Types of studies

Randomised controlled trials (RCTs).

Types of participants

Adult palliative care patients (aged 18 years or older) whose symptoms of anxiety were described by the trial authors as beyond what could be seen as normal in this patient group. This could be captured as a score equivalent to clinically significant symptoms on a validated scale, such as the Hamilton Anxiety Rating Scale (HAM‐A) (Hamilton 1959).

We included trials where anxiety was described as a disorder such as adjustment disorder, obsessive‐compulsive disorder, phobia, panic disorder, post‐traumatic stress disorder or GAD. We sought to verify that symptoms reported as clinically significant were measured using validated scales, and if the population was described as having an anxiety disorder (specifically GAD, substance‐induced anxiety, adjustment disorders, obsessive‐compulsive disorders, specific phobias such as in response to medical interventions, panic disorders and post‐traumatic stress disorder) that it was defined using the International Classification of Disease (ICD) (ICD‐10 2010) or the Diagnostic and Statistical Manual of Mental Disorders (DSM‐III 1980; DSM‐R 1987; DSM‐IV 1994; DSM‐IV‐TR 2000; DSM‐5 2013).

For the purposes of the review, we defined palliative care patients as people with a progressive, life‐limiting illness, no longer responsive to disease‐modifying treatments who were thought to be in the last year (or so) of life and were, or would be, eligible to receive palliative care. We did not include studies in which participants were in the last 24 to 48 hours of life. At this time, symptoms of anxiety may also be in part a manifestation of irreversible processes such as multiple organ failure, and treatment may differ from that considered in earlier phases of the disorder.

Types of interventions

Interventions for anxiety included any type of drug therapy, for example, 5‐HT3 receptor antagonists, anxiolytic agents, antidepressive agents, antipsychotic and atypical antipsychotic agents, benzodiazepines, butyrophenones, phenothiazines, antihistamines, barbiturates, sedative hypnotics, antiepileptic drugs, ketamine and beta‐blockers. We did not include any non‐conventional drugs, such as herbal medicines.

Comparator treatments included placebo; another drug therapy or different dose schedule; or a non‐drug intervention such as counselling, CBT or relaxation therapies.

Types of outcome measures

Electronic searches

To identify studies for inclusion in this review, we developed detailed search strategies for each electronic database searched. These were based on the search strategy developed for MEDLINE but revised appropriately for each database. The search strategy used in the original review run in 2003 is reported in Appendix 1 and the search strategies used in the first update of 2012 are presented in Appendix 2, Appendix 3, Appendix 4, Appendix 5, Appendix 6, and Appendix 7. The search strategies used in 2016 for the second update of this review are reported in Appendix 8. We also updated the list of drugs included in the search and re‐ran the search for these for all years to May 2016 (see Appendix 9).

Searching other resources

We searched the following trials and pharmaceutical industry trials registers.

Selection of studies

In this update, two review authors (SS and CM) independently assessed titles and abstracts for eligibility for inclusion in the review. We sought full‐text reports of all potentially relevant studies remaining after the initial assessment and two review authors (SS and CM) independently assessed these against our predefined inclusion criteria. We resolved any disagreements between the review authors by consulting a third review author (NP). Where we identified posters or conference abstracts which we considered potentially relevant, we sought full‐text reports of the study and, if unsuccessful, we contacted study authors to seek further information. Where English translations for studies published in another language were not available at the screening stage, we obtained full‐text reports and translated these initially using an electronic translator. We reported reasons for excluding full‐text reports. See Characteristics of excluded studies table.

Data extraction and management

We designed a data extraction form to collect the following data:

• publication details;

• study eligibility criteria;

• study details (e.g. aim, start and end date, ethics approval);

• participant characteristics (e.g. number of participants, age, sex, diagnosis of anxiety, study setting);

• description of intervention and comparator (e.g. duration of treatment, timing, delivery, number randomised to groups);

• outcome details (e.g. instrument used to evaluate anxiety, time points when outcomes were assessed, withdrawals).

Assessment of risk of bias in included studies

We planned for two review authors (SS and CM) to independently assess risk of bias for each study, using the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011), with any disagreements resolved by discussion. We planned to complete a 'Risk of bias' table for each included study using the 'Risk of bias' tool in Review Manager 5 (RevMan 2014).

We aimed to assess the following for each study.

• Random sequence generation (checking for possible selection bias). We planned to assess the method used to generate the allocation sequence as: low risk of bias (any truly random process, e.g. random number table; computer random number generator); unclear risk of bias (method used to generate sequence not clearly stated). We planned to exclude studies using a non‐random process (e.g. odd or even date of birth; hospital or clinic record number).

• Allocation concealment (checking for possible selection bias). The method used to conceal allocation to interventions prior to assignment determines whether intervention allocation could have been foreseen in advance of, or during, recruitment, or changed after assignment. We planned to assess the methods as: low risk of bias (e.g. telephone or central randomisation; consecutively numbered sealed opaque envelopes); unclear risk of bias (method not clearly stated). We aimed to exclude studies that did not conceal allocation (e.g. open list).

• Blinding of participants and personnel (checking for possible performance bias). We planned to assess the methods used to blind study participants and personnel from knowledge of which intervention a participant received. We aimed to assess methods as: low risk of bias (study stated that it was blinded and described the method used to achieve blinding, such as identical tablets matched in appearance or smell, or a double‐dummy technique); unclear risk of bias (study stated that it was blinded but did not provide an adequate description of how it was achieved). Studies that were not double‐blind were considered to have high risk of bias.

• Blinding of outcome assessment (checking for possible detection bias). We aimed to assess the methods used to blind study participants and outcome assessors from knowledge of which intervention a participant received. We planned to assess the methods as: low risk of bias (study had a clear statement that outcome assessors were unaware of treatment allocation, and ideally described how this was achieved); unclear risk of bias (study stated that outcome assessors were blind to treatment allocation but lacked a clear statement on how it was achieved). We considered studies where outcome assessment was not blinded as having a high risk of bias.

• Selective reporting (checking for reporting bias). We aimed to assess whether primary and secondary outcome measures were prespecified and whether these were consistent with those reported. We aimed to assess selective reporting as: low risk of bias (studies reported primary and secondary outcomes); high risk of bias (not all prespecified outcomes reported or only for certain data collection time points).

• Incomplete outcome data (checking for possible attrition bias due to the amount, nature and handling of incomplete outcome data). We planned to assess the methods used to deal with incomplete data as: low risk of bias (less than 10% of participants did not complete the study or used 'baseline observation carried forward' analysis), or both; unclear risk of bias (used 'last observation carried forward' analysis); high risk of bias (used 'completer' analysis).

• Size of study (checking for possible biases confounded by small size). We aimed to assess studies as being at low risk of bias (200 participants or greater per treatment arm); unclear risk of bias (50 to 199 participants per treatment arm); high risk of bias (fewer than 50 participants per treatment arm)

Measures of treatment effect

We planned to analyse dichotomous data as odds ratios and 95% confidence intervals (CI) and continuous data as mean difference and 95% CI.

Unit of analysis issues

For any RCTs using a cross‐over design, we planned to use only data from the first comparative phase prior to cross‐over. This decision was based on the possibility of a carry‐over of treatment effect from the drug evaluation or a comparative treatment.

Dealing with missing data

For this review, we expected a significant loss to follow‐up due to participants' declining health. We planned to report trial attrition rates in the 'Risk of bias' table. This would have included, if available, reasons for attrition per treatment arm. Where study data were missing but might have been available, we planned to contact the authors to obtain missing outcome data where possible. We planned not to exclude trials on the basis of missing data.

Assessment of heterogeneity

If meta‐analysis had been possible, we planned to use the I² statistic to assess heterogeneity among the trials in each analysis. If we identified substantial heterogeneity (i.e. I² greater than 50%), we aimed to report it and explore possible causes by performing prespecified subgroup analysis.

Assessment of reporting biases

If meta‐analysis had been possible using 10 or more studies, we planned to explore publication bias using Egger's test and by inspection of funnel plots for symmetry.

Data synthesis

We planned to combine study data to provide a pooled effect estimate using a fixed‐effect model. If we had found no substantial heterogeneity, we planned to use a random‐effects model to check the robustness of the fixed‐effect model. If substantial statistical heterogeneity had been observed, we would have used the random‐effects model a priori.

If we had found studies that reported a mixture of change‐from‐baseline and final value scores, we would have only combined data if the studies reported the outcome using the same measurement scale.

Where there were insufficient studies to undertake a meta‐analysis, we planned to combine individual studies in a narrative review.

Subgroup analysis and investigation of heterogeneity

If three or more studies reported relevant data, we planned to perform the following subgroup analyses:

• men versus women;

• mild or moderate anxiety versus severe anxiety;

• follow‐up no greater than one month versus follow‐up greater than one month.

Sensitivity analysis

We planned to perform sensitivity analyses to explore the influence of the following factors:

• excluding unpublished studies;

• excluding studies considered at high risk of selection bias in terms of adequate allocation concealment, detection bias in terms of blinded outcome assessment and attrition bias due to follow‐up of less than 80% of participants in each arm;

• excluding studies using the following filters:

  • industry funded;

  • non‐validated scales used for measuring effect;

  • non‐validated diagnostic criteria.