Venous access devices for the delivery of long-term chemotherapy: the CAVA three-arm RCT

Original sites

• Beatson West of Scotland Cancer Centre (BWoSCC), Glasgow (lead centre).

• Queen Elizabeth Hospital Birmingham, Birmingham.

• Christie Hospital, Manchester.

• Freeman Hospital, Newcastle upon Tyne.

• Guy’s Hospital, London.

• St James’s University Hospital, Leeds.

Additional sites

• University Hospital of North Durham, Durham.

• Darlington Memorial Hospital, Darlington.

• Northampton General Hospital, Northampton.

• Weston Park Hospital, Sheffield.

• Charing Cross Hospital, London.

• Royal United Hospital, Bath.

• Forth Valley Royal Hospital, Larbert.

• Cumberland Infirmary, Carlisle.

• Kent and Canterbury Hospital, Canterbury.

• Royal Cornwall Hospital, Treliske.

• Broomfield Hospital, Chelmsford.

• Sunderland Royal Hospital, Sunderland.

• Western General Hospital, Edinburgh.

Subcutaneously tunnelled central catheter (Hickman)

Introduced in 1979, subcutaneously tunnelled central catheters are commonly known as Hickman and consist of a thin plastic tube that is inserted into a central vein in the neck or upper chest region. It is ‘tunnelled’ under the skin for a few centimetres before exiting and has a Dacron™ cuff (DuPont de Nemours, Inc., Wilmington, DE, USA) to improve stability and minimise the risk of infection. Ultrasound is used to target the vein. These catheters are inserted by a variety of specialists (nurse practitioners, interventional radiologists, anaesthetists and surgeons) but, currently, nursing experience in placement is limited. Maintenance involves regular dressing change and weekly line flushing. The cost of a Hickman lies between those of PICCs and PORTs. Removal requires simple dissection to free the Dacron cuff.

Peripherally inserted central catheters

Introduced in 1975, a PICC line is a thin plastic tube that is inserted into a peripheral vein in the upper arm. These catheters are commonly inserted by nurse practitioners, but also by interventional radiologists. Ultrasound is used to target the vein. Maintenance involves regular dressing change and weekly line flushing. It is the cheapest device to purchase and the simplest device to place. When it is no longer required, removal of the line is straightforward.

Implantable chest wall PORTs

Introduced in 1981, a chest wall port is a small, coin-sized device with a silicone membrane that is buried just under the skin in a subcutaneous pocket. It connects to a thin plastic tube similar to the other two devices. The entire device is completely implanted with nothing exiting the skin. Ultrasound is used to target the vein. The PORT has to be punctured via the skin with a special needle each time it is used. PORTs are inserted by a variety of specialists (nurse practitioners, interventional radiologists, anaesthetists and surgeons) but, currently, nursing experience in placement is limited. Given that PORTs are totally implanted, there is no dressing requirement and flushing is needed only monthly. PORTs are the most expensive of the three devices and the most complicated to insert and remove, requiring a minor surgical procedure.

Primary outcome

The primary outcome was complication rate, a composite of the inability to aspirate blood, infection associated with the device (suspected, confirmed or exit site), venous thrombosis, pulmonary embolism related to the device, mechanical failure (line fracture, line separation from chest wall port, exposure of line cuff, exposure of chest wall PORT or breakdown of wound, chest wall port flip, line fallen out or line migration requiring intervention) and other complications.

Secondary outcome

The secondary outcome measures were:

• Incidence of individual complications – inability to aspirate blood from the device, venous thrombosis related to the device, pulmonary embolism related to the device, laboratory-confirmed bloodstream infection, suspected catheter-related bloodstream infection, exit site infection, mechanical failure and other complications.

• Complications per catheter-week – defined as the number of complications divided by the number of weeks that the device was in place.

• Time to first complication – defined as the time from randomisation to the first documented complication. Patients not experiencing a complication were censored at their device removal date or last available date (last chemotherapy date, last status assessment date or date of death) if the device was still in place at the end of the study.

• Duration of chemotherapy treatment interruptions – overall and for each individual complication.

• EuroQoL-5 Dimensions, three-level version (EQ-5D-3L) – a validated, generic, health-related quality-of-life (HRQoL) measure comprising five dimensions (i.e. mobility, self-care, usual activities, pain/discomfort and anxiety/depression). A value of 1 indicates perfect health, whereas a value of –0.594 indicates the worst health state (worse than dead). The visual analogue score for general health is reported on a scale of 0–100 (0 being the worst imaginable health state and 100 being the best).

• European Organisation for Research and Treatment of Cancer (EORTC) Quality of Life Questionnaire C30 (QLQ-C30) – comprising scores for the five functional scales (i.e. physical, role, emotional, cognitive and social), nine symptom scales (i.e. fatigue, nausea and vomiting, pain, dyspnoea, insomnia, appetite loss, constipation, diarrhoea and financial difficulties) and a global health status score, all on a scale of 0–100 (a high score represents a higher response level).

• Venous Access Device Questionnaire (see Report Supplementary Material 1) – comprising 16 different questions (relating to continuing to drive, getting in and out of a car, using public transport, going shopping, eating, hygiene, sleeping, mobility, usual work, exercise, hobbies, feeling of self-consciousness). These items are assessed on a scale of 1–4 (not at all, a little, quite a bit, very much; as piloted in the Glasgow feasibility study). 3

Hypothesis 1: peripherally inserted central catheters are non-inferior to Hickman in terms of complication rate

Based on the assumption that the Hickman complication rate is 55%, PICCs would be considered non-inferior if their complication rate is no more than 10% higher (i.e. 65%). To rule out this difference with 80% power, and a one-sided significance level of 2.5% required 778 patients in total using a 1 : 1 randomisation.

Hypothesis 2: PORTs have a lower complication rate and are more cost-effective than Hickman

The minimum requirement here was to demonstrate that PORTs have a lower complication rate than Hickman. Based on the assumption that the Hickman complication rate is 55%, we aimed to detect at least a 15% reduction with PORTs. To detect this reduction with 95% power, and a two-sided significance level of 5% required 550 patients in total using a 1 : 1 randomisation.

Hypothesis 3: PORTs have a lower complication rate and are more cost-effective than peripherally inserted central catheters

The minimum requirement here was to demonstrate that PORTs have a lower complication rate than PICCs. Based on the assumption that the PICCs complication rate is 55%, we aimed to detect at least a 15% reduction with PORTs. To detect this reduction with 80% power, and a two-sided significance level of 5% required 341 patients in total using a 1 : 1 randomisation.

Other amendments

• In the protocol, the definition of the primary end point was a composite of infection associated with the device (suspected or confirmed) and/or mechanical failure only. This was amended as more complications were specified as the study developed. It was an error that the protocol was not updated accordingly.

• The protocol incorrectly stated that the power for the PICCs versus Hickman comparison was intended to be 90%; this should have been corrected to state 80%.

• An analysis was planned based on complication event rate data13 to estimate the relative effect of the devices on infections (laboratory-confirmed, suspected catheter related and exit site infections) versus non-infections (all other complications combined) to allow an assessment of the similarity of these effects over time via a likelihood ratio test. The test was to be conducted by comparing the likelihood of a standard joint frailty model [modelling complication (no distinction of type) and device removal rates simultaneously] with the likelihood of a multivariate joint frailty model [modelling the two complication types (infection and non-infection) and device removal rates simultaneously]. Unfortunately, no software could be found to apply the multivariate joint frailty model successfully to the CAVA trial data and the analysis was abandoned.

Resource use and costs

All health-care resource use data were collected by a research nurse involved in the delivery of care to patients in the CAVA trial. These include:

• Procedure details relating to device insertion – staffing team (nurse, radiographer, anaesthesiologist, radiologist, doctor and surgeon), setting (theatre, treatment room, radiology department and bedside), type of anaesthesia and type of imaging. These components formed the basis of the intervention costs.

• Device removal – device removal for any reason during the trial period.

• Health-care resource use during follow-up period – number of admissions and length of stay, and outpatient visits. These were recorded monthly during the trial and formed the basis of the non-intervention costs.

Unit costs were based on national sources (Table 1). Where costs were not routinely available, resource use and unit costs were estimated through consultation with clinical experts. All unit costs were presented in Great British pounds for the price year 2017/18.

The total cost per patient was calculated by attaching relevant unit costs to the resource use data. In addition to the cost of the device, the cost associated with device insertion was calculated. This included the cost of the primary operator and any additional staff required for insertion, setting, imaging, anaesthesia and use of prophylactic antibiotics. Unit costs associated with staff and setting were applied to the time taken for device insertion based on data recorded within the trial. Similarly, unit costs for imaging, anaesthesia and the use of prophylactic antibiotics were applied to resource use data from the trial. The cost of device removal took into account staff time required to perform the procedure. The time required to remove each device was obtained from two separate clinical experts. Unplanned follow-up costs, including inpatient stays in hospital and outpatient visits recorded following device insertion (which were not part of the patient’s ongoing chemotherapy care), were collected. For the purposes of the analysis, device insertion costs (the device itself and device insertion costs) and unplanned follow-up costs (inpatient stay, outpatient visits, device removal and replacement costs) were also estimated alongside total costs.

Health-related quality of life

In accordance with NICE guidance on estimating cost-effectiveness,20 health outcomes were expressed as quality-adjusted life-years (QALYs) gained. Data on HRQoL were collected using the EQ-5D-3L at baseline, and monthly thereafter until device removal, death or the end of follow-up at 12 months. The EQ-5D-3L assesses five health domains (i.e. mobility, self-care, usual activities, pain/discomfort and anxiety/depression) according to three levels (i.e. no problems, some problems and extreme difficulty). Responses to EQ-5D-3L were converted into a single global measure of health utility for each time point. 21 For patients who had the device removed within the trial period, we did not collect data subsequent to device removal. Therefore, we estimated the quality-adjusted time spent on device per patient as a proxy for QALY gained. A baseline-adjusted AUC approach was used to estimate the QALYs gained in the trial, while adjusting for the life-years gained by the HRQoL experienced over the study period. 22

Missing data

Resource use data comprised data on inpatient stays and outpatient visits during the follow-up period. It was assumed that where a patient had no record of outpatient visits or inpatient admissions during the follow-up period, none had taken place. Therefore, we included resource use data only where they were reported and did not impute for missing resource use data.

Multiple imputation by chained equations was used to impute missing EQ-5D health–utility values. 23 For patients with missing values, health–utility values from all other available time points were used to predict missing values,24 while adjusting for the trial stratification factors (age, sex, BMI, device history and treatment arm).

Cost-effectiveness analysis

The cost-effectiveness analysis was based on the imputed ITT patient population. The mean total costs and QALYs were estimated by fitting generalised linear models to the data and adjusting for age, sex, BMI, device history, treatment arm and baseline EQ-5D (QALY estimates only). The appropriate family for the generalised linear model was selected based on the results of the modified Park’s test. The final cost model was based on the log-link and gamma family; the final QALY model was based on the identity link and Gauss family. Based on the estimation of the final statistical model, the total cost and QALY difference between arms were based on the marginal prediction. All cost and QALY calculations were conducted in Stata^® 14 (StataCorp LP, College Station, TX, USA).

Non-parametric bootstrapping was used to explore uncertainty in our estimates of mean cost and QALYs and to describe how this uncertainty affects the model outcomes. A 1000-iteration bootstrap was undertaken to estimate the 95% confidence intervals (CIs) for mean cost and QALYs for each device. The resultant distribution of mean costs and QALYs was then presented graphically on the cost-effectiveness plane.

Where appropriate, cost-effectiveness was expressed as incremental cost-effectiveness ratios (ICERs), calculated as the difference between devices in mean total costs divided by mean total QALYs. In addition, cost-effectiveness was also expressed as net monetary benefit (NMB) based on a willingness-to-pay threshold of £20,000. The NMB is a measure of the health benefit, expressed in monetary terms:

where E = effectiveness, WTP = willingness-to-pay threshold and C = cost.

When comparing two devices, a positive incremental NMB indicates that the device of interest is cost-effective compared with the alternative. Cost-effectiveness acceptability curves were used to present the uncertainty in the decision regarding the most cost-effective option over a variety of monetary willingness-to-pay thresholds.

Sensitivity and scenario analyses

Several sensitivity and scenario analyses were undertaken to test the robustness of the cost-effectiveness estimates. These included:

• Exploring the results at 3 and 6 months. There have been suggestions from the literature and from clinical practice that device-related complications typically occur within the first 6 months of device placement. The time to first complication was also investigated in the CAVA trial. A sensitivity analysis was conducted by restricting the analysis to 3 and 6 months of follow-up.

• Exploring the results of patients with solid tumours only. The CAVA trial population included patients with solid tumours and haematological malignancies. There have been suggestions from the literature and from clinical practice that these patients are managed differently. This has also been reflected by the small number of patients with haematological malignancies being recruited into the trial. A sensitivity analysis was carried out by restricting the analysis to patients with solid tumours only.

• Exploring a nurse-led model of care. There are variations across clinical practice in how care is provided for patients requiring venous access devices for chemotherapy, specifically with regard to device implantation, removal and replacement of PORTs. Where Hickman or PORTs devices are required, these are typically inserted in a theatre or radiology department by a radiologist. The associated resource requirements are greater than those for a PICC device. Furthermore, waiting times for theatre space and radiologist availability may also increase the time taken to start a patient on treatment. Personal communication with clinicians currently involved in the routine insertion of PORTs suggested that this procedure can be performed by a nurse-led service. A sensitivity analysis was carried out to examine this scenario. For such a nurse-led service, we assumed that three nurses would be required and that the procedure was undertaken in a treatment room (instead of in an ‘air handled’ room). We also included the cost of regular device maintenance required for Hickman and PICCs, delivered by a nurse in the community, which is currently standard practice but was not captured within the clinical trial. This involved a 30-minute visit per week from a district nurse.

Sampling and recruitment

Following ethics approval of the study protocol, patients were approached in chemotherapy clinics and day care units and were invited to participate in the study. A patient information leaflet was sent out by post or handed out to interested patients. The research team followed up the initial contact with a telephone call to discuss any queries and to recruit participants. Initial plans to hold one focus group of 10 patients were modified when recruitment proved more difficult than anticipated. Barriers to recruitment included ill health, treatment schedules, travel issues, work commitments and a reluctance to join a group discussion. Therefore, following consultation with the CAVA project management group, the decision was taken to carry out three separate focus groups. A convenience sample of nine patients receiving treatment for cancer at the lead centre (BWoSCC, Glasgow) took part in three separate focus groups, each with three participants. Four male and five female patients participated. Participants’ ages ranged from 48 to 66 years. All participants had solid tumours, including two with metastatic disease. Three participants had experience of an implanted PORT, three participants had experience of a PICC line and two participants had experience of a tunnelled cuffed central catheter. One participant had experience of both a PICC and a tunnelled cuffed central catheter. Four participants had previously participated in a RCT.

Clinical staff were also invited to be interviewed because it was anticipated that there would be variations in attitudes regarding equipoise across specialties and roles, and differences in local practice, which may act as facilitators of and barriers to randomisation. One-to-one, semistructured interviews were conducted with 23 clinical staff (five nurse managers, four research nurses, four oncologists, six interventional radiologists, three haematologists and one anaesthetist) from six different centres (BWoSCC, Glasgow; Queen Elizabeth Hospital Birmingham, Birmingham; St James’s University Hospital, Leeds; Christie Hospital, Manchester; Freeman Hospital, Newcastle upon Tyne; University Hospital of North Durham, Durham). The research team made initial contact with the relevant individuals via the local principal investigator at each site and clinical staff information sheets were e-mailed to potential participants.

Procedure

Focus groups with patients took place in a private room at the lead centre and were facilitated by one of the authors (MS), who has received considerable training in qualitative research methodology and has a master of laws (medical law and health-care ethics), as well as a background in nursing. The moderator, who had no prior relationship with the participants, was accompanied by an assistant, but no one else was present during data collection. Signed consent to participation and to audio-recording of the discussion was obtained. A focus group guide was used (see Report Supplementary Material 2), which encompassed questions on participants’ attitudes, experiences and preferences relating to the three devices, and on participants’ understanding of the study design and willingness to participate in a randomised trial of the three devices. Learning aids included laminated photographs of the devices in situ, copies of the patient information sheet (PIS) for the trial and a demonstration of the three devices. Focus groups lasted approximately 1 hour and were digitally recorded.

The same qualitative researcher visited each centre to interview staff. Written consent to participation and to audio-recording of the interview was obtained prior to participation. A semistructured interview guide was used (see Report Supplementary Material 3) that aimed to elicit information on perceived barriers to and facilitators of recruitment, as well as the attitudes to the three devices and to RCTs. Feedback was obtained on trial materials including PISs and consent forms. All interviews were digitally recorded.

Analysis

All recordings, along with field notes taken during data collection, were transcribed verbatim, anonymised and thematically analysed. 25 The QSR NVivo10 (QSR International, Warrington, UK) software program was used to facilitate data analysis. First, initial codes were identified, based on careful reading and re-reading of the data by two members of the research team independently. Second, these codes were sorted into potential themes. Finally, the themes were refined through repeated investigation of both similar and anomalous examples.

Sampling and recruitment

Forty-two patients enrolled in the CAVA trial participated in eight focus groups. Participants were purposively sampled from the CAVA trial’s six largest recruitment centres: BWoSCC, Glasgow; St James’s Hospital, Leeds; Christie Hospital, Manchester; Freeman Hospital, Newcastle upon Tyne; University Hospital of North Durham, Durham; and Hammersmith Hospital, London. Three further participants who had agreed to participate were too unwell to attend, one each at Leeds, Durham and London. To include a range of perspectives and experiences, participants at each site were chosen for maximum variation in terms of age, sex, cancer diagnosis and device allocation, as well as positive and negative clinical experiences with CVADs. Eligible participants were initially contacted by local trial nurses with whom they had prior contact and who provided information sheets in person or by mail.

Originally, six focus groups (one at each centre) comprising participants with a mix of all three CAVA trial devices was planned. This plan was amended for two reasons. First, one site (Christie Hospital) had ceased recruitment at the time of focus group planning and only participants with PORT devices remained on the trial. Second, it was felt that, although mixed-device groups would be well suited to comparisons between devices, single-device groups could offer greater insight into attitudes and experiences of each device. The design was amended to include four large mixed-device groups (4–11 participants) and four smaller single-device groups (two PORTs only, one Hickman only, one PICC only; two or three participants). The two additional groups were sampled at the trial’s Glasgow site, which had higher recruitment rates than the other sites.

Clinical staff (nurses, oncologists, radiologists and anaesthetists) from each of the trial’s centres were contacted via e-mail by the research team, provided with copies of the information sheet and invited to be interviewed. The decision was made to also invite trial staff from non-clinical research backgrounds to take part, as it was felt that these staff could offer useful insights, particularly regarding the organisational and administrative aspects of the trial. Twenty-six one-to-one semistructured interviews with clinical staff from 13 centres were carried out.

Procedure

Focus groups took place on site at six CAVA trial centres in quiet meeting rooms and were moderated by one of the authors (CR), a female psychologist (PhD; Doctor of Philosophy) and an experienced qualitative researcher who had no prior relationship with the participants. A trial nurse attended part of one focus group (Leeds) to address patient queries; otherwise no other persons were present. Prior to each session, details about data collection, analysis and use were discussed, and written informed consent was obtained. The moderator started by explaining her own background and role in the trial. She then reminded participants about the purpose of the broader trial and current focus group, using A4 cards depicting each device type and reiterating current clinical equipoise. A focus group guide was used to ensure that all relevant topics would be addressed (see Report Supplementary Material 2). Topics included CAVA trial participation and day-to-day experiences relating to their device. To create a communication situation resembling a naturally occurring interaction, interference with the discussion was kept to a minimum. Focus group discussions lasted approximately 1 hour and were audio-recorded with participants’ permission. To assist with transcription and analysis, relevant field notes were compiled after each focus group.

Interviews with staff took place in person or by telephone by the same researcher. Face-to-face interviews took place in interviewees’ offices or in quiet rooms on site. As above, a semistructured interview guide was used to facilitate data collection (see Report Supplementary Material 3). Topics covered included the interviewee’s knowledge, experiences and opinions regarding the use of CVADs in the context of anticancer treatment, as well as their experience of participating in the CAVA trial. Interviews lasted, on average, 30 minutes and were audio-recorded using a digital voice recorder.

Analysis

All recordings, along with field notes taken during data collection, were transcribed verbatim, anonymised and uploaded to the QSR NVivo 10 qualitative software program. Data were analysed using thematic analysis, ‘a method for identifying, analysing and reporting patterns (themes) within data’. 25 Transcripts were read and re-read to ensure familiarity. A coding framework was developed based on patterns and repeated topics identified in the data. Data were coded, and coded chunks of data were grouped into initial themes. These processes were conducted by a single researcher in the first instance and reviewed by two further researchers at different stages. Data were then re-read and the appropriateness of themes interrogated. Particular attention was paid to similarities and differences across device types, and to discrepancies between developed themes and the data. As a final step, the specifics of each theme were refined, and clear definitions for each theme were formulated.

Taking part in randomised controlled trials

All participants were in favour of patients being invited to participate in clinical research. Factors influencing this were both self-benefiting and altruistic, while including an appreciation of the serious nature of cancer, a desire to help future patients and an understanding that progress and current practice is based on the results of previous clinical trials, together with a hope of being randomised to a device not currently available outside the study. As one of the participants said:

Well I think it is, because had I not taken part in the first clinical trial, then I wouldn’t be still here. It’s as simple as that. I think that clinical trials are absolutely essential.

Female, Hickman, prior experience of a RCT

Similarly, another patient with no previous experience of RCTs said:

I think for future care as well, em, the experiences of patients now helps to form better care for people coming along, you know, kind of after us if you like.

Female, PICC, no prior experience of a RCT

Views on randomisation

All patients expressed an understanding of the process and reasons for randomisation. The majority either had participated in or would be prepared to participate in a RCT, acknowledging that this is an informed decision. Yet, despite professing to understand the concept of RCTs, participants’ comments indicated a degree of confusion over common trial terms (e.g. ‘trial’ was equated with ‘experiment’ and ‘treatment arm’ with ‘chemotherapy regime’). One patient considered that there should be a degree of patient choice in device allocation, assuming that this would not affect the numbers required to be allocated to each device:

I think it’s an informed choice, you know, If you are prepared to go ahead with the trial, it is made very clear to you, you know, it is randomised, you’ve got a chance of having, you know, one of the three and if you’re happy with that then, you know, yeh, go ahead.

Female, PICC, no prior experience of a RCT

Other factors influencing this decision were the severity of illness and whether or not participation would have an impact on other issues, such as treatment schedules:

If I chose the Hickman or the PORT, it would delay my next chemotherapy by a week. And my last chemotherapy finished on my birthday and I was determined to get to that last one on that day.

Female, PICC, no prior experience of a RCT

Certain patients, however, stated that they would not want to take part in a randomised trial of the three devices, as they wished to retain an element of choice in relation to line type:

I can understand why you want to do it, ehm . . . but eh, I know that given the choice I wouldn’t want to be randomly selected for a Hickman line or one that is implanted . . .

Male, PICC, no prior experience of a RCT

Study documentation

Comments on the language used in the PIS varied between ‘straightforward’ and ‘jargon free’ and, by contrast, ‘patronising’. Similarly, views on the length of the PIS also varied. Although several participants were not concerned about the length of the information sheet, considering that this would be what was expected, other views were that it was too long and would be overwhelming for patients. One participant, in particular, felt strongly that the design and layout of the PIS could be significantly improved. He considered that, initially, patients should be provided with a summary sheet, that is, the ‘technical specification’. The opportunity to find out further information should then be in the form of frequently asked questions. This could also be supplemented by links to a website (that could be accessed while attending for treatment). The longer document could be retained for reference.

Familiarity with the devices

Patients were not neutral in their stance towards the three devices. Some had received more than one device over the course of their treatment and this experience had coloured their judgement. For instance, a patient who previously had a PICC mentioned:

Well, probably not the PICC, I wouldn’t want that . . . That looped inside you.

Male, Hickman, prior experience of a RCT

Others were influenced by family members’ and friends’ experiences of the devices:

My wife had the Hickman line and eh, she . . . got a lot of infections, she eventually at one stage got septicaemia. She was on holiday at the time and had to be airlifted off one of the islands so that was a bit of a trauma.

Male, PICC, no prior experience of a RCT

As in every area of medical care, the level of patients’ knowledge and understanding of the subject varied widely. Many had conducted their own internet research on the three devices before deciding whether or not to enter the study. However, other patients admitted to being reluctant to know too much about the disease or treatment, preferring information to be filtered or summarised by health-care professionals.

Getting staff on board

It was clear that the CAVA trial would necessitate a relatively complex patient pathway crossing several different specialties, including oncology, haematology and radiology, as well as having implications for primary care staff. The need to have all staff informed, involved and committed to the study was a recurrent theme with both nursing and medical staff across all specialties and centres; this was acknowledged to be a priority and essential for effective recruitment and was frequently described as getting colleagues ‘on board’:

I have no concern, I mean I am absolutely confident that the staff, the nursing in particular, will be very supportive, but we need everybody to be absolutely onboard.

Oncologist, Glasgow

The CAVA champion was perceived to be instrumental in informing, advising and motivating staff at each centre, providing a point of contact for clinical and nursing staff and raising awareness of the study. Several initiatives were reported to be under way across centres to ensure that staff were kept up to date with the progress of study set-up, including sending out slides and flyers to advise of the study and holding meetings to introduce the study to relevant colleagues and visiting key clinicians and departments to raise awareness. The qualitative study, particularly the interviewing of clinical staff, also appeared to play a significant part in raising awareness of the study and stimulating discussion and debate.

The heavy workload typical of all departments treating patients with cancer, the number of RCTs carried out in cancer centres and the adverse impact that this can have on trial recruitment was not an unexpected finding and the need to ensure that staff are fully informed and motivated with regard to the study was reiterated. In addition, it became apparent that the oncology and haematology departments were much more familiar with drug trials than with device trials, raising the risk of a device trial being overlooked. The concept of ‘trial fatigue’ and the potential for a non-drug trial to be less visible was a valuable prompt to raise considerations of adequate notification of the study, particularly to those carrying out ongoing care, ensuring complete data collection and complications reporting. Initiatives to raise visibility of the study included colour coding patient-held diaries and case record forms, along with adding proformas to the front of case notes in clinics:

We are very busy, that can be the over-riding sort of factor in saying: ‘OK, fine, I’ll just get on with the standard work rather than spend another half an hour or an hour explaining a trial that the patient may or may not take up’.

Oncologist, Glasgow

Clinical staff perceptions and preferences

A significant finding of the pre-trial qualitative study was a lack of equipoise regarding the three devices for particular patient groups. Haematologists in all but one centre expressed concern regarding the suitability of PORTs in their patient subgroup. In addition to issues of staff inexperience with the device, concern was expressed over the risks posed by the more invasive nature of the procedure to insert and remove the device in thrombocytopenic and neutropenic patients. Consequently, in some centres there was a lack of engagement by haematologists in the study (Manchester and Durham). In Leeds, haematologists intended to randomise between PICCs and Hickman-type devices only, whereas in Glasgow concern was also expressed regarding the use of PICCs, which were considered to present difficulties in transfusing blood products and other supportive therapies required for acute leukaemic and transplant participants:

We are actually not familiar with the PORT in adult haematology or oncology really . . . It would be a major practical problem to be inserting it into our haematology patients in terms of the insertion itself, there is a possibility that the bleeding risk and the infection risk could create more problems for my type of patients, because these are acute leukaemia patients . . .

Haematologist, Leeds

Newcastle was the only centre where the haematologists were completely on board, willing to randomise between all three devices and to introduce PORTs for chemotherapy delivery in haematology patients.

With the exception of the issues discussed above, most clinicians and nursing staff interviewed seemed to be proactive in the conduct of RCTs and ostensibly at ease with introducing the concepts of equipoise and randomisation to cancer patients:

This is our, if you like, ‘bread and butter’. We are involved, myself and my colleagues . . . are involved in trials in general; Phase III trials, and randomisation is the core part of it. So, we do discuss randomisation with patients on a regular basis and I don’t see this as a complex trial at all, I think it is quite straightforward.

Oncologist, Glasgow

Logistics of device insertion

The logistics of the line insertion appointments was also a recurrent theme across clinical staff interviews. Several participants thought that the element of randomisation between devices could create an additional level of complexity in the process of commencing a patient on chemotherapy. This was particularly relevant in oncology and haematology, where the time from referral for line insertion to treatment can be short and any perceived potential delays to chemotherapy delivery may impact negatively on patient recruitment. Services at each centre varied significantly, and there was acknowledgement that there might be a scarcity of slots for PORT insertion in some centres. In others, PORT insertion for chemotherapy was in addition to existing workload:

I think, obviously you know kind of access to interventional radiology might be the decider here because if there is a waiting list to put a PORT in, what have you, then your treatment targets have to be met either for cancer waiting time target or from a clinical need target type of thing.

Oncologist, Newcastle

The introduction of PORTs for chemotherapy delivery also created a requirement for PORT removal appointments, either because of complications or because removal was requested by patients at the end of treatment. In many centres, this was in addition to current workload and, therefore, raised the issue of accommodating this requirement. Responses to this varied across centres; in one, this was not considered an appropriate use of interventional theatre time and arrangements were made for a surgical referral for PORT removal when necessary, whereas, in others, interventional radiologists were prepared to undertake line removal for study patients as required.

Lack of experience and training

Experience of PORTs varied widely between centres. In several centres, PORTs were already in use for chemotherapy delivery; however, in others experience was limited and the need for additional training was apparent. Staff interviews in Glasgow highlighted a lack of confidence and experience with PORTs as a major area of concern. Despite a programme of training, the learning curve generated by the new device resulted in additional referrals to interventional radiology for assistance with maintenance issues and additional hospital visits by patients for procedures normally carried out by the primary care team. The qualitative interviews did highlight, however, the nursing staff’s enthusiasm and commitment to developing the necessary skills:

So that’s a big buzz, eh, you know when we problem solve in a successful way. And I know that once we get used to PORTs, we’ll love them.

Oncology nurse, Glasgow

Attitudes towards trial materials

Clinical staff in all centres raised a few minor issues with trial materials, specifically the PIS. It was felt that illustrations were not as clear as they could be and that the inclusion of necklaces was distracting and did not demonstrate best practice. Some inconsistencies with local practice were raised, specifically the reference to pumps that were not universally used across all centres. It was felt that greater clarity was required in two areas: (1) that the radiation dose was less than that of a radiograph and (2) that monthly questionnaires did not involve a further hospital visit. Finally, two clinicians raised the point that the references to randomisation by ‘computer’ and ‘tossing a coin’ were considered not to be concepts that patients feel comfortable with in relation to decision-making. A summary of these points was forwarded to the project manager and the Trial Management Group for further discussion at a forthcoming Trial Management Group meeting, and the PIS was modified as appropriate.

Hickman versus peripherally inserted central catheters comparison: recruitment

Recruitment to the Hickman versus PICCs comparison was challenging and was consistently below the target recruitment rate. The recruitment target was revised in April 2015, but overall recruitment remained below the target (Figure 3). In total, 424 patients from 15 sites contribute to the PICCs versus Hickman comparison (see Report Supplementary Material 5). This is notably less than the required 778 patients to ensure 80% power for our non-inferiority comparison. The resultant power of this comparison is 54%.

FIGURE 3.

Recruitment projections for the PICCs vs. Hickman comparison.

PORTs versus Hickman comparison: recruitment

Recruitment for the PORTs versus Hickman comparison was above the target recruitment rate (Figure 4). As a consequence, the two-way randomisation option for this comparison closed at the end of November 2015. However, patients continued to accrue to this comparison via the three-way randomisation until recruitment ended in February 2018. A target of 550 patients was required and a total of 556 patients were recruited from 10 sites (see Report Supplementary Material 5), which results in a final power for this comparison of 95%.

FIGURE 4.

Recruitment projections for the PORTs vs. Hickman comparison.

PORTs versus peripherally inserted central catheters comparison: recruitment

The initial recruitment rate for the PORTs versus PICCs comparison was lower than the expected rate. The recruitment rate was reassessed in April 2015 and a revised target was produced. Subsequently, the trial recruited to target for this comparison (Figure 5). A total of 341 patients were required from 11 sites for the PORTs versus PICCs comparison to ensure 80% power for the final comparison (see Report Supplementary Material 5).

FIGURE 5.

Recruitment projections for the PORTs vs. PICCs comparison.

Inability to aspirate blood

The proportion of patients who experienced one or more occurrences of an inability to aspirate blood was larger in the PICCs arm (n = 45; 21%) than in the Hickman arm (n = 33; 16%). A simple flush was commonly performed in approximately one-third of all ‘actions undertaken’ to manage the complications. Other procedures included thrombolytic locks (21% of procedures undertaken) and infusion, further imaging and manipulation of the device in situ. Only one patient in each arm had a complication classified as SIR category C or above. As a consequence of an inability to aspirate blood, 36% (n = 16) of the patients with this complication in the PICCs arm had their device removed, compared with 12% of these patients (n = 4) in the Hickman arm. Chemotherapy was prematurely stopped in three patients (7%) in the PICCs arm.

Laboratory-confirmed infection

Laboratory-confirmed bloodstream infection was recorded in 5% (n = 10) of patients in the PICCs arm, compared with 19% (n = 41) in the Hickman arm. In both arms, these infections were primarily identified by culture of recognised pathogens on blood samples and were treated with antibiotics. Five patients in the PICCs arm and 34 patients in the Hickman arm had a complication classified as SIR category C or above; one death was recorded in the Hickman arm. As a consequence of laboratory-confirmed bloodstream infection, 60% (n = 6) of the patients with this complication in the PICCs arm had their device removed, compared with 66% (n = 27) of those patients in the Hickman arm. Chemotherapy was prematurely stopped in six of the Hickman arm patients.

Suspected catheter-related bloodstream infection

Suspected catheter-related bloodstream infection was recorded in 5% (n = 10) of patients in the PICCs arm, compared with 9% (n = 18) of patients in the Hickman arm. In both arms, these infections primarily presented as fever and were treated with antibiotics. Eight patients in the PICCs arm, compared with 14 patients in the Hickman arm, had a complication classified as SIR category C or above. As a consequence of suspected catheter-related bloodstream infection, 20% (n = 2) of the patients with this complication in the PICCs arm had their device removed, compared with 61% (n = 11) of those patients in the Hickman arm. Chemotherapy was prematurely stopped in two of the PICCs arm patients.

Exit site infection

Exit site infection was recorded in 2% (n = 4) of patients in the PICCs arm, compared with 9% (n = 19) of patients in the Hickman arm. All infections were confirmed and antibiotics were given in the majority of cases. One patient in the PICCs arm and four patients in the Hickman arm, had a complication classified as SIR category C or above. As a consequence of exit site infection, no patients in the PICCs arm and three patients in the Hickman arm had their device removed. Only one patient in the Hickman arm had their chemotherapy prematurely stopped.

The mean time to first infection (any infection) was similar in both arms: 33 ± 1.2 weeks in the Hickman arm and 33 ± 8 weeks in the PICCs arm. The median time to first infection (any infection) cannot be estimated because less than half of the patients experienced an event.

Venous thromboembolism and pulmonary embolism

Venous thrombosis was uncommon: 6% (n = 13) of patients in the PICCs arm and 5% (n = 10) of patients in the Hickman arm experienced venous thrombosis, of whom one patient in the PICCs arm reported two instances of venous thrombosis. Non-inferiority of PICCs in terms of venous thrombosis could not be concluded at a significance level of 2.5% for the logistic regression analysis (OR 1.30, 95% CI 0.55 to 3.08). The most frequent site for venous thrombosis in the PICCs arm was upper extremity veins (71%), whereas in the Hickman arm it was the internal jugular (40%) and subclavian (40%) veins. Patients were primarily treated with anticoagulants. Six patients in the PICCs arm and three patients in the Hickman arm had a complication classified as SIR category C or above. As a consequence, six patients in the PICCs arm and five patients in the Hickman arm had their device removed. Chemotherapy was prematurely stopped in no patients in the PICCS arm and in two patients in the Hickman arm.

Pulmonary embolism (PE) was rare, occurring in 3% (n = 6) of patients in the PICCs arm and 2% (n = 4) of patients in the Hickman arm. All patients received anticoagulants. As a consequence, none of the PICCs arm patients and one of the Hickman arm patients had their device removed. Chemotherapy was prematurely stopped in two of the PICCs arm patients and one of the Hickman arm patients.

Mechanical failure and other complications

Mechanical failure was reported in 15% (n = 31) of patients in the PICCs arm, compared with 3% (n = 7) of patients in the Hickman arm. Of these patients, six in the PICCs arm and three in the Hickman arm reported line fracture, 10 patients in the PICCs arm and two in the Hickman arm reported ‘device fallen out’, and 15 patients in the PICCs arm and two patients in the Hickman arm reported line migration that required intervention. Four patients in the PICCs arm and two patients in the Hickman arm had a complication classified as SIR category C or above. As a consequence, 28 patients in the PICCs arm and seven in the Hickman arm had their device removed. Chemotherapy was prematurely stopped in one patient in the Hickman arm.

Other complications were reported for 11% (n = 23) of patients in the PICCs arm, compared with 8% (n = 16) of patients in the Hickman arm. Two patients in the PICCs arm and three patients in the Hickman arm had a complication classified as SIR category C or above. As a consequence, 12 patients in the PICCs arm and five patients in the Hickman arm had their device removed. Chemotherapy was prematurely stopped in one Hickman patient.

Health-related quality of life

EuroQoL-5 Dimensions index values were calculated by patient and time point according to standard conventions (a value of 1 indicates no problems on any of the five dimensions and a value of –0.594 indicates maximum problems on all five dimensions). Health state scores were reported on a scale of 0–100, where 0 is the worst imaginable health state and 100 the best. Overall, there was a trend of small improvement over time in both the index values and the health state score (Figure 8). During the first 6 months, the study arms were similar. However, from around 7 months onwards, both the index values and the health state scores tended to be marginally higher and less variable in the Hickman arm than in the PICCs arm (see Figure 8). For each patient, the AUC described by the EQ-5D index values and health state scores were standardised by the time spent in the study, leading to values on the same scale as the original data. The standardised AUCs were then adjusted by subtracting the baseline value. The mean adjusted standardised AUC for the original EQ-5D index value data was –0.009 ± 0.01 for the PICCs arm and –0.007 ± 0.01 for the Hickman arm. The pooled estimate over the imputed data resulted in mean values of –0.095 ± 0.43 for the PICCs arm and –0.129 ± 0.03 for the Hickman arm. These scores were compared using the Mann–Whitney U-test, and a two-sided p-value was found not to be statistically significant at the 5% level. The mean adjusted standardised AUC for the original EQ-5D health state score data was –13.93 ± 2.42 for the PICCs arm and –8.79 ± 1.99 for the Hickman arm. The pooled estimate over the imputed data resulted in mean values of –8.57 ± 1.99 for the PICCs arm and –7.71 ± 1.73 for the Hickman arm. The difference in scores was not statistically significant at the 5% level (see Appendix 3).

FIGURE 8.

The PICCs vs. Hickman comparison for (a) EuroQoL-5 Dimensions index value and (b) health states. The error bars represent ± 1 standard error.

Quality of life: Quality of Life Questionnaire C30

The QLQ-C30 scale scores were calculated in accordance with standard EORTC conventions, resulting in values between 0 and 100; a high score represents a higher response level. During the first 6 months, the study arms were similar (Figure 9); however, from month 8, there was a tendency for apparently more markedly different patterns between the arms. The results were consistent at the 5% significance level for both the unadjusted and the adjusted approaches for all of the scales; there were no statistically significant differences between arms (see Appendix 4).

FIGURE 9.

The EORTC QLQ-C30 scores for the PICCs vs. Hickman comparison.

Venous access device-related quality of life

The worst score reported during the study for each question was established and compared across the two arms for the 403 ITT patients with a device fitted (excluding those patients with a different device fitted because of technical insertion failure) (Figure 10). There were statistically significant differences via the unadjusted p-values at the 5% level for hygiene and hobbies; however, these were not statistically significant when adjusted for the other questions. Both of these results appear to favour the Hickman over the PICC (see Appendix 5).

FIGURE 10.

Device-specific questionnaire for the PICCs vs. Hickman comparison.

Inability to aspirate blood

Similar proportions of patients in the PORTs arm (n = 38; 15%) and in the Hickman arm (n = 42; 14%) experienced one or more occurrence of an inability to aspirate blood; this accounts for 63 complications in the PORTs arm and 60 complications in the Hickman arm. A simple flush was commonly performed to manage the complications. Other procedures included thrombolytic locks and infusion, further imaging and manipulation of the device in situ. Six patients in the PORTs arm and five patients in the Hickman arm had a complication classified as SIR category C or above. As a consequence of an inability to aspirate blood, 5% (n = 2) of the PORTs patients with this complication and 29% (n = 12) of the Hickman patients had their device removed.

Laboratory-confirmed infection

Laboratory-confirmed bloodstream infection was recorded in 6% (n = 14) of patients in the PORTs arm, compared with 16% (n = 49) of patients in the Hickman arm. In both arms, these infections were primarily identified by culture of recognised pathogens on blood samples and were treated with antibiotics. Twelve patients in the PORTs arm and 39 patients in the Hickman arm had a complication classified as SIR category C or above. As a consequence of laboratory-confirmed bloodstream infection, 50% (n = 7) of the patients with this complication in the PORTs arm had their device removed compared with 76% (n = 37) in the Hickman arm. Chemotherapy was prematurely stopped in two of the PICCs arm patients and seven of the Hickman arm patients.

Suspected catheter-related bloodstream infection

Suspected catheter-related bloodstream infection was recorded in 8% (n = 19) of patients in the PORTs arm, compared with 5% (n = 15) of patients in the Hickman arm. In both arms, these primarily presented as fever and were treated with antibiotics. Fourteen patients in the PORTs arm compared with nine patients in the Hickman arm had a complication classified as SIR category C or above. As a consequence of suspected catheter-related bloodstream infection, 37% (n = 7) of the patients with this complication in the PORTs arm had their device removed compared with 60% (n = 9) in the Hickman arm. One patient in each of the two arms had chemotherapy prematurely stopped.

Exit site infection

Exit site infection was recorded in 4% (n = 10) of patients in the PORTs arm, compared with 9% (n = 26) of patients in the Hickman arm. All infections were confirmed and antibiotics were given in the majority of cases. Three patients in the PORTs arm and nine patients in the Hickman arm had a complication classified as SIR category C or above. As a consequence of exit site infection, no patients in the PORTs arm and six patients in the Hickman arm had their device removed. One patient in the PORTs arm and two in the Hickman arm had their chemotherapy prematurely stopped.

The mean time to first infection (any infection) was 42 ± 0.9 weeks for patients in the PORTs arm, compared with 37 ± 1.2 weeks for patients in the Hickman arm. The median time to first infection (any infection) could not be estimated because less than half of the patients experienced an event.

Venous thromboembolism

Venous thrombosis was uncommon: 1% (n = 3) of patients in the PORTs arm and 2% (n = 7) of patients in the Hickman arm. No statistically significant difference was detected using logistic regression. Patients were primarily treated with anticoagulants. Thromboses occurred in three patients in the PORTs arm, occurring in the superior vena cava, upper extremity and internal jugular veins. In the seven patients in the Hickman arm in whom thromboses occurred, the majority of cases occurred in the internal jugular or subclavian vein. One patient in the PORTs arm and two patients in the Hickman arm had a complication classified as SIR category C or above. As a consequence, one patient in the PORTs arm and three patients in the Hickman arm had their device removed. No patients had their chemotherapy prematurely stopped.

Pulmonary embolism was rare and occurred in 1% (n = 3) of patients in the PORTs arm and 1% (n = 4) of patients in the Hickman arm. All but one patient received anticoagulants. None of the patients had their device removed as a consequence of PE. Chemotherapy was prematurely stopped in one of the Hickman patients.

Mechanical failure and other complications

Mechanical failure was reported in 1% (n = 2) of patients in the PORTs arm, compared with 3% (n = 9) of patients in the Hickman arm. Both patients in the PORTs arm reported exposure of the PORT through the wound. In the Hickman arm, two patients reported line fracture, one reported exposure to fixation cuff, two reported ‘device fallen out’ and four reported line migration that required intervention. Two patients in the Hickman arm had a complication classified as SIR category C or above. As a consequence, all patients in the Hickman arm who had mechanical failure had their device removed. Chemotherapy was prematurely stopped in one patient in the Hickman arm.

Other complications were reported in 6% of patients in the PORTs arm (n = 14) and in 6% of patients in the Hickman arm (n = 17). Two patients in the PORTs arm and one patient in the Hickman arm had a complication classified as SIR category C or above. As a consequence, five patients in each arm had their device removed. Chemotherapy was prematurely stopped in two PORTs patients.

Health-related quality of life

Overall, there was little or no change in HRQoL over time. During the first 6 months, the study arms were similar. However, from month 7 onwards the mean index values and health state scores tended to be marginally higher and less variable for patients in the PORTs arm; however, at almost all time points, the standard error bars overlap (Figure 13). The mean adjusted standardised AUC for the original EQ-5D index values was –0.003 ± 0.02 in the PORTs arm and –0.010 ± 0.01 in the Hickman arm. The pooled estimate over the imputed data resulting in mean values was –0.157 ± 0.03 for the PORTs arm and –0.139 ± 0.02 for the Hickman arm. These scores were not statistically significantly different at the 5% level. The mean adjusted standardised AUC for the original EQ-5D health state score data were –9.72 ± 1.47 for patients in the PORTs arm and –10.90 ± 1.67 for patients in the Hickman arm. The pooled estimate over the imputed data resulting in mean values were –11.39 ± 1.73 for the PORTs arm and –8.74 ± 1.53 for the Hickman arm. The difference in these scores was not statistically significant at the 5% level (see Appendix 8).

FIGURE 13.

The PORTs vs. Hickman comparison for (a) EQ-5D index value; and (b) health state.

Quality of life: Quality of Life Questionnaire C30

The mean QLQ-C30 global health status was very similar across the two arms at all time points, with the exception of month 11 (Figure 14). The analyses of the AUC data over the duration of device insertion showed no statistically significant difference between the devices for this scale. There were statistically significant differences at the 5% significance level in terms of cognitive functioning (unadjusted p = 0.040 in favour of Hickman) and appetite loss (unadjusted p = 0.007 in favour of PORTs), as assessed by the QLQ-C30; however, after adjusting for multiple comparisons, these differences were no longer of statistical significance (see Appendix 7).

FIGURE 14.

The EORTC QLQ-C30 scores for the PORTs vs. Hickman comparison. The error bars represent ± 1 standard error.

Venous access device-related quality of life

The worst score reported during the study for each question was established and compared across the two arms for the 545 ITT patients with a device fitted (excluding those patients with a different device fitted owing to technical insertion failure) (Figure 15). There were statistically significant differences via the unadjusted p-values at the 5% level for 11 of the 16 questions. These remained statistically significant when adjusted for the other analyses. Hence, both of these results appear to favour PORTs over Hickman (see Appendix 8).

FIGURE 15.

Device-specific questionnaire for the PORTs vs. Hickman comparison.

Inability to aspirate blood

Similar proportions of patients in the PORTs arm (n = 23; 16%) and the PICCs arm (n = 37; 19%) experienced one or more occurrence of the inability to aspirate blood; this accounts for 33 complications in the PORTs arm and 55 complications in the PICCs arm. A simple flush was commonly performed to manage the complications. Other procedures included thrombolytic locks and infusion, further imaging and manipulation of the device in situ. Two patients in the PICCs arm had a complication classified as SIR category C or above. As a consequence of an inability to aspirate blood, 4% (n = 1) of PORTs patients with this complication and 27% (n = 10) of PICCs patients had their device removed. Three PICCs patients had their chemotherapy stopped prematurely.

Laboratory-confirmed infection

Laboratory-confirmed bloodstream infection was recorded in 5% (n = 8) of patients in the PORTs arm, compared with 4% (n = 7) of patients in the PICCs arm. In both arms, these infections were primarily identified by culture of recognised pathogens on blood samples and were treated with antibiotics. Six patients in the PORTs arm and four patients in the PICCs arm had a complication classified as SIR category C or above. As a consequence of laboratory-confirmed bloodstream infection, 63% (n = 5) of the patients with this complication in the PORTs arm had their device removed, compared with 86% (n = 6) of those patients in the PICCs arm. Chemotherapy was prematurely stopped in one patient in the PORTs arm and two patients in the PICCs arm.

Suspected catheter-related bloodstream infection

Suspected catheter-related bloodstream infection was recorded in 5% (n = 8) of patients in the PORTs arm, compared with 3% (n = 5) of patients in the PICCs arm. In both arms, these infections primarily presented as fever, and all patients were given antibiotics. Six patients in the PORTs arm, compared with five patients in the PICCs arm, had a complication classified as SIR category C or above. As a consequence of suspected catheter-related bloodstream infection, 50% (n = 4) of the patients with this complication in the PORTs arm had their device removed, compared with 60% (n = 3) of those patients in the PICCs arm. One patient in the PICCs arm had chemotherapy prematurely stopped.

Exit site infection

Exit site infection was recorded in 3% (n = 4) of patients in the PORTs arm, compared with 2% (n = 4) of patients in the PICCs arm. All infections were confirmed and antibiotics were given in the majority of cases. Two patients in the PICCs arm had a complication classified as SIR category C or above. As a consequence of exit site infection, no patients in the PORTs arm and three patients in the PICCs arm had their device removed. No patients had their chemotherapy prematurely stopped.

The mean time to first infection (any infection) was 43 ± 01.2 weeks in the PORTs arm, compared with 44 ± 1.1 weeks in the PICCs arm. The median times to first infection (any infection) could not be estimated because less than half of the patients experienced an event.

Venous thrombosis and pulmonary embolism

Venous thrombosis was recorded in 2% (n = 3) of patients in the PORTs arm and 11% (n = 22) of patients in the Hickman arm. This difference was found to be statistically significantly different at the 5% significance level (OR 0.14, 95% CI 0.04 to 0.50). Patients were primarily treated with anticoagulants. Two patients in the PORTs arm and nine patients in the PICCs arm had a complication classified as SIR category C or above. As a consequence, one patient in the PORTs arm and nine patients in the PICCs arm had their device removed. One patient in the PORTs arm had their chemotherapy prematurely stopped.

Pulmonary embolism was rare and occurred in 2% (n = 3) of patients in the PORTs arm and 1% (n = 1) of patients in the PICCs arm. All patients received anticoagulants. No patients with PE had their device removed or their chemotherapy prematurely stopped.

Mechanical failure and other complications

Mechanical failure was reported in 3% (n = 4) of patients in the PORTs arm, compared with 11% (n = 21) of patients in the PICCs arm. Of these patients, five in the PICCs arm reported line fracture, one in the PORTs arm reported exposure of the PORT through their wound, one in the PORTs arm and eight in the PICCs arm reported ‘device fallen out’, and two in the PORTs arm and eight in the PICCs arm reported line migration that required intervention. Two patients in the PICCs arm had a complication classified as SIR category C or above. As a consequence, one patient in the PORTs arm and 19 in the PICCs arm who had mechanical failure had their device removed. No patient had their chemotherapy prematurely stopped.

Other complications were reported in 11% (n = 16) of patients in the PORTs arm and 10% (n = 19) in the PICCs arm. Three patients in the PORTs arm and two in the PICCs arm had a complication classified as SIR category C or above. As a consequence, six patients in the PORTs arm and 12 patients in the PICCs arm had their device removed. No patient had their chemotherapy prematurely stopped.

Health-related quality of life

Overall, there was little or no change in HRQoL over time. During the first 6 months, the study arms were similar. Following that, however, the mean index scores look to be marginally higher in the PICCs arm (Figure 18). The mean adjusted standardised AUCs for the original EQ-5D index values were –0.031 ± 0.01 for the PORTs arm and –0.017 ± 0.01 for the PICCs arm. The pooled estimate over the imputed data resulting in mean values were –0.216 ± 0.04 for the PORTs arm and –0.113 ± 0.04 for the PICCs arm. These scores were not statistically significantly different at the 5% level. The mean adjusted standardised AUCs for the original EQ-5D health state score data were –17.09 ± 2.84 for the PORTs arm and –19.08 ± 2.84 for the PICCs arm. The pooled estimate over the imputed data resulting in mean values were –14.51 ± 1.99 for the PORTs arm and –9.78 ± 2.21 for the PICCs arm. The difference in these scores was not statistically significant at the 5% level (see Appendix 11).

FIGURE 18.

The PORTs vs. PICCs comparison for (a) EQ-5D index value and (b) health state. The error bars represent ± 1 standard error.

Quality of life: Quality of Life Questionnaire C30

The mean QLQ-C30 global health status was very similar across the two arms at all time points, with the exception of months 7 and 10 (Figure 19). The analysis of the AUC data over the duration of device insertion showed no statistically significant difference between the devices for this scale. There was a statistically significant difference at the 5% significance level in terms of constipation (unadjusted p = 0.029 in favour of PICCs), as assessed by the QLQ-C30; however, after adjusting for multiple comparisons, this statistical significance was lost (see Appendix 11).

FIGURE 19.

The EORTC QLQ-C30 scores. The error bars represent ± 1 standard error.

Venous access device-related quality of life

The worst score reported during the study for each question was established and compared across the two arms for the 325 ITT patients who had a device fitted (excluding those patients who had a different device fitted owing to technical insertion failure) (Figure 20). There were statistically significant differences favouring PORTs over PICCs at the 5% significance level for 8 of the 16 questions relating to hygiene, exercise, hobbies, self-consciousness, socialising, at risk of damaging device, at risk of infection and impact on quality of life. The adjusted p-values were all < 0.001, with the exception of self-consciousness (p = 0.016), socialising (p = 0.009) and risk of infection (p = 0.003) (see Appendix 12).

FIGURE 20.

Device-specific questionnaire for the PORTs vs. PICCs comparison.

Device insertion

The resource required for device insertion varied across the three devices. The insertion of Hickman-type devices was most commonly led by radiologists in a radiology department (46–48%). Nurse-led and anaesthesiologist-led insertions in a procedure or treatment room accounted for > 20% of the patients who had a device fitted (Table 11). The insertions of PICCs were most commonly nurse led and performed in a procedure or treatment room (67–73%); radiologist-led insertions were less frequent (11–13%). Similar to Hickman insertions, PORTs insertions were primarily radiologist led in a radiology department. Nurse-led PORTs insertions varied substantially (2–24%). All devices were inserted under local anaesthetic, with the exception of five PORTs patients who received general anaesthetic. The use of prophylactic antibiotics was uncommon and non-antimicrobial dressings were the most commonly applied across all three devices.

Device insertions for the majority of cases required an additional nurse to be present during the procedure (86–95% among Hickman patients and 81–97% among PORTs patients) (Table 12); however, this was less frequent for PICCs (65–81% of patients). Similarly, the use of an additional radiographer was common for the insertion of Hickman-type devices and PORTs but was less common, for PICCs. During the procedure for each of the devices, it was not common to have an additional anaesthesiologist, radiologist, doctor or surgeon present. The majority of patients received ultrasound imaging (Table 13). Fluoroscopy was far more frequently used in patients receiving Hickman-type devices and PORTs. Approximately one-quarter of patients receiving a PICC required the use of fluoroscopy. The use of chest X-rays was more common with PICCs than with Hickman-type devices and PORTs.

Unplanned outpatient visits

In the PICCs versus Hickman comparison, the proportion of patients reporting unplanned outpatient visits was similar when comparing PICCs (28%) with Hickman (28%) during the follow-up period (Table 14). The mean and median number of outpatient visits were similar across the two arms. In the PORTs versus Hickman comparison, 13% of the PORTs patients reported outpatient visits, compared with 27% of Hickman patients. The mean number of visits in the PORTs arm was approximately half that of the Hickman arm. In the PORTs versus PICCs comparison, 17% of PORTs patients reported outpatient visits compared with 28% of PICCs patients. The mean number of outpatient visits was substantially smaller in the PORTs arm (0.250 ± 0.63) than in the PICCs arm (0.476 ± 1.19).

Unplanned inpatient admissions

In the PICCs versus Hickman comparison, 14% of the PICCs patients reported unplanned inpatient admissions during the follow-up period, compared with 20% of the Hickman patients (Table 15). The mean length of stay for these patients was shorter in the PICCs arm (1.198 ± 5.05 days) than in the Hickman arm (2.279 ± 6.92 days). In the PORTs versus Hickman comparison, 12% of PORTs patients and 18% of Hickman patients reported unplanned inpatient admissions. The mean length of stay was shorter in the PORTs arm (1.05 ± 4.23 days) than in the Hickman arm (1.50 ± 4.59 days). In the PORTs versus PICCs comparison, a similar proportion of patients reported unplanned inpatient admission in both arms (12% of PORTs and 11% of PICCs). Patients in the PORTs arm were associated with a longer length of stay (1.45 ± 6.37 days) than patients in the PICCs arm (0.50 ± 2.02 days). Regardless of the device inserted, few patients reported extended hospital stay. The maximum reported length of stay ranged from 19 days in the PICCs arm (when compared with the PORTs arm) to 56 days in both the PICCs arm and the Hickman arm (when compared with each other).

Hickman versus peripherally inserted central catheters comparison

There was a total of 424 patients included in the PICCs (n = 212) versus Hickman (n = 212) comparison, who reported 110 and 103 complications, respectively (Table 17). The predicted mean total cost, which accounted for the skewed distribution and stratification factors in the trial, was £1708 in the PICCs arm, compared with £3262 in the Hickman arm. This lower cost (–£1553) in the PICCs arm was associated with lower mean QALY gained (–0.009 QALYs; equates to a loss of 3 days of perfect health over a 12-month period) and a higher complications rate (52% in the PICCs arm and 49% in the Hickman arm). However, the differences in QALYs and in complication rates were not statistically significant. Compared with Hickman-type devices, PICCs were associated with an ICER of £172,556 saved per QALY lost and, based on a cost-effectiveness ceiling ratio of £20,000, a NMB of £1373.

A visual illustration of the uncertainty surrounding the mean cost and QALY gain associated with each device over 1000 iterations of bootstrap analysis plotted on the cost-effectiveness plane is presented in Figure 23. The x-axis represents the incremental QALY gain, and the y-axis represents the incremental cost, associated with the use of PICCs compared with Hickman-type devices. The majority of the costs and QALY estimates lie in the south-west quadrant of the cost-effectiveness plane. Despite some uncertainty relating to the magnitude of the difference in costs between the two devices, the overall finding is that PICCs were associated with lower costs and a decrement in QALY gains when compared with Hickman. Irrespective of the willingness-to-pay threshold, the probability of the Hickman-type device being cost-effective compared with PICCs is low (see Figure 23).

FIGURE 23.

The PICCs vs. Hickman (a) cost-effectiveness plane and (b) acceptability curve.

PORTs versus Hickman comparison

There was a total of 556 patients included in the PORTs (n = 253) versus Hickman (n = 303) comparison, who reported 73 and 131 complications, respectively (Table 18). The predicted mean total cost was £2436 in the PORTs arm, compared with £2481 in the Hickman arm. There was a very small difference in mean QALYs gained (0.004) favouring PORTs. Furthermore, the PORTs arm was associated with a substantially lower complication rate (29%) than the Hickman arm (43%). In this scenario, where the cost of PORTs was lower than that of Hickman with additional gain in QALYs, PORTs dominate Hickman. Similarly, PORTs dominate Hickman from the perspective of costs and complications averted. Compared with Hickman, PORTs were associated with an ICER of –£11,250 per QALY gained (note that a negative ICER in this case owing to PORTs being dominant) and, based on a cost-effectiveness ceiling ratio of £20,000, a NMB of £125.

A visual illustration of the uncertainty surrounding the mean cost and QALY gain associated with each device over 1000 iterations of bootstrap analysis plotted on the cost-effectiveness plane is presented in Figure 24. The majority of the simulated mean costs and mean QALYs were centred around the origin; there is little difference in costs and QALYs between the two arms.

FIGURE 24.

The PORTs vs. Hickman cost-effectiveness plane.

PORTs versus peripherally inserted central catheters comparison

There was a total of 346 patients included in the PORTs (n = 147) versus PICCs (n = 199) comparison, who reported 47 and 93 complications, respectively (Table 19). The predicted mean total cost was £2706 in the PORTs arm, compared with £1041 in the PICCs arm. There was a difference in mean QALYs gained between the two arms (0.741 in the PORTs arm and 0.759 in the PICCs arm; –0.018 equates to a loss of 6.5 days of perfect health over a 12-month period); this was not statistically significant. However, the PORTs arm was associated with a substantially lower complications rate (32%) than the PICCs arm (47%). In this scenario, PORTs were dominated by PICCs (the cost of PORTs was greater and the QALYs were lower). Compared with PICCs, PORTs were associated with an ICER of –£56 per QALY gained (note that a negative ICER in this case owing to PORTs being dominated) and, based on a cost-effectiveness ceiling ratio of £20,000, a NMB of –£2025. However, PORTs were associated with a 15% reduction in complication rates. The incremental cost per complication averted (based on the number of complications averted per 100 patients) was £104.

A visual illustration of the uncertainty surrounding the mean cost and QALY gain associated with each device over 1000 iterations of bootstrap analysis plotted on the cost-effectiveness plane is presented in Figure 25. The majority of the simulated mean costs and mean QALYs were centred around the origin; there is little difference in costs and QALYs between the two arms. Despite some uncertainty related to the magnitude of the difference in costs between the two devices, the overall finding is that PORTs were associated with greater costs and a small decrement in QALY gains than PICCs.

FIGURE 25.

The PORTs vs. PICCs cost-effectiveness plane.

Time horizon of 3 and 6 months

The majority of the complications reported in the CAVA trial occurred within the first 6 months of the follow-up period. In addition, the data for the EQ-5D were also more complete in the first 6 months than the final 6 months of the follow-up. Therefore, the sensitivity analysis was conducted by restricting the data to 3 and 6 months of follow-up. In the PICCs versus Hickman comparison, at 3 months, PICCs were associated with lower costs (–£1077) but also with a small decrement in QALYs gained (–0.020); similar findings were observed at 6 months (Table 20). These results were in line with the base case. In the PORTs versus Hickman comparison, at 3 months, PORTs were associated with lower costs (–£135) but also with a small decrement in QALYs gained (–0.009; equivalent to 3 days of perfect health); similar findings were observed at 6 months. This small QALY decrement was not observed with the base case, which showed an incremental QALY gain of 0.004 at 12 months. In the PORTs versus PICCs comparison at 3 months, PORTs were associated with greater costs (£1039) but with a small decrement in QALYS gained (–0.010; equivalent to 3.5 days of perfect health); similar findings were observed at 6 months. This is in line with the results of the base case.

Solid tumours only

A sensitivity analysis was performed on patients with solid tumours only (Table 21). The main impact seen in this analysis is the reduction in mean total costs, primarily driven by the mean unplanned follow-up associated with PORTs. The mean unplanned follow-up costs associated with PORTs reduced from £2436 in the base case to £2096 when compared with Hickman; similarly, a reduction from £2706 (base case) to £2286 was observed when compared with PICCs. In the PICCs versus Hickman comparison, the results were consistent with the base case. In the PORTs versus Hickman comparison, PORTs were associated with lower costs (–£59) and a small decrement in QALYs gained (–0.001; equivalent to less than 1 day of perfect health) than Hickman. In the PORTs versus PICCs comparison, the results were consistent with the base case.

Nurse-led service of care

The health-care resource use required for PORTs insertion varied across the sites within the CAVA trial (Table 22). A scenario analysis based on a nurse-led service of care was conducted. Under this scenario, it was assumed that the insertion of PORTs will be led by a senior nurse, with the support of two others. Overall, this reduced the cost of PORTs insertion and subsequently the total costs. When compared with Hickman, PORTs are no longer associated with greater costs (–£297). At lower costs and when there is no difference in QALYs, PORTs become the dominant option. When compared with PICCs, the incremental cost of PORTs was less than that of the base case (£1194 vs. £1665); however, PICCs remained the dominant strategy (lower costs and greater QALYs gained).

Acceptability of central venous access devices

Most participants expressed satisfaction with their device and, in general, devices were regarded as having less of an impact than other aspects of participants’ journeys, including the side effects of chemotherapy, surgical interventions and the effects of their illness more broadly. Participants with experience of chemotherapy via peripheral cannulation said that their device facilitated treatment, making administration less painful and easier (for themselves and staff). Participants tended to frame satisfaction in terms of the role of their device in their treatment and in their overall journey. They saw their device as part of their treatment and, hence, part of the reason that they were getting better or were ‘still here’: ‘It’s part of my treatment so it’s part of my life, it’s there’ (female, PICC). In addition, when expressing satisfaction with their device, many participants referred to having had a problem-free experience with their device. That being said, our sample included people who had experienced various difficulties and complications with each of the devices. We found that even these patients remained satisfied with their overall experience. Complications, including partial or minor malfunctions, did not appear to affect acceptability or satisfaction, as long as the device remained functional.

Living with a central venous access device

Notwithstanding acceptance, living with a CVAD presented distinct challenges, which necessitated meaningful adjustments and adaptations. Although the experience of individual participants differed substantially, we provide an inclusive account of the wide range of experiences described, highlighting points of difference between devices.

Patterns of preference

Among our sample, participants tended to prefer their own device and most participants stated that, if they had to choose a CVAD they would choose the one that they currently had. Noticeable differences between devices were evident with regard to how satisfaction and preferences were formulated and expressed. For instance, we observed differences in the comparators participants used when discussing device satisfaction. Participants with PICCs and Hickman-type devices tended to compare their devices (favourably) with peripheral cannulation. Participants in a Hickman-only focus group compared Hickman and PICCs and preferred the former, which they felt was less obtrusive. Unless prompted, participants with either of these devices tended not to make comparisons with PORTs. When prompted, there was ambivalence. Some were unsure about the idea of a needlestick for device access or, among those with PICCs, a device in their chest. Some claimed to lack sufficient knowledge about PORTs to compare. Others, although happy with their current device, professed an interest in PORTs. By comparison, participants with PORTs consistently compared them with PICCs and Hickman-type devices and explicitly regarded PORTs as superior CVADs:

The whole three [CVADs] were an option and I was hoping that it would be the PORT just because you didn’t see anything.

Female, Hickman

Furthermore, participants with PICCs and Hickman-type devices tended to express satisfaction with their device in personal terms. In the following extract, when ‘participant 1’ was asked if he would recommend PICCs to others he was reluctant, stressing the personal nature of his preference. This was not the case with ‘participant 2’, who had a PORT device:

Participant 1: Well I wouldn’t suggest, I wouldn’t give anybody advice on it because it’s up to them what that they do, it’s everybody’s personal choice what they like, em, but I would just say, well I find this a lot easier you know, so that’s . . . [to participant 2] obviously you’ve had a PORT . . .

Male, PICC

Participant 2: I would recommend the PORT.

Female, PORT

As illustrated here, participants who preferred PORTs were more forthright in their preferences, which tended to be based on the characteristics of the device itself, rather than familiarity or personal factors. This forthrightness was noted with interest by one participant with a PICC:

Loving the fact that the PORT owners are really enthusiastic, there is this sense of like evangelical ‘yeah this is great’. Whereas with the PICC line we are all indifferent, different experiences, it’s not as clear cut, whereas you guys are.

Male, PICC

Participants who could directly compare PORTs with other devices were unequivocal in their preference for PORTs and a positive regard for PORTs held even among those who had experienced painful placement procedures and device complications:

So, it [PORT] hasn’t worked for me, but if it had worked I would think it was the best one.

Female, PORT

Reflecting on the provision of PORTs by the NHS, participants advocated for greater access to PORTs, arguing that the benefits justify the additional costs:

And I feel quite strongly the very small difference between the Hickman-line costs and PORT costs and the overall benefits to people who are going, suffering enough anyway, their life should be made easier. And it’s a small difference, you know. And as I say I have had the benefit of trying these different things, and I find just, there’s no comparison.

Female, Hickman: prior experience of PORT

Experiences of trial participation

Most participants intimated a prosocial motivation for taking part in the CAVA trial. They were motivated to help other patients who came later to ‘give something back’ to the hospital or health system or to advance cancer treatment and, as one participant put it, help to ‘end this horrible disease’. Part of this motivation was a result of attaching a high value to research. One participant described a culture of research at the centre he attended and reported pre-emptively asking if there were any trials that he could take part in. For others, personal or family connections to the NHS or to health research reinforced motivation:

I signed up for all the trials when I got diagnosed, so owt [‘anything’] I could do to help the research.

Male, PORT

Participants’ decisions to take part in the trial also depended on their opinions regarding the CVADs under investigation and their feelings regarding randomisation. When invited to take part, many participants reported a preference for one device (most often PORTs), either a long-standing or a newly formed preference based on the information that they received about the three devices. Indeed, for some patients, being informed about the trial prompted them to do their own research into different vascular access options. These participants tended to form strong preferences. Others did not form opinions of their own but reported observing preferences among staff and inferring the superiority of that device (again, usually PORTs). PORTs were not routinely offered or available to NHS patients at several trial centres. This was a key incentive for many. Some of these patients reported that the randomisation process felt like taking part in a lottery; they felt lucky or unlucky, depending on the outcome:

I actually wanted a PORT and I was told the only way I could have one in Glasgow was by paying for it. Or by joining the trial. So, that’s why I joined the trial and I was thinking about paying for a PORT, but because I ended up having weekly chemo [chemotherapy] I didn’t mind having the PICC line, because it meant I was in here every week to have bloods, so it would get re-dressed then.

Male, PICC

Some participants professed not to prefer any device. These participants tended to feel that they did not know much about any of the devices, that all were equally unknown and there was nothing to be lost by leaving device selection to chance. Some were well informed but unable to choose; opting for randomisation was a way of solving their dilemma. Finally, some participants reported having been confused about the nature of randomisation when first informed about the trial, but stated that this had been addressed before participation.

For the vast majority of participants, the trial appeared to be a source of information and new knowledge about vascular access options and/or an opportunity to gain access to otherwise unavailable devices. For many, it also represented a source of additional support. For instance, one participant who was nervous about having a line said that the trial nurse took the time to explain the device and procedure to her, which was very helpful. A few patients, however, reported confusion or conflation of the trial with treatment as usual. Two patients, for example, referred to their PORT devices as ‘the CAVA’ or ‘the CAVA PORT’. Furthermore, one participant who had a PORT device and lived in an area in which PORTs were available only as part of the trial thought that the care and maintenance that she received for her PORT was also only part of the trial:

This is what I’m not clear of on the study, as to when the study comes to an end, what happens to us all going around with PORTs?

Female, PORT

Following device placement, further contact with the trial varied. Participants received and completed regular questionnaires (i.e. self-report measures of quality of life). Sometimes, questionnaires were brought to them in person by CAVA trial staff, who sat and went through the questionnaire with them. For some participants, this was an appreciated extra point of contact with a member of clinical staff who had a good understanding of their devices: an opportunity to ask questions or just chat about their experience.

Acceptability of central venous access devices

All devices were regarded as acceptable by participants, and no device was considered inherently superior. Interviewees felt that all three had a place and purpose in the context of anticancer treatment. It was also noted that direct comparisons between different CVADs were not always reasonable, as several independent factors meant that certain devices cannot/should not be used under certain circumstances, such as a specific treatment regimen, patient vasculature and other patient factors (e.g. support and access to transport):

I think we’ve learned, eh, more through routine clinical practice than the trial, that they all have risks and complications and none of them are perfect. You can’t say oh, this one’s head and shoulders above. And that’s why the trial is so important, because I honestly do not know which is best.

Oncologist

Although interviewees consistently acknowledged a lack of empirical evidence comparing the PICC, Hickman and PORT devices, they were willing to express personal perspectives. These tended to be couched as subjective opinions, based on personal observations, experience or common sense. Perceived advantages and disadvantages primarily concerned device safety, including complications (e.g. thrombosis, haematoma, movement/displacement of device and clot), infections and the durability of the device. Additional points of comparison included ease of use and maintenance, compatibility with local services, ease of insertion and costs.

Experiences of trial participation

The CAVA trial was regarded positively by participants, who saw it as necessary research, addressing important questions. The trial itself was seen as straightforward to run, and participant information materials were seen as well designed and helpful. At some centres, taking part had increased or even initiated the use of devices that were not commonly used prior to the CAVA trial, typically PORTs. This was attributed to an increased awareness of these devices as an option for patients and an increased capacity to place and manage PORTs:

And we had free PORTs to start with . . . I think, we ended up with about six or eight free PORTs, which was great, and they just have this kind of get off the ground and allow people to see them.

Non-clinical researcher

The CAVA trial was also seen as an extra source of support for patients. Trial staff became an extra point of contact for patients with queries about their device. Contact necessitated by the trial (e.g. administering quality-of-life questionnaires) presented patients with an opportunity that they might not otherwise have had, namely to chat about concerns. A few participants also suggested that the very fact that the trial was taking place would raise awareness of the general dearth of evidence relating to CVADs, and hopefully inspire further work:

Sadly, like all these trials, it will probably raise as many questions as it answers, but I think it will, it will put the sort of problem on the map and it will, even if it doesn’t give all the answers we’d like, I think it will start to focus people’s attention as to the fact that there is no evidence on this and will encourage others to sort of build on that. So, I think it will sort of get a momentum because you’ll need, you’ll need a whole series of trials, you know, like is it better to put lines on the right or the left, you know, are big lines better than small lines, and all this stuff.

Anaesthetist

Post-trial expectations

Interviewees believed that results in three domains would be important for informing practice: safety, cost and patient outcomes. Regarding safety, staff were interested in evidence relating to infection rates, device durability and complications, including device malfunction, movement, blood clots and thrombosis. Some wished for evidence regarding the effects of quality of care (handling, staff experience and proper placement) on these factors. Discussions of cost-effectiveness mainly centred on one issue: interviewees suspected that the belief that PICCs are the most cost-effective of the three CVADs (and PORTs the least) was not correct and were curious to see if this suspicion would be borne out by the evidence. Participants were also interested in patient-reported outcomes, with many expecting that PORTs would have the least negative impact on quality of life.

Almost all participants expressed a willingness to respond to new evidence. Most wished that, whatever the outcome of the CAVA trial, there would be an appropriate response broadly and/or locally. Some interviewees hoped that results would find one device to be consistently superior across domains, which would greatly simplify device selection. Others held that this would not be likely or, in fact, desirable:

And eh, I wouldn’t like to see a situation whereby the end result is that this device is better than the other, because there are too many factors involved and I don’t believe you can say that.

Nurse (vascular access)

Among those who expected a ‘winner’, most believed that PORTs would be identified as superior, certainly safer and preferred by patients. These interviewees hoped that PORTs would become a more widely available option for patients. Indeed, most interviewees emphasised the importance that all three devices be made equally available in future, which, in practice, often meant increasing the availability of PORTs. In any case, it was hoped that the trial findings would be used to support evidence-based decisions and recommendations both in individual cases and in terms of service provision.

Finally, in broader terms regarding the future of central venous access in chemotherapy, most participants felt that early intervention was important (i.e. avoidance of cannulation), emphasising the experience of patients or the importance of preserving patients’ veins for purposes beyond anticancer treatment. In general, our analysis found that, regardless of the outcomes that they had expected or hoped for, interviewees were clear that all three devices had a place within anticancer treatment and that none should be confined to the shelf, unless it was found to pose an unacceptable risk.