Evidence review

Clinical and technical evidence

Regulatory bodies

A search of the Medicines and Healthcare Products Regulatory Agency website revealed no manufacturer Field Safety Notices or Medical Device Alerts for this device. No reports of adverse events were identified from a search of the US Food and Drug Administration (FDA) database: Manufacturer and User Device Facility Experience (MAUDE).

Clinical evidence

Of the 27 relevant publications identified, 22 were excluded because they were abstracts. Consequently, 5 studies are included in this briefing. Of these, 2 were feasibility studies, and 3 were pilot studies. A total of 138 people were included in the studies. Two of the pilot studies include largely overlapping cohorts, with different outcomes.

Pilot studies

Konda et al. (2013; table 2) carried out a multinational, multicentre, prospective pilot study. The aims of the study were to assess the diagnostic accuracy of Cellvizio to differentiate benign and potentially malignant cysts, and the safety of the technique. A total of 57 people scheduled for endoscopic ultrasound with fine-needle aspiration (EUS‑FNA) had needle-based confocal laser endomicroscopy (nCLE) imaging. The final diagnosis was either based on histopathological diagnosis of specimens from people who had had surgery (n=14), or by clinical diagnosis after a blinded committee consensus review of images by 5 investigators (n=43). Results from surgical and non-surgical patients (n=26) were analysed to provide definitions for images associated with various pancreatic structures. In the remaining 31 patients, a blinded committee consensus review of the nCLE images suggested that the presence of epithelial villous structures was associated with potentially malignant cysts (p=0.004) and provided diagnostic accuracy of 71%, sensitivity of 59%, specificity of 100%, positive predictive value (PPV) of 100%, and negative predictive value (NPV) of 50%. The overall complication rate was 9% and included pancreatitis (1 mild case, 1 moderate case), transient abdominal pain (n=1), and intracystic bleeding not needing any further measures (n=3).

Napoleon et al. (2015a; table 4) carried out a prospective multicentre pilot study to identify and validate new diagnostic criteria for nCLE of pancreatic cysts. A total of 31 people admitted for EUS‑FNA for pancreatic cysts had nCLE. Final diagnosis was based on histopathology in patients who had had surgery (n=7), or on conclusive FNA results (n=11). For the remaining patients (n=13), 6 investigators and 2 pathologists reviewed all available information to make a final diagnosis by committee consensus. A superficial vascular network pattern visualised on nCLE was identified as the significant diagnostic criterion for benign cysts by non-blinded consensus review by 6 investigators. The overall diagnostic accuracy, sensitivity, specificity, PPV, and NPV for diagnosing benign cysts were 87%, 69%, 100%, 100% and 82% respectively. Interobserver agreement was substantial (kappa=0.77). A single event of mild acute pancreatitis was recorded, resulting in an overall complication rate of 3.2%.

Napoleon et al. (2015b; table 5) carried out a retrospective analysis of prospectively collected data (from largely the same patients described in Napoleon et al. 2015a) to determine new nCLE diagnostic criteria to characterise pancreatic cysts and to validate the performance of these criteria (n=33). The final diagnosis was based on histopathological diagnosis of specimens from people who had had surgery (n=20). A committee consensus review of all information produced the final diagnosis in those who had not had surgery (n=13). New nCLE criteria were identified for various cyst types (mucinous cystic neoplasms [MCN], pseudocysts [PC] and cystic neuroendocrine neoplasms [NEN]). People with cystic NEN were excluded from the subsequent validation of the identified criteria due to the small number (n=2). nCLE yielded a conclusive diagnostic result in 74% (n=23) of 31 surgical and non-surgical participants. Of these, 87% (n=20) were correct and 13% (n=3) were incorrect when compared with final diagnosis. Interobserver agreement for cyst types ranged from 0.35 for potentially malignant cysts (indicating fair agreement) to 1.00 for benign cysts (indicating perfect agreement). The overall agreement for all cysts was kappa=0.72 (indicating substantial agreement).

Feasibility studies

A US-based single-centre, prospective study was carried out by Konda et al. (2011; table 1) to evaluate the feasibility of using the Cellvizio system in 18 people having EUS‑FNA for the evaluation of pancreatic cysts. Images were obtained in 17 of 18 people, but technical challenges were encountered in 6 of 18 attempts to image. There were no device malfunctions. Good to very good quality images were obtained in 10 of the 17 people. Image quality was rated as moderate in 5 people, and low or poor in 2 people. Two serious adverse events occurred; both were pancreatitis requiring hospitalisation.

In a single-centre, prospective, single-arm feasibility study performed in the US, Nakai et al. (2015; table 3) investigated the feasibility, safety, and diagnostic accuracy of cystoscopy and nCLE in the clinical diagnosis of potentially malignant cysts. Thirty people having EUS‑FNA for the evaluation of pancreatic cysts were examined using 3 sequential EUS-guided procedures: nCLE with the Cellvizio system, followed by cystoscopy with the SpyGlass system, and then FNA. In people who had had surgery (n=2), the final diagnosis was based on histopathology. In people who had not had surgery (n=28), diagnosis was established by 2 independent, blinded investigators based on clinical presentation, combined with image findings on EUS, CT, and MRI, as well as fluid analysis and cytology on EUS. Image quality was reviewed by 1 non-blinded investigator. The procedure was technically successful except for 1 probe exchange failure. Image acquisition was 100% successful, with 90% of the nCLE images and 67% of the cystoscopy images rated as excellent quality. In all patients, diagnostic accuracy for nCLE in potentially malignant cysts was 87%, sensitivity was 77%, and specificity was 100%. In high certainty cases (n=18), the diagnostic accuracy of nCLE for potentially malignant cysts was 89%, sensitivity was 80%, and specificity was 100%. Pancreatitis developed in 2 people after the procedures. The procedures were carried out sequentially, so it is unclear which (if any) caused the complications.

Recent and ongoing studies

Seven ongoing or in-development trials on Cellvizio for pancreatic cysts were identified in the preparation of this briefing.

NCT01563133: a French interventional study designed to measure the diagnostic accuracy of Cellvizio in the characterisation of pancreatic cysts, lymph nodes near the gastrointestinal tract and pancreatic masses. This is an ongoing study with a study completion date of June 2017 (last updated April 2016).

NCT01770405: a US-based, prospective, observational study to assess the diagnostic performance of Cellvizio in diagnosing masses and cystic tumours of the pancreas, lymph nodes and submucosal lesions of the GI tract. This is an ongoing study with an estimated study completion date of September 2016.

NCT02523170: a UK-based, interventional study to determine the safety and efficacy of Cellvizio in patients with suspected cystic tumours of the pancreas. This is an ongoing study with an estimated study completion date of July 2017.

NCT02494388: a multinational, prospective, observational study to determine the diagnostic accuracy of Cellvizio in a cohort of patients with cystic pancreatic tumours. This is an ongoing study with an estimated study completion date of June 2017.

NCT01734967: a multinational, prospective, interventional study aiming to describe confocal imaging criteria for pancreatic masses, lymph nodes or liver metastases identified during EUS procedures for pancreatic cancer staging. It will also evaluate the feasibility and safety of examination using Cellvizio. This is an ongoing study with an estimated study completion date of December 2017.

NCT02516488: a US-based, prospective, observational study to assess the accuracy of Cellvizio for diagnosing cystic pancreatic lesions. This is an ongoing study with an estimated study completion date of November 2016.

TCTR20140402001: a Thai prospective, observational study to investigate the efficacy of Cellvizio for differentiating solid pancreatic masses. Recruitment was completed in February 2014.

Costs and resource consequences

There are no available data relating to the prevalence of pancreatic cysts in the UK population. There are more than 9,000 newly diagnosed cases of pancreatic cancer in the UK each year, with an incidence rate of around 14 per 100,000 people (Cancer Research UK 2016). It is estimated that 7,000 to 14,000 EUS procedures are needed in the UK each year for the diagnosing, assessing and staging likely pancreatobiliary cancers (Meenan et al. 2011). EUS is also available at 65 centres in the UK, with an average of 85 EUS‑FNA procedures done annually at each unit (Meenan et al. 2011). Cellvizio is designed to be used as an adjunct to EUS‑FNA and so this can provide some indication of potential NHS usage.

According to the manufacturer the device is being used in 3 NHS hospitals.

Because Cellvizio is used as an adjunct to EUS‑FNA there would be no need to change the way current services are delivered if the system were adopted. Training is needed to operate the system. No other additional facilities or technologies are needed alongside the technology.

No published evidence on the resource consequences of adopting Cellvizio for the relevant indication was identified in the systematic review of evidence.

Strengths and limitations of the evidence

In general, the studies considered in this briefing are small feasibility or pilot studies to establish technical feasibility, safety, diagnostic criteria, and diagnostic accuracy. All the studies included prospectively selected cohorts including 1 retrospective analysis of a prospectively selected cohort. No randomised controlled trials were identified.

Most sample sizes were relatively small, ranging from 18 (Konda et al. 2011) to 57 (Konda et al. 2013). The studies included 138 patients in total. No studies reported the use of a sample size calculation. Smaller sample sizes can lower the power and generalisability of a study.

Two studies were based in the USA (Konda et al. 2011; Nakai et al. 2015), 1 was a multinational study based in the US and France (Konda et al. 2013), and 2 studies were based in France (Napoleon et al. 2015a; Napoleon et al. 2015b). No published results of UK-based studies were found. The reported studies may not be generalisable to the NHS. A UK-based multicentre clinical trial is currently ongoing with an expected completion date of July 2017.

All 3 studies assessing diagnostic accuracy (Konda et al. 2013; Napoleon et al. 2015a; Nakai et al. 2015) used blinded-consensus review to compare Cellvizio with the reference standard. In Nakai et al. (2015), the investigator assessing image quality was not blinded to other clinical information. In Konda et al. (2013) and Napoleon et al. (2015a) investigators carrying out consensus review of images to obtain diagnostic criteria for pancreatic cysts were not blinded. Blinding minimises observer bias.

None of the studies carried out direct parallel comparisons of Cellvizio with EUS‑FNA. All studies assessed Cellvizio during an EUS‑FNA procedure. Only Nakai et al. (2015) provided an in‑study comparison, but each patient had EUS with cystoscopy (SpyGlass), followed by nCLE (Cellvizio), and then FNA sequentially within the same procedure. Tissue trauma caused by invasive imaging procedures may have affected the results of any subsequent tests.

Pancreatic cysts have heterogeneous pathology and this contributes to challenges in diagnosis. The reference standard used to produce a final diagnosis varied between the studies. In all 3 studies assessing diagnostic accuracy, the final diagnosis was based on either a stringent gold standard (histopathology or positive cytology) or a committee consensus based on all available information including clinical presentation, combined with image findings from EUS, CT, and MRI, as well as fluid analysis and cytology from EUS‑FNA. Study cohorts varied in the proportion of patients with a true criterion standard diagnosis by surgical histology compared with consensus review of non-surgical information. For example, in Nakai et al. (2015) only 2 of 30 patients had surgery and in Napoleon et al. (2015b) 20 of 33 patients had had surgery. This may produce challenges for direct comparison of the results because confidence levels, in particular for the less stringent methods, can vary.

The type of pancreatic cyst varies by gender. Serous cystadenomas and mucinous cystic neoplasms are predominantly found in women and intraductal papillary mucinous neoplasms and pseudocysts have similar prevalence in men and women. Nakai et al. (2015) and Napoleon et al. (2015a and 2015b) had more than twice the number of women than men in their sample, whereas in Konda et al. (2011 and 2013) the number of men and women included was about equal. In all studies, patients were recruited prospectively, but small sample sizes may result in cohorts being less representative of the wider population.

Cellvizio is designed to be used as an adjunct to standard EUS‑FNA, so adverse events with Cellvizio may be expected to be similar to EUS‑FNA. A prospective safety and feasibility study involving 128 people having EUS‑FNA reported an overall complication rate of 2.4%, which included pancreatitis (de Jong et al. 2011). But, Konda et al. (2011 and 2013), Nakai et al. (2015) and Napoleon et al. (2015a) showed complication rates of between 3.2% and 11%. The smaller sample sizes may have amplified the results. In addition, Nakai et al. (2015) carried out 3 sequential procedures in their study. It is unclear which procedure (if any) caused the complications. A review by Krishna et al. (2016) found that combining data from the clinical trials described in Konda et al. (2011 and 2013), and Napoleon et al. (2015a) resulted in an overall rate of pancreatitis after the procedure of 4.3%, which is higher than in the de Jong et al. (2011) study. If there is prolonged contact of the needle probe with pancreatic tissue during Cellvizio imaging compared with EUS‑FNA alone, this may be related to an increased risk of pancreatitis. The comparative safety of Cellvizio with EUS‑FNA needs further investigation.

The studies by Konda et al. (2011 and 2013) and Napoleon et al. (2015a and 2015b) were sponsored by the manufacturer. This introduces potential bias in the reporting of outcomes.