. For example, one study discussed parent judgement of child outcome when the more valid measure, but non-significant finding, of child report was available ( Levy 2010 ). A posteriori selection of outcome measures is a significant problem in this field. As per our protocol we were uninfluenced by the primary reporting of measures and focused on the best measure available in each domain. This field needs to take account of reporting biases and establish standards to improve the reporting of a priori decisions regarding measurement. Second, we attempted to review evidence of trials with a dominant parent intervention component. This meant we were inevitably going to combine trials with varying amounts of parenting content. Although we planned subgroup analyses, the data were not of sufficient quantity and quality to enable such an investigation. For some analyses we combined studies that were designed specifically with parents as the sole focus, and in others they were part of a combined treatment. Further, the philosophy of some treatments (e.g. MST) was antithetical to our strategy of determining an individual as a treatment target, however, we included them in this study. It should be noted that significant findings in this review emerged when there was homogeneity of approach, homogeneity of outcome measurement and a larger n. Third, it should be noted that we had some difficulties in data retrieval due to incomplete and partial data reporting. Data were sometimes reported graphically, and ns, means and/or standard deviations were often missing. We wrote to all 31 first authors an average of two emails. Complete outcome data (i.e. sample size, means, standard deviations) were available from the published paper in 13 trials ( Barakat 2010 ; Connelly 2006 ; Hoekstra-Weebers 1998 ; Kashikar-Zuck 2005 ; Kashikar-Zuck 2012 ; Laffel 2003 ; Ng 2008 ; Palermo 2009 ; Seid 2010 ; Stehl 2009 ; Wade 2006 ; Wade 2006b ; Wade 2011 ). Seven authors provided data in response to our requests ( Ambrosino 2008 ; Askins 2009 ; Celano 2012 ; Levy 2010 ; Niebel 2000 ; Sahler 2002 ; Sahler 2005 ). Other authors were unable or unwilling to provide additional data or did not respond. The non-production of data is a problem in science ( Data's shameful neglect 2009 ), and has been particularly discussed in psychology ( Wicherts 2006 ; Wicherts 2011 ). We support the general move toward central registries for all trial data. Fourth, piecemeal and repeat publication was found in five cases where multiple manuscripts were published from the same trial. In particular, one study ( Ellis 2005 ) was reported six times in five different journals while another trial ( Wysocki 1999 ) was reported five times in four different journals, with variable citation of previous publications in later publications. Such practices are unhelpful, create confusion and increase unnecessary labour ( American Psychological Association 2011 ). Many journals now have policies regarding publication of multiple manuscripts from the same trial, including a detailed description of previous publications from that trial and a statement regarding the unique contribution of the present manuscript (e.g. Drotar 2010 ). Finally, replication by other research teams independent to the therapy progenitors is uncommon. For example, Ellis and colleagues are the only group who have evaluated MST in young people with diabetes ( Ellis 2004 ; Ellis 2005 ). Similarly, PST for children and adolescents with TBI has not been evaluated by any research team outside of Wade and colleagues ( Wade 2006 ; Wade 2006b ; Wade 2011 ). Finally, some therapy approaches have been used exclusively within an illness group. Most notably, CBT was the only intervention evaluated for children with chronic pain. This review was limited to the analysis of 14 conditions. Other studies in other conditions may be instructive. As is common practice within the Cochrane Pain, Palliative and Supportive Care Group (PaPaS), we did not searched grey literature. It is always possible that trials of parent-focused interventions were undertaken but unreported in peer-reviewed publications. We consider it unlikely that any such trials exist in the grey literature but this should be acknowledged. Only RCTs were included in this review. However, therapists were not blind to the therapy being delivered. Bias is most likely due to small sample sizes. Unpublished studies are always possible but unlikely given that there appear to be few barriers to publishing small, negative or poor quality studies. Bias in the field may be due largely to the lack of available studies. Only a handful of reviews have also investigated psychological interventions for parents of children with a chronic illness. Our results are consistent with a previous meta-analysis regarding the effectiveness of CBT in reducing child symptoms in young people with chronic pain ( Eccleston 2009a ). Our results were somewhat consistent with a meta-analysis of psychological paediatric oncology interventions, which showed no effects on child behaviour or child mental health but positive effects for parent mental health and parent behaviour ( Pai 2006 ). Our results are not consistent with previous reviews of psychological interventions that included parents of children with diabetes, which reported positive effects on child symptoms and family functioning ( Armour 2005 ; Grey 2000 ; Harris 2010 ; McBroom 2009 ). Previous reviews of psychological interventions that included parents of children with asthma or skin diseases were inconclusive due to lack of trials that met inclusion criteria ( Ersser 2007 ; Yorke 2009 ). Notably, disagreements between the present meta-analysis and previous reviews may be attributed to differences in inclusion criteria, selection of outcome measures and/or selection of comparator group. One prior review indicated that psychological interventions which included coping skills training for adolescents and young adults with chronic illness (cancer, diabetes, juvenile idiopathic arthritis, sickle cell disease and asthma) and their parents/families had mixed effects on child psychosocial functioning and family functioning ( Sansom-Daly 2011 ). We were unable to find any previous reviews that compared results from individual psychological therapies across chronic illness conditions for parent outcomes or child symptoms. Therefore, we cannot draw conclusions regarding consistency of our results by treatment type for parent outcomes or child symptoms with previous reviews.

To evaluate the effectiveness of psychological parent interventions on reducing the distress associated with parenting a child with a chronic illness. To evaluate the effectiveness of psychological parent interventions on reducing the primary symptom or behavioural expression of illness for the child. To assess primary outcomes and adverse events of different parent interventions in the 14 different conditions (see ‘Types of participants’).