Safety of midazolam for sedation of HIV-positive patients undergoing colonoscopy


Correspondence: Elke S. Backman, PharmD, Department of Pharmacy, Massachusetts General Hospital, 55 Fruit Street, GRB-005, Boston, MA 02114, USA. Tel: 617 726 2503; fax: 617 726 7653; e-mail:



Because of concerns regarding interactions between midazolam and antiretroviral therapy (ART), alternative sedatives are sometimes used during procedural sedation. Our objective was to compare outcomes in patients on ART who received intravenous (iv) midazolam vs. iv diazepam, a second-line agent, during colonoscopy.


We conducted a retrospective analysis of adult HIV-positive patients who underwent colonoscopy over a 3.5-year period. Primary outcomes were sedation duration, nadir systolic blood pressure (SBP), nadir oxygen saturation, abnormal cardiac rhythm, and change in level of consciousness using a standardized scale. We calculated rates of adverse events according to benzodiazepine use and identified risk factors for complications using univariate and multivariate analyses.


We identified 136 patients for this analysis: 70 received midazolam-based sedation and 66 received a diazepam-based regimen. There were no significant differences between the two groups with respect to sedation duration (mean 48.0 vs. 45.7 minutes for the midazolam and diazepam groups, respectively; P = 0.68), nadir SBP (mean 97.0 vs. 101.6 mmHg; P = 0.06), nadir oxygen saturation (mean 94.6 vs. 94.8%; P = 0.72) or rate of abnormal cardiac rhythm (11.4 vs. 19.7%; P = 0.18). More patients in the midazolam group experienced a depressed level of consciousness (91% vs. 74% in the diazepam group; P = 0.0075), but no patient required reversal of sedation or became unresponsive.


We did not find evidence that patients who received midazolam for procedural sedation had clinical outcomes statistically different from those who received diazepam. These findings should be confirmed in prospective studies or in a randomized controlled trial.


Midazolam is a rapidly acting benzodiazepine widely used as a sedative and anaesthetic induction agent in patients undergoing colonoscopy. However, because of concerns regarding potential drug interactions between midazolam and antiretroviral therapy (ART), some clinicians use general anaesthesia instead in HIV-positive patients undergoing colonoscopy. Other providers have patients stop ART for 1 to 2 days before a planned procedure, although interrupting ART is not a recommended strategy from the perspective of HIV disease management [1]. Diazepam, an alternative agent, utilizes multiple metabolic pathways and is less likely to interact with HIV protease inhibitors (PIs); however, midazolam is favoured by most clinicians because of its more rapid onset of action, shorter half-life (1.5 h vs. 20 h for diazapem), amnestic effects, and lower incidence of pain on injection or phlebitis compared with diazapem [2-4].

We sought to evaluate the safety of intravenous (iv) midazolam in HIV-positive patients on ART who received procedural sedation for colonoscopy. Our objective was to compare clinical outcomes and identify risk factors for complications in patients who received midazolam vs. diazepam.


Study population

HIV-positive patients who were 18 years of age or older and who underwent out-patient colonoscopy at Massachusetts General Hospital between 1 January 2007 and 30 June 2010 were identified through the Research Patient Data Registry (RPDR), a clinical database for the Partners Health-Care System. Inclusion criteria were: (1) completion of colonoscopy; (2) receipt of ART for a minimum of 4 weeks at the time of the colonoscopy; (3) receipt of midazolam- or diazepam-based sedation.

Data collection

Gastrointestinal (GI) endoscopy procedure flow sheets were reviewed for each patient. Flow sheets included scores assigned by the providers for selected physiological parameters. Demographic data, past medical history, and medication data were documented for each patient and confirmed via review of the electronic medical record.


Primary outcomes were sedation duration, nadir SBP, nadir oxygen saturation, rate of abnormal cardiac rhythm, and change in level of consciousness. Level of consciousness was assessed using an institutional scale on which a score of 2 represents awake and oriented, 1 represents arousable or disoriented, and 0 represents unresponsive. Secondary outcomes included change from baseline in respiratory rate, heart rate, airway status, quality of respirations, skin perfusion, SBP, pain score, procedure time, and total monitoring time required as documented in the post-procedure report.

Statistical analysis

Univariate analyses of associations between exposures and outcomes were performed using Student's t-test for continuous variables and the Wald χ2 test or Fisher's exact test for categorical variables. Multivariate analyses were completed using logistic regression models. Covariates included in the analyses were concomitant cytochrome (CYP) 3A4 inhibitors, chronic opioid use, chronic benzodiazepine use, liver disease, age and gender. In a separate analysis, outcomes were analysed by four treatment groups: PI/midazolam; PI/diazepam; non-PI/midazolam; non-PI/diazepam. A P-value < 0.05 was considered to be statistically significant.


Patient characteristics

We identified 344 HIV-positive patients receiving ART who were scheduled for a colonoscopy. Of this group, 208 were excluded for a variety of reasons; the most common reason for exclusion was cancellation or postponement of the procedure by the patient (n = 127). The excluded patients were similar to the study group with respect to age (median 52 years), sex (82% male), and CD4 T-cell count (mean 571 cells/μL; P = 0.82). The final cohort consisted of 136 HIV-positive patients on ART who underwent out-patient colonoscopy with midazolam- or diazepam-based sedation (Fig. 1).

Figure 1.

Consort diagram. ART, antiretroviral therapy; EGD, oesophagogastroduodenoscopy.

Demographic characteristics in the two groups were similar (Table 1). A greater proportion of patients in the midazolam group had liver disease (mostly viral hepatitis) compared with the diazepam group (31% vs. 17%, respectively; P = 0.04). Most liver disease was attributable to hepatitis C virus infection (72%; liver disease was mild to moderate in these patients). The remainder had a diagnosis of steatohepatitis or alcoholic hepatitis. The groups were similar with respect to physical characteristics, although there was a trend towards higher body mass index (BMI) in the midazolam group (mean 28 vs. 26 kg/m2 in the diazepam group; P = 0.05). No patient had significant renal disease. PI-based ART was used by 53% of patients in the midazolam group and 61% of patients in the diazepam group (P = 0.55).

Table 1. Baseline characteristics of HIV-positive patients who underwent out-patient colonoscopy receiving midazolam- or diazepam-based sedation
CharacteristicsMidazolam (n = 70)Diazepam (n = 66)
  1. Data are number (%) of patients, unless otherwise indicated.
  2. ARV, antiretroviral; BMI, body mass index; CYP, cytochrome P450; NNRTI, nonnucleoside reverse transcriptase inhibitor; NRTI, nucleoside reverse transcriptase inhibitor; PI, protease inhibitor; SBP, systolic blood pressure; SD, standard deviation.
  3. One patient in each group was on a double-PI-containing regimen.
  4. *Two patients on unboosted atazanavir and two patients on nelfinavir.
  5. Nelfinavir.
Age (mean ± SD)53 ± 754 ± 7
Male56 (80)56 (85)
Female14 (20)10 (15)
Current alcohol use29 (41)32 (48)
Current tobacco use18 (26)16 (24)
Current illicit drug use6 (9)5 (8)
Height (inches) (mean ± SD)68 ± 469 ± 3
Weight (lbs) (mean ± SD)183 ± 45174 ± 29
BMI (mean ± SD)28 ± 626 ± 3
CD4 count (cells/μL) (mean ± SD)596 ± 298592 ± 313
Undetectable HIV-1 RNA level59 (84)57 (86)
Liver disease22 (31)11 (17)
Baseline SBP (mean ± SD)123 ± 15126 ± 17
Concomitant CYP 3A4 inhibitors3 (4)2 (3)
Concomitant chronic opioids7 (10)6 (9)
Concomitant chronic benzodiazepines11 (16)12 (18)
ARV regimen  
NNRTI-based31 (44)25 (38)
Efavirenz22 (31)25 (38)
Nevirapine6 (9)0
Etravirine3 (4)0
NRTI only20
PI-based37 (53)40 (61)
Atazanavir/ritonavir21 (30)18 (27)
Darunavir/ritonavir4 (6)0
Lopinavir/ritonavir8 (11)16 (24)
Fosamprenavir/ritonavir1 (1)3 (5)
Tipranavir/ritonavir02 (3)
Saquinavir/ritonavir01 (2)
Unboosted PI4*1
Raltegravir-based01 (2)

Procedural sedation

Of the 136 patients in the cohort, 70 (51%) received midazolam-based sedation and 66 (49%) received diazepam-based sedation. Oxygen saturation was monitored in all patients and everyone received supplemental oxygen, as per institutional guidelines (the amount of supplemental oxygen administered was not collected). Colonoscopy procedures were performed by 34 different providers, 28 of whom used midazolam-based sedation in selected patients. Sedation was administered by a nurse under the supervision of an attending physician; most often a gastroenterologist. The mean total midazolam dose administered was 4.1 ± 1.2 mg (0.05 ± 0.02 mg/kg) with a mean initial bolus of 2 ± 0.2 mg. The mean diazepam dose was 7.7 ± 2.4 mg (0.10 ± 0.03 mg/kg) with a mean initial bolus of 3.4 ± 1.5 mg. Most patients received adjunctive opioids such as meperidine ± fentanyl. Meperidine was given to 69 (99%) of the patients who received midazolam-based sedation (mean total dose 79 mg meperidine), and to 57 (86%) of the patients who received diazepam-based sedation (mean total dose 89 mg meperidine).


There were no significant differences between the groups with respect to duration of sedation (length of time consciousness score < 2, mean 48.0 vs. 45.7 minutes for the midazolam and diazepam groups, respectively; P = 0.68), nadir SBP (mean 97.0 vs. 101.6 mmHg, respectively; P = 0.06), or nadir oxygen saturation (94.6 vs. 94.8%, respectively; P = 0.72) (Table 2). Eight patients in the midazolam group experienced a newly abnormal cardiac rhythm vs. 13 patients in the diazepam group (P = 0.18); no patient had a newly abnormal cardiac rhythm requiring treatment. Sinus bradycardia was the most frequently observed abnormal rhythm, occurring in six patients (9%) in the midazolam group and eight patients (12%) in the diazepam group. No patient became unresponsive with a consciousness score of 0 at any time or required a reversal agent, but more patients in the midazolam group experienced a depressed level of consciousness at some point during the procedure (91.4 vs. 74.2%; P = 0.0075). All patients remained responsive to verbal or tactile stimulation; no patient in either group was unarousable to painful stimulation.

Table 2. Outcomes related to procedural sedation for endoscopy with midazolam-based vs. diazepam-based sedation in HIV-positive patients on antiretroviral therapy
VariableMidazolam (n = 70)Diazepam (n = 66)P-value
  1. Data are number (%) of patients, unless otherwise indicated.
  2. BBS, bilateral breath sounds; HR, heart rate; mmHg, millimetres of mercury; RR, respiratory rate; SBP, systolic blood pressure; SD, standard deviation; TV, tidal volume.
Anaesthesia score   
Airway < 2 (clear airway)001.0
Respirations < 2 (clear/BBS, good TV)001.0
Skin perfusion   
Score = 2 (pink/dry)64 (91)65 (98)0.06
Score = 1 (pale/mottled/diaphoretic)6 (9)1 (2)
Score = 0 (cyanotic)00
Score = 2 (awake/oriented)6 (9)17 (26)0.0075
Score = 1 (arousable/disoriented)64 (91.4)49 (74.2)
Score = 0 (unresponsive)00
Blood pressure   
Score = 2 (SBP ± 20 mmHg)15 (21)17 (26)0.38
Score = 1 (SBP ± 20–40 mmHg)45 (64.3)35 (53)
Score = 0 (SBP ± 40 mmHg)10 (14.3)14 (21)
Cardiac rhythm   
Score = 2 (no change)62 (89)53 (80)0.18
Score = 1 (newly abnormal)8 (11.4)13 (19.7)
Score = 0 (requiring treatment)00
Lowest RR (mean breaths per min)15.815.70.63
RR range, change from baseline (breaths per min) (mean ± SD)2.8 ± 1.73.5 ± 2.20.04
Lowest HR (mean beats per min)62640.90
HR range, change from baseline (beats per min) (mean ± SD)17.6 ± 6.219.2 ± 9.60.25
Lowest oxygen saturation (mean %)94.694.80.72
Oxygen saturation range (%) (mean ± SD)4.4 ± 2.24.3 ± 1.70.85
Lowest SBP (mmHg) (mean ± SD)97 ± 13.9101.6 ± 14.10.06
Highest SBP (mmHg) (mean ± SD)134 ± 5142 ± 180.01
Length of time consciousness score < 2 (min) (mean ± SD)48 ± 28.645.7 ± 29.90.68
Pain scale (range 0–10; worst pain is 10)   
Score > 318 (26)10 (15)0.10
Score = 1–310 (14)18 (27)
Score = 042 (60)38 (58)
Procedure time (min) (mean ± SD)21 ± 1422 ± 150.76
Total time (min) including post-procedure monitoring (mean ± SD)98 ± 3197 ± 300.86

In terms of other outcomes, the diazepam group experienced a significantly greater change from baseline in respiratory rate, which decreased by 3.5 breaths per minute vs. a decrease of 2.8 breaths per minute in the midazolam group (P = 0.04). There were no significant differences in the overall blood pressure score—which compares intra-procedure SBP with pre-procedure SBP—in the midazolam group compared with the diazepam group (P = 0.38). The time to nadir SBP following sedative administration was also similar between the two groups (44 vs. 47 min for the midazolam and diazepam groups, respectively; P = 0.46).

We also compared outcomes in patients on PI-based ART vs. non-PI-based ART. There were no significant differences between these groups in terms of age, sex, liver disease, baseline SBP, or concomitant use of CYP 3A4 inhibitors, chronic opioids or chronic benzodiazepines. There were no significant differences between the four treatment groups (PI/midazolam; PI/diazepam; non-PI/midazolam; non-PI/diazepam) in the outcomes of lowest SBP, heart rate, nadir oxygen saturation, nadir respiratory rate, consciousness score < 2, sedation duration, or newly abnormal change in cardiac rhythm.

Multivariate analysis

We performed multivariate analysis for four discrete outcomes – attainment of consciousness score < 2, sedation duration, nadir SBP, and cardiac rhythm score – adjusting for possible confounders based on clinical relevance and significance in univariate analysis. In adjusted analyses, we found that midazolam use (vs. diazepam use) was significantly associated with a consciousness score < 2 (odds ratio = 3.80; 95% confidence interval 1.33; 10.88) but not with cardiac rhythm score, nadir SBP, or sedation duration (data not shown).


Most endoscopic procedures are performed with the patient under moderate (‘conscious’) sedation, during which the patient is able to respond purposefully to verbal or tactile stimulation [5]. Adequate sedation can both facilitate successful completion of the procedure and increase patient tolerability by decreasing anxiety and pain.

In this retrospective analysis of 136 adult HIV-positive patients on ART who underwent a colonoscopy, we assessed the safety of midazolam vs. diazepam based on multiple physiological outcomes. We found that the patients in our cohort safely received midazolam-based sedation, with comparable haemodynamic, cardiac, and respiratory outcomes compared with patients who received diazepam. We did observe a difference in consciousness score between the midazolam and diazepam groups, consistent with the known interaction between midazolam and ART. However, no patient suffered an outcome indicative of over-sedation, such as requirement for reversal of sedation, hospitalization or unresponsiveness. It is notable that more patients in the diazepam group compared with the midazolam group (26% vs. 9%, respectively) were awake and oriented during the procedure. Had patients in the diazepam group been given more diazepam to achieve a level of moderate sedation, it is likely that the duration of sedation would have been longer in the diazepam group.

Concerns regarding the safety of midazolam use in HIV-positive patients on ART arise from the possibility of drug interactions resulting in increased levels of midazolam [6-8]. Metabolism of midazolam is almost entirely dependent on hydroxylation by CYP 3A4 [2]. Guidelines indicate that iv midazolam can be co-administered with ART with close clinical monitoring [9-11]. In addition, it has been demonstrated in a number of small studies that, with careful titration, drugs such as fentanyl, meperidine and midazolam, can be administered safely for short-term procedures in HIV-positive patients on ART [12-15].

Previous studies investigating the clinical impact of co-administration of midazolam and ART have yielded conflicting results. In a cohort of 143 ambulatory patients undergoing bronchoscopy and receiving midazolam (total doses ranged from 2 to 15 mg; average 7.5 mg), no patient suffered an adverse outcome or required intubation [16]. By contrast, in a recent study of 241 HIV-positive in-patients who received iv midazolam prior to bronchoscopy [17], the incidence of prolonged sedation (defined as > 90 minutes) was 9.8% in patients receiving PIs compared with 1.58% in those not taking ART. One possible reason for the difference between these findings and our results is the fact that the previous cohort consisted of in-patients, in whom concomitant comorbidities may have predisposed patients to complications from sedation; indeed, 10% of the patients in the PI group had pre-existing respiratory distress or altered mental status prior to the procedure and 29% were diagnosed with Pneumocystis jirovecii pneumonia. By contrast, we examined only ambulatory patients and compared outcomes between patients on ART receiving iv midazolam and those receiving iv diazepam. It is important to acknowledge that the risks of iv midazolam may differ by procedure and the presence of medical comorbidities.

In our institution, based on pharmacokinetic principles and available evidence, we generally do not empirically reduce the dose of midazolam for HIV-positive patients undergoing procedural sedation. Peak benzodiazepine concentrations after a single dose or multiple doses over a very short period of time (e.g. 20 minutes) are not affected by drug half-life or clearance. Furthermore, in the context of short-term use, the recovery time for benzodiazepines is mainly dependent on redistribution kinetics rather than half-life.

Limitations of this study include its retrospective design and the limited sample size. To address the possibility of confounding, we conducted a multivariable analysis adjusting for variables such as age, sex, liver disease, and concomitant medications including CYP 3A4 inhibitors and chronic opioid use. We also examined outcomes in four different subgroups – midazolam- and diazepam-based sedation with and without concomitant PIs – and could not detect an association between the treatment groups and any of the outcomes. An additional limitation is that our institutional endoscopic anaesthesia scoring system may not be sensitive enough to detect the optimal level of sedation.

In conclusion, we demonstrated that HIV-positive out-patients undergoing colonoscopy who received iv midazolam for procedural sedation had similar clinical outcomes to those who received diazepam. Based on these findings, we conclude that the use of iv midazolam can be considered for HIV-positive patients on ART with close clinical monitoring. These findings should be confirmed in prospective studies or in a randomized controlled trial.


RTG is supported by grants from the National Institutes of Health (NIH) (R01 AI066992-04A1 and NIH G08LM008830-01) and by grants to the AIDS Clinical Trials Group (NIH U01 AI 694722) and the Harvard University Center for AIDS Research (NIH 2P30 AI060354-06).

Conflicts of interest: RTG has received grant support from Tibotec and Gilead.