Ambulatory AnesthesiologySection Editor: Peter S. A. Glass

A Randomized, Controlled, Double-Blind Trial ofPatient-Controlled Sedation with Propofol/RemifentanilVersus Midazolam/Fentanyl for Colonoscopy

Jeff E. Mandel, MD, MS*

Jonathan W. Tanner, MD, PhD*

Gary R. Lichtenstein, MD†

David C. Metz, MD†

David A. Katzka, MD†

Gregory G. Ginsberg, MD†

Michael L. Kochman, MD†

BACKGROUND: Patient-controlled sedation (PCS) with propofol has been advocatedas a method for dealing with the narrow therapeutic window for moderatesedation, but previous studies have methodologic limitations. We hypothesizedthat, by using remifentanil in conjunction with propofol and using PCS in botharms of the study, we could demonstrate marked improvements in facility usecompared with fentanyl plus midazolam.METHODS: Fifty patients undergoing elective colonoscopy were randomized (withconcealed allocation) to midazolam/fentanyl (group MF) or propofol/remifentanil(group PR) administered via PCS. Time intervals for sedation and recovery,perceptions by patient, nurse, and gastroenterologist, and need for anesthesiologistintervention were assessed.RESULTS: Group PR patients were sedated and recovered significantly more rapidlythan did group MF (P � 0.0001). In the group PR, recovery room time was actuallyshorter than procedure room time. Patient, nurse, and gastroenterologist percep-tions were equivalent between the groups. Two patients in group PR requiredanesthesiologist intervention for arterial desaturation exceeding the primary safetyend point.CONCLUSIONS: PCS with propofol/remifentanil yields superior facility throughputcompared with midazolam/fentanyl when used in an appropriate care setting.(Anesth Analg 2008;106:434–9)

Colorectal cancer is the second most common causeof cancer-related mortality in the United States, ac-counting for 55,000 deaths annually.1 Use of colonos-copy for screening has been advocated as an effectivemeans for reducing deaths from colon cancer. Anestimated 14.5 million colonoscopies were performedin the United States in 2002.2 In the United States,colonoscopy is considered to be sufficiently uncom-fortable as to require sedation. Sedation is not withoutrisks, and complications attributable to sedation mayexceed technical complications by a factor of 10.3

Sedation with propofol has been advocated as analternative to the more commonly used combinationof midazolam and narcotic sedation. Propofol is asso-ciated with faster onset and offset of sedation and

higher patient satisfaction, but may require resuscita-tion maneuvers more often than conventional seda-tion. Patient-controlled sedation (PCS) has been advocatedto improve titration of sedation with propofol. Theapproach taken in previous studies of PCS is toattempt to minimize the respiratory and hemody-namic side effects of propofol, whereas the currentstudy attempts to maximize the time savings associ-ated with propofol sedation by the addition of signifi-cant doses of remifentanil and use of an anesthesiologistfor rescue when needed. We used PCS in both arms ofthe study to avoid bias in the dosage of midazolamand fentanyl, and evaluated the equivalence of seda-tion by assessment of blinded participants. Our hy-pothesis was that the combination of propofol andremifentanil would significantly reduce the time tosedation and ambulation without assistance when com-pared with midazolam and fentanyl at equivalent levelsof patient, nurse, and gastroenterologist satisfaction.

METHODSThe protocol was approved by the IRB of the

University of Pennsylvania School of Medicine. Fiftypatients scheduled for elective outpatient colonoscopywith conscious sedation were recruited from the prac-tices of the gastroenterologist investigators. Patientsknown to have failed sedation with conventionaldrugs or to have significant comorbidity were not

From the Departments of *Anesthesiology and Critical Care, and†Medicine, University of Pennsylvania School of Medicine, Phila-delphia, Pennsylvania.

Accepted for publication September 19, 2007.Reprints will not be available from the author.Address correspondence to Jeff E. Mandel, MD, MS, Department

of Anesthesiology and Critical Care, 780 B Dulles Building, Hospitalof the University of Pennsylvania, 3400 Spruce St., Philadelphia, PA19104. Address e-mail to mandelj@uphs.upenn.edu.

Copyright © 2008 International Anesthesia Research SocietyDOI: 10.1213/01.ane.0000297300.33441.32

Vol. 106, No. 2, February 2008434

explicitly excluded from participation. After recruit-ment, patients were randomized to one of two groupsby sequentially numbered envelopes. The anesthesi-ologist prepared a 60-mL syringe loaded with themixture of study drugs and programmed a Graseby3300 syringe pump (Marcal Medical, Millersville, MD)as shown in Table 1.

The midazolam-fentanyl mixture was prepared bycombining 10 mL of midazolam (1 mg/mL), 5 mL offentanyl (50 �g/mL), and 5 mL of saline. An initialbolus of 4 mL and demand dose of 1 mL with 1 minlockout were programmed. The propofol-remifentanilmixture was prepared by dissolving 1 mg remifentanilin 2.5 mL of propofol (10 mg/mL) and adding 1 mL ofthis mixture to 39 mL of propofol. The admixture wasperformed immediately before initiation of the study;no evidence of deemulsification of propofol wasnoted. An initial bolus of 2.5 mL and demand dose of0.75 mL with a zero lockout were programmed. Theactual lockout for group propofol-remifentanil (PR)was dependent on bolus and syringe size; for the 60-mLsyringe with 0.75-mL bolus, this was approximately 13 s.

The infusion pump was connected to a pigtailadapter placed at the hub of the IV catheter with adead-space of �0.1 mL. An infusion set with a checkvalve was used to avoid retrograde flow. The carrierfluid was observed to assure continuous infusionwithout obstruction.

All patients were monitored according to standardpractice in our institution, with continuous pulseoximetry, electrocardiogram, and noninvasive arterialblood pressure measured at 5-min intervals, and oxy-gen 2 L/min was administered by nasal cannula. AnAdvanced Cardiac Life Support certified nurse waspresent throughout the procedure, with primary re-sponsibility for record keeping and patient safety. Thenurse was permitted to undertake normal measures todeal with desaturation, including verbal stimulationand chin lift.

Patients were instructed to press the button as oftenas they felt the need to do so, as there was a lockout tolimit the amount of medication and an anesthesiolo-gist to intervene. Nurses were instructed to respond tothe question “should I press the button” with theanswer “If you want to press the button, do so.”Patients were instructed that after the procedure, wewanted to see how quickly they could ambulate, and

that a research assistant would be with them to assessthis.

Blinding was achieved by covering the clearLexan™ shell of the pump with opaque labels, andshrouding the microbore infusion tubing with opaqueplastic sheathing used for chemotherapeutic infusions.The gastroenterologist, nurse, and patient were un-aware of the lockout period. The anesthesiologist wasblinded before allocation, but was aware of the treat-ment regimen while present to intervene for adverseevents. Safety end points included arterial desatura-tion �85% for more than 60 s, hypotension �90 mmHg systolic or 20% decrease from baseline persistingon repeat determination 1 min later, or inability totolerate the procedure.

Times were recorded using a synchronized timesource for the following events: initiation of sedation,insertion of the colonoscope, cecal intubation, removalof colonoscope, clearance for transportation to postan-esthesia care unit (PACU), and ability to walk withoutassistance. Initiation of sedation was the time at whichthe pump was activated to administer the initialloading dose. Insertion of the colonoscope was deter-mined by the gastroenterologist based on clinicalassessment of patient readiness. Cecal intubation andremoval of colonoscope were announced by the gas-troenterologist. Clearance for transportation was de-termined by the endoscopy suite nurse. Assessment ofreadiness to ambulate was made by a blinded ob-server who remained with the patient for the durationof the PACU stay. The observer was instructed thatthe patients were required to be able to walk as if theywere leaving the PACU unescorted. The intent of theselection of time points was to closely resemble thenormal clinical workflow.

The collected times were used to derive the follow-ing intervals:

Procedure time � Colonoscope out–Colonoscope inTime to Sedation � Colonoscope in–Initiate sedationTime to Recovery � Ambulation–Colonoscope outProcedure room time � Transport–Initiate sedationPACU time � Ambulation–Transport

Satisfaction of gastroenterologist, nurse, and pa-tient were obtained by a 7-point Likert scale. Thegastroenterologist and nurse were asked to evaluatetwo statements: “The patient was adequately sedated”

Table 1. Medication Regimens

Group midazolam/fentanyl (MF) Group propofol/remifentanil (PR)Drug content in syringe Fentanyl 250 �g Remifentanil 400 �g

Midazolam 10 mg Propofol 400 mgSaline 5 mL

Concentrations Fentanyl 12.5 �g/mL Remifentanil 10 �g/mLMidazolam 0.5 mg/mL Propofol 10 mg/mL

Volume in syringe 20 mL 40 mLInitial bolus 4 mL 2.5 mLDemand dose 1 mL 0.75 mLLockout 1 min 13 s

Vol. 106, No. 2, February 2008 © 2008 International Anesthesia Research Society 435

(statement 1) and “The patient was safe” (statement 2).Patients were asked to evaluate three statements: “Ifelt comfortable with controlling my own sedation”(statement 3), “I was comfortable throughout theprocedure” (statement 4), and “I would have thisanesthetic again, and encourage others to do so”(statement 5).

StatisticsA sample size estimate was calculated by using

previously published data comparing PCS propofol/alfentanil to conventional sedation.4 A study com-prised of 20 patients per group, and a P � 0.05 wasdetermined to have a 95% power. We sought approvalfor 25 patients per group to insure against incompletestudies due to adverse events and procedures aborteddue to inadequate bowel preparation.

Patients were randomly allocated to two groups of25 by the randperm function of MATLAB (The Math-works, Natick, MA). A blocking of 10 and 40 was usedto permit safety review after the first 10 cases.

Times were compared by two-sided Student’s t-testfor two samples assuming equal variance, as weredemographic data such as age and body mass index(BMI), and presented as mean � sd. Comparison ofsatisfaction scores was performed by ordered logitregression for group, gender, age, height, weight, andBMI. Linear regression of sedation and ambulationtimes was performed to examine possible confound-ing by gender, age, height, weight, BMI, and gastro-enterologist. All statistics were performed with STATA 8(Stata Corp, TX).

RESULTSForty-nine patients completed the study; one pa-

tient was excluded for noncompliance with bowelpreparation. The groups were equivalent for age, BMI,gender, and number of ASA III patients. There wereno significant differences between groups in the num-ber of polypectomies, the number of patients requiringposition change to achieve cecal intubation, and inoverall procedure time (exclusive of time to sedation).These values are summarized in Table 2.

Patients in group midazolam-fentanyl (MF) con-sumed 6.4 � 2 mg midazolam and 161 � 49 �gfentanyl (mean � sd). Patients in group PR consumed115 � 42 mg propofol and 115 � 42 �g remifentanil.

Times to sedation and ambulation and time spent inthe procedure room and recovery room are summa-rized in Figure 1.

Patients in group MF required significantly moretime to achieve sedation—7.6 � 3.6 min vs 3.4 � 1.3min in group PR. Time to ambulation without assis-tance was also longer in group MF—36.4 � 5.3 min vs9.2 � 4.0 min. Recovery room time was longer ingroup MF—32 � 25 min vs 4.9 � 4.3 min. Thesedifferences were significant at P � 0.0001. Age, gen-der, height, weight, BMI, and gastroenterologist werenot confounders for these intervals. Procedure roomtime was not significantly different between groups.Of interest, recovery room time was appreciablyshorter than procedure room time for group PR.

Agreement with statements 1 and 2 by nurses andgastroenterologists and statements 3, 4, and 5 bypatients were analyzed by ordered logit regression.No correlation was found between ratings and group,gender, age, height, weight, or BMI. Of the 343 re-sponses, only 14 ratings were lower than neutral.

Figure 1. Times to sedation and ambulation and time spent inthe procedure room and recovery room. Group propofol/remifentanil (PR); group midazolam/fentanyl (MF).

Table 2. Group Comparison

Group midazolam/fentanyl(MF) (24)

Group propofol/remifentanil(PR) (25)

Age (mean � sd) 57.7 � 10.8 60.5 � 9.6Body mass index (kg/m2) (mean � sd) 28.1 � 6.0 28.6 � 5.9Gender (male/female) 11/13 13/12ASA III 1 2Polypectomies 10 10Position change 13 13Procedure duration (min) (mean � sd) 21 � 12.3 19 � 9.9

436 Patient-Controlled Sedation for Colonoscopy ANESTHESIA & ANALGESIA

More group PR patients had low safety ratings, moregroup MF patients had low satisfaction ratings.

The study was not designed to detect differences inthe rate of cardiopulmonary complications, but endpoints for anesthesiologist intervention were requiredfor study approval. There were no significant differencesin hemodynamic safety end points (two in group PR,one in group MF). In group PR, two patients requiredintervention by the anesthesiologist, in both cases, twobreaths of 100% oxygen via a Mapleson circuit andfacemask. Both of these episodes occurred shortly aftercecal intubation, when stimulation diminished mark-edly, and both patients were awake and responsive toverbal commands within 60 s of the intervention. Nopatients in group MF required intervention. There wereno technical complications of colonoscopy, and all pa-tients completed colonoscopy without non-PCS seda-tion.

The success of blinding could not be assessed withcomplete certainty. Although gastroenterologists andnurses would periodically express opinions as towhich arm the patient was in, the only primary datacollection point subject to observer bias was colono-scope insertion time, which generally preceded thetime at which the gastroenterologist would have anyopinion on the form of sedation. Patients uniformlyfailed to express an opinion as to the arm of the studyinto which they perceived to be allocated.

DISCUSSIONThe practice of sedation for procedures such as

colonoscopy exposes the patient to risk of morbidityand mortality, and appropriate safeguards for patientsafety must be preserved. Economic pressure to de-liver care under cost constraints has led to calls forchanges in Food and Drug Administration labeling ondrugs such as propofol to permit their use in settingsother than the traditional anesthesia care model.5

Although involvement of anesthesia providers incolonoscopy has been demonstrated to reduce sedation-related adverse events,6 concern related to the costs ofsuch sedation practices led the American College ofGastroenterology, the American GastroenterologicalAssociation, and the American Society for Gastroin-testinal Endoscopy to state their position that “theroutine assistance of an anesthesiologist/anesthetistfor average risk patients undergoing standard upperand lower endoscopic procedures is not warranted.”7

Current trends suggest a reduction in reimbursementfor anesthesia-directed procedural sedation8 at a timewhen many anesthesia departments are unable toprovide coverage for the current volume of proceduralsedation.9 A demonstration that anesthesiologist in-volvement in sedation practices might increase facilityefficiency and reduce overall cost of care would beuseful information in guiding policy on an importantpublic health issue.

The use of propofol by nonanesthesiologists has beenreported in numerous prospective studies, and has been

associated with significantly shorter time to sedation andrecovery with higher patient satisfaction.10

The narrow therapeutic window for moderate se-dation with propofol has prompted use of PCS toavoid unintended general anesthesia. In patients ran-domized to propofol-alfentanil PCS or midazolam-pethidine, the rate of over-sedation was lower withPCS, time to discharge was shorter, and all patientsexpressed satisfaction with their care.4 These resultshave been confirmed in other reports.11,12 PCS withmidazolam has been compared with PCS with propo-fol in patients undergoing third molar extraction13 andcataract extraction.14 Patient satisfaction was equiva-lent, but recovery time was shorter with propofol.

The limitations of previous studies include themanagement of narcotics and lack of concealed allo-cation. Fentanyl and demerol have been given at thediscretion of the physician, and alfentanil doses givenin conjunction with propofol fall short of those neces-sary to reduce propofol requirements. Pharmacoki-netic simulation of recovery time from infusions ofpropofol and alfentanil has demonstrated that thefastest recovery is achieved using the lowest feasibledose of propofol relative to alfentanil, and lack ofsensitivity of wakeup time to infusion duration whenremifentanil is used as the narcotic.15 We used fixedratios of narcotic and sedative to permit a singlesyringe to be used in each arm of the study, eliminat-ing the need for discretionary narcotic use and facili-tating concealed allocation. The ratio of remifentanil topropofol was high enough to expect a propofol-sparing effect, but the optimality of the ratio was notassessed.

The primary end points in this study were timeintervals. Time to sedation and time to recovery weresignificantly shorter in group PR, and differences weremore striking than those reported in previous efforts.Ambulation time was shorter than procedure time,which might permit reduced PACU staffing ratiosfrom the current recommendations of 3 or 4 PACUslots per endoscopy suite.16 The magnitude of the costsavings with such an approach is currently unknown.

Perceptions were a secondary end point in thestudy. There were minor differences in perceptionsbetween groups, and between nurses and gastroenter-ologists, but no significant differences were found.Patient satisfaction was typically high. There was onepatient in group MF who was dissatisfied with theprocess of controlling his or her own sedation,whereas all other patients were accepting. The extentto which this result can be extended to a widerpopulation is unclear. Satisfaction with patient-controlled analgesia has been demonstrated to beaffected by the psychological test locus of control.17

PCS applied as the sole option for sedation wouldlikely result in lower satisfaction scores.

The doses of midazolam and fentanyl exceededthose typically administered by the same gastroenter-ologists during routine conscious sedation. The doses

Vol. 106, No. 2, February 2008 © 2008 International Anesthesia Research Society 437

we report for propofol are comparable with those re-ported by Bright et al.4 using PCS propofol/alfentanil,and approximately half that reported by Sipe et al.10

using nurse-administered propofol sedation.The primary safety end point was arterial desatu-

ration below 85% for 60 s. Although 90% saturation isan often-used criterion for significant desaturationduring sedation, there appears to be no validation ofthis cutoff. Moller et al.,18 in an observational study of296 patients, noted episodes of desaturation between86% and 90% in 53% of patients that were undetectedby anesthesiologists in 70% of cases, and when de-tected, were discovered on average after 70 s. Sever-inghaus et al.19 subjected 50 volunteers to deliberatehypoxia intended to yield arterial saturation of 55%for 30–45 s without deleterious effects. Thus, thesafety end point seemed only slightly below a levelthat most anesthesiologists would fail to detect byclinical signs, and significantly above levels intention-ally produced by Severinghaus et al. The primarysafety end point was intended to mimic a process inwhich the nurse would initiate a response to desatu-ration with backup from an anesthesiologist in anearby location. Desaturation requiring supplementaloxygen is a relatively common occurrence in colonos-copy; a prospective study of 3196 patients noted anincidence of 4.4%.20

The combination of propofol and remifentanil hasbeen demonstrated to yield significant respiratorydepression.21 This propensity for respiratory adverseevents was seen in this study; in two patients in groupPR, the anesthesiologist intervened due to the safetyend point. In both instances, the desaturation occurredimmediately subsequent to cecal intubation and re-solved promptly without heroic measures. We con-clude that the presence of an anesthesiologist wasbeneficial. The frequency of intervention was entirelya function of the specified end point. The recoveryfrom respiratory depression with the propofol-remifentanil mixture is rapid. Had a lower saturationthreshold been chosen, or a longer time, or hadpatients been administered oxygen, the rate of inter-vention may have been different, perhaps zero. Theprimary safety end point was not intended to measurerespiratory depression but, rather, the ability of pa-tients to recover from transient desaturation throughlow-skill interventions by Advanced Cardiac Life Sup-port certified nurses. Clinicians would not normallyelect to defer intervention during desaturation for 60 s.We cannot recommend this form of sedation to thosenot prepared to administer resuscitative measurespromptly, and do not suggest that the methods usedin this study are standard clinical practice.

The applicability of this form of sedation to otherendoscopic procedures, or sedation in general, isunknown. An efficient sedation protocol could signifi-cantly improve treatment applicability by increased

throughput in procedure rooms, greater patient accep-tance, and decreased PACU costs. The extent to whichPCS with anesthesiologist rescue can form the basis ofa safe, cost-effective alternative to a two-tiered systemof conscious sedation by nurses and deep sedation ina monitored anesthesia care model will require furtherstudy.

ACKNOWLEDGMENTSE. Andrew Ochroch, MD, MPH, assisted in statistical

analysis. Bhavisha Dhillon, RN, Laura Chang, BS, andMin Wang, BA, assisted in postprocedure assessment. Dr.Mandel acknowledges honoraria for lectures from AbbottLaboratories.

REFERENCES

1. Greenlee RT, Hill-Harmon MB, Murray T, Thun M. Cancerstatistics, 2001. CA Cancer J Clin 2001;51:15–36

2. Seeff LC, Richards TB, Shapiro JA, Nadel MR, Manninen DL,Given LS, Dong FB, Winges LD, McKenna MT. How manyendoscopies are performed for colorectal cancer screening?Results from CDC’s survey of endoscopic capacity. Gastroen-terology 2004;127:1670–7

3. Sieg A, Hachmoeller-Eisenbach U, Eisenbach T. Prospectiveevaluation of complications in outpatient GI endoscopy: asurvey among German gastroenterologists. Gastrointest Endosc2001;53:620–7

4. Bright E, Roseveare C, Dalgleish D, Kimble J, Elliott J, ShepherdH. Patient-controlled sedation for colonoscopy: a randomizedtrial comparing patient-controlled administration of propofoland alfentanil with physician-administered midazolam andpethidine. Endoscopy 2003;35:683–7

5. Rex DK. Review article: moderate sedation for endoscopy:sedation regimens for non-anaesthesiologists. Aliment Pharma-col Ther 2006;24:163–71

6. Vargo JJ, Holub JL, Faigel DO, Lieberman DA, Eisen GM. Riskfactors for cardiopulmonary events during propofol-mediatedupper endoscopy and colonoscopy. Aliment Pharmacol Ther2006;24:955–63

7. Rex DK. Three challenges: propofol, colonoscopy by under-trained physicians, and CT colonography. Am J Gastroenterol2005;100:510–3

8. Hannenberg AA. Payment for procedural sedation. Curr OpinAnaesthesiol 2004;17:171–6

9. Lalwani K. Demographics and trends in nonoperating-roomanesthesia. Curr Opin Anaesthesiol 2006;19:430–5

10. Sipe BW, Rex DK, Latinovich D, Overley C, Kinser K, BratcherL, Kareken D. Propofol versus midazolam/meperidine foroutpatient colonoscopy: administration by nurses supervised byendoscopists. Gastrointest Endosc 2002;55:815–25

11. Heuss LT, Drewe J, Schnieper P, Tapparelli CB, Pflimlin E,Beglinger C. Patient-controlled versus nurse-administered seda-tion with propofol during colonoscopy. A prospective random-ized trial. Am J Gastroenterol 2004;99:511–8

12. Lee DW, Chan AC, Sze TS, Ko CW, Poon CM, Chan KC, SinKS, Chung SC. Patient-controlled sedation versus intrave-nous sedation for colonoscopy in elderly patients: a prospec-tive randomized controlled trial. Gastrointest Endosc2002;56:629 –32

13. Rudkin GE, Osborne GA, Finn BP, Jarvis DA, Vickers D.Intra-operative patient-controlled sedation. Comparison ofpatient-controlled propofol with patient-controlled midazolam.Anaesthesia 1992;47:376–81

14. Pac-Soo CK, Deacock S, Lockwood G, Carr C, Whitwam JG.Patient-controlled sedation for cataract surgery using peribul-bar block. Br J Anaesth 1996;77:370–4

15. Vuyk J, Mertens MJ, Olofsen E, Burm AG, Bovill JG. Propofolanesthesia and rational opioid selection: determination ofoptimal EC50-EC95 propofol-opioid concentrations that as-sure adequate anesthesia and a rapid return of consciousness.Anesthesiology 1997;87:1549 – 62

16. Marcus SN. Efficiency in endoscopy centers. Gastrointest En-dosc 2006;64:765–7

438 Patient-Controlled Sedation for Colonoscopy ANESTHESIA & ANALGESIA

17. Johnson LR, Magnani B, Chan V, Ferrante FM. Modifiers ofpatient-controlled analgesia efficacy. I. Locus of control. Pain1989;39:17–22

18. Moller JT, Johannessen NW, Berg H, Espersen K, Larsen LE.Hypoxaemia during anaesthesia—an observer study. Br J An-aesth 1991;66:437–44

19. Severinghaus JW, Naifeh KH, Koh SO. Errors in 14 pulseoximeters during profound hypoxia. J Clin Monit 1989;5:72– 81

20. Nelson DB, McQuaid KR, Bond JH, Lieberman DA, WeissDG, Johnston TK. Procedural success and complications oflarge-scale screening colonoscopy. Gastrointest Endosc 2002;55:307–14

21. Nieuwenhuijs DJ, Olofsen E, Romberg RR, Sarton E, Ward D,Englbers F, Vuyk J, Mooren R, Teppema LJ, Dahan A.Response surface modeling of remifentanil-propofol interac-tion on cardiorespiratory control and bispectral index. Anes-thesiology 2003;98:312–22

Vol. 106, No. 2, February 2008 © 2008 International Anesthesia Research Society 439