Non-physician endoscopists: A systematic review

Table 1 Summary table of all competency research regarding medical and nurse and other non-physician endoscopists

Publication	Setting	Non-physician proce-dures	Supervision	True randomization of patients?	Procedurists	Potential biases	Outcome parameters	Modality	Outcome
Rosevelt et al[8], 1984	Metropolitan Tertiary Centre, United States	825	Yes	No	1 NE	Patient selection biasEndoscopist selection biasLevel of assistance not documented	Polyp detection rateComplications	FS	Polyp detection rate of 8.7%No complications
Schroy et al[4], 1988	Metropolitan Tertiary Centre, United States	100	Yes	No	1 NE	Patient selection biasEndoscopist selection biasLevel of assistance not documentedConcordance criteria not given	Polyp detection rateConcordance of findings with expert opinion	FS	Polyp detection rate of 36%Video review showed k = 0.72 concordance with GCNurse sensitivity = 75%, specificity = 94%
DiSario et al[15], 1993	Metropolitan Tertiary Centre, United States	80	Yes	Yes	5 NEs5 GRs	Endoscopist selection biasLevel of assistance not documented1 NE excluded due to incompetency after trainingComparison of outcome groups for missed lesions not calculated	Insertion depthIdentification of anatomyComplicationsProcedure timeMissed lesions	FS	Polyp detection rate of 24%Insertion depth, complications and procedure time similar between groupsNEs missed more lesions and missed more anatomy
Maule[7], 1994	Metropolitan Tertiary Centre, United States	1881	Yes	No	4 NEs2 GCs	Endoscopist selection biasComplicated patient referred away from NEsLevel of assistance not documented	Insertion depthComplicationsPolyp detection ratePatient satisfaction	FS	GCs had significantly deeper insertion depthsSimilar polyp detection rateSimilar patient satisfaction
Moshakis et al[16], 1996	Metropolitan Tertiary Centre, United Kingdom	50	Yes	No	1 NE1 GC	Endoscopist selection biasPatient selection biasNE was compared to GC who performed trainingLevel of assistance not documentedMethod of quality scoring not given	Insertion depth“Quality and accuracy”	FS	Insertion depth, quality and accuracy were similar between comparison groups
Duthie et al[6], 1998	Metropolitan Tertiary Centre, United Kingdom	205	Not specified	No	1 NE	Endoscopist selection biasPatient selection biasCriteria for “successful procedure” not given	Successful procedure when compared to various other imaging modalities	FS	93% of procedures considered “successful”
Schoenfeld, Cash et al[17], 1999	Metropolitan Tertiary Centre, United States	114	Yes	No	1 NE3 GFs3 Surgical Consultants	Endoscopist selection biasPatient selection biasLevel of assistance not documented	Depth of insertionProcedure timePolyp detection ratePatient satisfactionComplications	FS	Surgeons had less depth of insertion than NEs or GFsNEs had longer procedures than GFs or surgeonsPolyp detection rate similarNo complications
Schoenfeld, Lipscomb et al[10], 1999	Metropolitan Tertiary Centre, United States	151	Yes	Yes	3 NEs4 GCs	Endoscopist selection biasPatient selection biasLevel of assistance not documentedHigh threshold for detecting difference in polyp detection rate	Polyp detection rateDepth of insertionComplications	FS	Polyp detection rates similar between groups (43%-45%)GCs had much greater depth of insertionNo complications
Wallace et al[9], 1999	Metropolitan Tertiary Centre, United States	2323	Yes	No	1 NE2 PAs15 GCs	Endoscopist selection biasPatient selection biasLevel of assistance not documented	Depth of insertionPolyp detection rateComplications	FS	GCs had significantly greater depths of insertion compared with NE+PAsPolyp detection rate similar between groups (23%-27%)No complications
Schoen et al[26], 2000	Metropolitan Tertiary Centre, United States	660	Not specified	No	1 PE1 GC1 MC	Endoscopist selection biasPatient selection biasLevel of assistance not documented	Patient satisfaction	FS	Similar patient satisfaction between groups
Shapero et al[27], 2001	Metropolitan Tertiary Centre, Canada	488	No	No	2 NEs	Endoscopist selection biasPatient selection bias Level of assistance not documented	Polyp detection rateComplicationsDepth of insertionProcedure time	FS	Average depth of insertion 52.9 cm8.4 min average procedure timePolyp detection rate of 15.4%
Jain et al[28], 2002	Metropolitan Tertiary Centre, United States	5000	No	No	Not specifiedNo physicians	Unknown endoscopistsPatient selection biasLevel of assistance not documentedNo comparison	Polyp detection rateNo complications	FS	No major complications polyp detection rate of 13.3%
Meenan et al[29], 2003	Metropolitan Tertiary Centre, United Kingdom	25	Yes	No	1 NE4 GFs	Endoscopist selection biasPatient selection biasLevel of assistance not documentedViews by NE were limited to the esophagus	Adequacy of views obtain by radial ultrasound by endoscopy	EGD	NEs had consistently lower quality scores
Smale et al[30], 2003	Metropolitan Tertiary Centre, United Kingdom	1487	Yes	No	2 NEs15 mixed medical/surgical physicians	Endoscopist selection biasPatient selection biasLevel of assistance not documentedComplicated patients excludedRetrospective and prospective	Differences in sedationPatient satisfaction	EGD	No difference between groups for sedation or patient satisfactionSubjectively, nurses reporter fewer studies as normal
Wildi et al[22], 2003	Metropolitan Tertiary Centre, United States	40	Yes	No	1 NE1GC	Endoscopist selection biasPatient selection biasLevel of assistance not documentedMethod of assessment not given	Concordance of findings with GC	EGD	NE had sensitivity of 75% and specificity of 98% with GC as gold standard
Nielsen et al[12], 2005	Metropolitan Tertiary Centre, Denmark	69	Yes	No	2 NEsUnknown number of physicians	Endoscopist selection biasPatient selection biasLevel of assistance not documented	Patient satisfaction	FS	Nurses had better patient satisfaction than physicians
Meining et al[3], 2007	Metropolitan Tertiary Centre, United Kingdom	190	Yes	Yes	2 NEs1 GC2 GRs1 Physician not specified1 MC	Endoscopist selection biasPatient selection biasLevel of assistance not documentedSubjectivey, nurses tended to focus on the entire examination whilst physicians focused on the reason for referral	Adequacy of views for entire procedureDuration of procedureUse of sedation	EGD	Nurses had twice the amount of adequate views however took twice as long on average.Nurses used sedation more frequently
Williams et al[19], 2006 Williams et al[21], 2009	Nation-wide Metropolitan Tertiary Centres, United Kingdom	957	Yes	Yes	30 NEs67 physicians (not specified)	Significantly higher numbers of patients changed schedule from physician to nurse (due	Patient satisfactionAdequacy of viewsDepth of insertionEndoscopic procedures performedDuration of examinationComplications	EGD	Patient satisfaction favoured nursesNo difference for depth of insertionNurses took biopsies for histology in upper endoscopy and FS more frequently than physicians.There were more normal histology findings for nurses
Richardson et al[20], 2009							Need for assistanceNeed for subsequent follow up and investigationCost-benefit analysis		Nurses were more likely to report sedation and procedural details whilst physicians were more likely to report diagnosis and suggested treatment.Nurses took biopsies for H. pylori more frequentlyNo major differences in final diagnoses frequency between 2 groups.No serious complicationsSimilar need for assistance.Nurses had greater follow-up cost per procedure whilst physicians had greater labor costs per procedure.Physicians had greater overall costs per procedure but greater patient improvement.Physicians were 87% more likely to be cost-effective than nurse endoscopists.
Koornstra et al[11], 2009	Metropolitan Tertiary Centre, Netherlands	300	Yes	No	2 NEs1 GF1 GC	Endoscopist selection biasPatient selection biasLevel of assistance not documented	Caecal intubation rateCaecal intubation timeComplicationsPatient satisfaction	Col	Similar caecal intubation rates/times (80%-90%) between GF and NEs but much lower/longer than GC after 150 procedures.Patient satisfaction similar between GF and NEs, less than for GC.Similar complication rate.
Maslekar, Hughes et al[31], 2010	Metropolitan Tertiary Centre, United Kingdom	308	Not specified	No	1 NE1 PA/TSeveral physicians not specified	Endoscopist selection biasPatient selection biasLevel of assistance not documentedNE and PA/Ts had less colonoscopies more FS	ComplicationsPatient satisfaction	Col and FS	No difference between all 3 groups
Maslekar, Waudby et al[32], 2010	Metropolitan Tertiary Centre, United Kingdom	26	Yes	No	1 Surgical Registrar1 PA/T	Endoscopist selection biasPatient selection biasLevel of assistance not documentedPatients needing resection excluded	Accuracy of endoscopists to gauge position in colon	FS	PA/T accuracy of 70% with Registrars accuracy of 80%, not statistically significant.
Shum et al[18], 2010	Metropolitan Tertiary Centre, HK	119	Yes	No	1 NE	No comparison group	Mean procedure timeDepth of insertionComplications	FS	9.4 min average procedure time53.5 cm average depth of insertionNo major complications
Limoges-Gonzalez et al[44], 2011	Metropolitan Endoscopy Centre, United States	50	Yes	Yes	1 NE2 GCs	Endoscopist selection biasLevel of assistance not documented	Adenoma detection rateCaecal intubation rateCaecal intubation timePatient satisfactionSedation useComplications	Col	Adenoma detection rate higher in NE (42%) than GCs (17%)All other parameters similar across both groups
de Jonge et al[33], 2012	Multi-metropolitan tertiary centre, Netherlands	162	Not specified	No (retrospective)	6 NEs113 Staff not specified including GCs, GFs, surgeons, MCs	Data was retrospective review of reports	Overall caecal intubation rateAdenoma detection rate	Col	NEs and GFs and GCs found more adenomas and had greater caecal intubation rates (94%) than nongastroenterology staff, especially surgical
van Putten et al[34], 2012	Multi-metropolitan tertiary centre, Netherlands	1000	Yes	No	10 NEs	Endoscopist selection biasPatient selection bias	Unassisted caecal intubation rateWithdrawal timeAdenoma detection rateAssistance requirementsPatient satisfactionComplications	Col	Unassisted caecal intubation rate of 94%23% of colonoscopies required assistance from GCWithdrawal time of 10 minAdenoma detection rate of 23%1 perforation and 1 onset of atrial fibrillation95% of patients satisfied with procedure
Massl et al[5], 2013	Multi-metropolitan tertiary centre, Netherlands	866	Yes	No	7 NEs8 GFs	Endoscopist selection biasNEs had significantly lower ASA scores on patientsLevel of assistance not reported	Unassisted caecal intubation rateCaecal intubation timeComplicationsPolyp detection rate	Col	Unassisted caecal intubation rate was significantly lower 77% for NE than GFs (88%).Polyp detection rate (45%), complications, withdrawal and intubation times were similar between groups.Crude cost-analysis showed a saving of €7.61 per colonoscopy where 1 GC supervises 3 NEs. Did not account for higher need for repeat colonoscopies due to incomplete procedures

Col: Colonoscopy; EGD: Esophagogastroduodenoscopy; FS: Flexible sigmoidoscopy; NE: Nurse endoscopist; GF: Gastroenterology fellow; GC: Gastroenterology consultant (attending); GR: Gastroenterology registrar/resident; PA/T: Physician’s assistant or technician; MC: Medical (non-gastroenterology) consultant.

Table 2 Assessment of the risk of bias assessments for studies included into this systematic review and key characteristics of data analysis

Publication	Sequence generation	Allocation concealment	Blinding of participants, personnel and outcome assessors	Incomplete outcome data	Selective outcome reporting	Other sources of bias	Study hypothesis and power calculation
Rosevelt[8], 1984	No randomization	No	No	Not specified	Likely, report was intended to describe a successful training program		No hypothesis, no statistics
Schroy et al[4], 1988	No randomization, review of videotape	No	No	Not specified	Report of an established service model. Review of videotapes Quality assurance		No statistics
DiSario[15], 1993	Computer generated randomization	Not specified	No	Not specified	Aim was to demonstrate that “.registered nurses could be trained to perform the flexible sigmoidoscopy in a similar to resident physicians’		Not powered to demonstrate equivalence, no formal power calculations
Maule[7], 1994	No randomization	Not specified	No	Not specified	The study was done to confirm that training of nurse endoscopists is feasible.		Hypothesis defined (no difference), no power calculation for equivalence study, oucome parameters not specified a priori
Moshakis et al[16], 1996	No randomization, no comparator	Not specified	No	Not specified	Report describes the successful training of one (1) nurse endoscopist		No hypotheses, no statistical analysis
Duthie et al[6], 1998	No randomization	No	No	Not specified	Evaluation of a training program that was developed and implemented by the authors (self-fulfilling prophecy)	Not evident	No hypothesis, no power calculation
Schoenfeld et al[17], 1999	No randomization, patients allocated to the ‘first available provider’	No	No	Not specified	No evidence	Not evident	Several outcome parameters specified, but no hypothesis tested, no power calculation for equivalence.
Schoenfeld et al[10], 1999	Randomization of veterans referred for flexible sigmoidoscopy. Computer generated randomization	No	unknown	Not specified	Justifies the implemented clinical service model.		Several outcome parameters listed, but no specific hypothesis, power calculation provided (to identify differences, but not targeting equivalence)
Wallace et al[9], 1999	No randomization, nurse-coordinator assigned eligible patients to a physician or non-physician endoscopists based upon ‘daily staffing assignments and patient time preference’	No	Unknown	Not specified	Justifies the implemented clinical service model.		No hypothesis stated, no power calculation
Schoen et al[26], 2000	No randomization	No	No	Not specified	Study targeted to demonstrate the good tolerability of flexible sigmoidoscopy	Gender distribution of patients was not equivalent across examiners, and the nurse practitioner did not have trainees working with her.	No hypothesis stated, no proper power calculation
Shapero et al[27], 2001	No randomization, allocation not clear	No	No	Not specified	Data justify the implemented clinical practice	Data are generated in the setting of CRC screening with flexible sigmoidoscopy, highly selective cohort.	Not done
Jain et al[28], 2002	No randomization	No	No	Not specified	Justification of implemented clinical practice	CRC screening utilizing flexible sigmoidoscopy, selective cohort	Not done
Meenan et al[29], 2003	No randomization	No	No	Not specified	Assessement of training progress		Not done
Smale et al[30], 2003	No randomization, part one retrospective analysis of endoscopy data base, second part prospective data collection	No	No	Not specified	Review and justification of clinical practice		Not done
Wildi et al[22], 2003	No randomization	No	No	Not specified	Sequential procedures Nurse endoscopist followed by physyician, potential effect of sequence.		Not done
Nielsen et al[12], 2005	No randomization	No	No	Not specified	Quality assurance of existing training program		Not done
Meining et al[3], 2007	No details in relation to the randomization process are provided. Patients unequally allocated to endoscopist or nurse	No	No	Reported but uneven numbers of ‘Randomization failures (33 vs 0). Considerable number of patients excluded (only 367 out of 641 reported)	Review and justification of clinical practice	Primary outcome parameter was stated as “appropriate diagnosis”, this outcome parameter was not reported.	Not done
Williams et al[19], 2006 Williams et al[21], 2009 Richardson et al[20], 2009	Randomization of patients to nurse or physician endoscopy	No	No	Properly reported	Primary outcome parameter not related to endoscopic. Measured with Gastrointestinal Symptoms Rating scale up to one year after procedure	Only patients suitable to be serviced by nurse endoscopists included. Numerically more patients from the nurse cohort were lost of follow-up without specified reasons (286 vs 269). A trend for more patients with weight loss in the physicians cohort, more patients in the physicians’’ cohort had previously barium enema (suggesting more chronic or relapsing symptoms)	Authors make reference to required sample sizes. Total number of patients completed was below the required sample size
Koornstra et al[11], 2009	It is stated that patients were randomly allocated, no information is given on allocation. Proportion of inpatients lower in the nurse group. No evidence for ethic approval or consent of patients. Training of nurse and medical staff was not identical	No	No	No information provided	Multiple endpoints reported	The authors developed a training program and with their data they aimed to confirm that their training program delivered (self-fulfilling prophecy).	Not powered to verify equivalence
Maslekar et al[31], 2010	Patients were allocated by administrative staff into the nurse or medical specialist group.	No	No	Incomplete response data cited as reason for exlusion (48/561 excluded), no intention to treat analysis	Study justifies an implemented service model that aims to address shortage of medical specialists	The instrument was unlikely to detect group differences. Variable mixture of flexible sigmoidoiscopy and colonoscopy across groups	No power calculation
Maslekar et al[32], 2010	No randomization	No information given	No	Not reported	Justifies implemented service and training model	For flexible sigmoidoscopies the validity of the endoscopists impression of maximal extension was tested. A priori unlikely to identify difference.	No power calculation
Shum et al[18], 2010	No randomizatiom, no comparator	No	No	No information provided	Justifies the implemented training model		No
Limoges-Gonzalez et al[44], 2011	It is stated that patients were randomly allocated, no information is given on allocation.	No	No	No information provided	Justifies the implemented service model	Postprocedure questionnaire was administered after (at least) 30 min of recovery. Drug effects likely to blunt potential differences	No power calculation
de Jonge et al[33], 2012	Routine quality data were used, no randomization.	No	No	No information provided	Data were partly retrospective data, partly prospective data, no justification given.	?	No power calculation
van Putten et al[34], 2012	Allocation of patients by secretatial staff, no randomization	No	Not reported	Not specified	Justifies and implemented service model	Significant differences in comorbidities (more severe in the Gastroenterologists group), differences in source of referral. Outcome assessment limited to immediate salary comparisons not total costs including pathology and follow-up.	No power calculation
Massl et al[5], 2013	It is stated that patients allocated by administrative staff, endoscopists assigned to lists randomly based on availability	No	No	79/2025 procedures not included due to drop out of 1 nurse endoscopist for unspecified reasons	Justifies the implemented service model	Patients younger than 18 years or referred for therapeutic procedures were excluded from the nurse endoscopist group only. Drop out of nurse endoscopist not justified.	Power calculation done.Appropriate numbers achieved.

Red cell = high risk of bias, yellow cell = unclear risk of bias, green cell = low risk of bias. CRC: Colorectal cancer.