Patient-to-patient transmission of hepatitis C virus (HCV) during colonoscopy diagnosis

On February 2006, two regular blood donors were independently diagnosed with acute HCV infection. Both individuals had undergone colonoscopy procedures at the same endoscopy unit on December 19^th, 2005. These two patients (C2 and C3) had tested negative for HCV prior to colonoscopy. Thus, the time of putative transmission was considered to be comprised between the last HCV negative test of these patients and the dates of their onset of symptoms (Figure 1 and Table 1). The initial wide case-finding investigation focused on recorded HCV status of patients attending the colonoscopy unit from July 2005 to February 2006 and did not reveal any suspect of nosocomial infection. Next, a more limited and detailed case-finding search was carried out among all attendants to the same endoscopy unit between December 12^th, 2005 and December 26^th, 2005. Apart from the two index cases, 39 patients matched temporal criteria, but three were excluded due to exitus. The active search for HCV-infected cases among these 36 eventually exposed patients (30 of which had gone through colonoscopy and the remaining 6 through gastroscopy) revealed two asymptomatic HCV-positive individuals (C1 and C4). The epidemiological questionnaire indicated that patient C1, but not patient C4, had also undergone a colonoscopy on the same day as patients C2 and C3.

HCV subtypes for patients C1, C2 and C3 were 1b, while that of C4 was subtype 2c. This patient was considered to be unrelated to the outbreak and was removed from further analyses. Relevant features and nosocomial risk factors of patients C1, C2 and C3 are shown in Table 1. On December, 19^th, 6 other patients had undergone endoscopy diagnosis (3 gastroscopies and 3 colonoscopies) (Figure1). Data extracted from epidemiological questionnaires (Table 1) strongly indicated that patient C1 was the likely source of infection of the two incident cases. One additional patient underwent gastroscopy between colonoscopies practiced on patients C1 and C2 (Figure 1) but tested negative for HCV.

The epidemiological investigation focused on determining the transmission route. The procedures that could have led to patient-to-patient infection were examined: (a) blood collection and peripheral intravenous line placement and (b) sedation in the endoscopy room, later followed by colonoscopy. In the former, usual practices and decision-making about materials largely depend on the nurse in charge. Patient-nurse assignment was not recorded although it could be traced after interviews, and written protocols were not available. In spite of this, interviews to nurses did not reveal any failure in disinfection standards applied to blood-related practices.

The same anaesthesiologist, endoscopist, two nurses and auxiliary nurse performed anesthesia, colonoscopy procedure and nurse assistance, respectively, on all patients that underwent endoscopies on December 19^th, including patients C1, C2 and C3. Written records documented that these three patients received parentally midazolam, propofol and anfentanil for anesthetic purposes. The patient that underwent gastroscopy between colonoscopies practiced on patients C1 and C2 received propofol and lidocaine. It could not be checked whether medication vials were multi- or mono-dose and whether they were used on more than a patient. Written protocols for sedation were not available. Interviews of medical personnel related to sedation procedures did not reveal any failure in preventive measures of infection.

The endoscopy unit had two colonoscopes, for which traceability failures were detected in patient assignment and disinfection records. A registered processing label recovered from the supply unit of endoscopes along with data from questionnaires established that patient order in sedation and colonoscopy was C1, C2 and C3 (Table 1). Biopsy specimens were obtained only from patient C3 by means of sterilized forceps provided by the central sterilization unit. Although it was not possible to verify whether the same colonoscope had been used with patients C1, C2 and C3, examination of the endoscope disinfection routines could not rule out this possibility. General standard guidelines for decontamination included manual cleaning of equipment followed by automatic disinfection. According to the available written protocol, obstruction of colonoscope channels with gastrointestinal debris could not be ruled out completely. In this procedure, colonoscope channels were flushed and rinsed with water and externally washed. After removing the valves, colonoscopes were immersed in enzymatic detergent and brushed manually. Individual clearance of all colonoscopy channels was not checked. Next, channels were rinsed with cleansing solution and introduced into the automatic-washer disinfector. As for the automatic step, correct disinfection of colonoscopes could not be checked due to a failure of the printer connected to the disinfection device. In addition, technical staff members in charge of colonoscope disinfection had not received specific training.

Partial NS5B sequences obtained from patients C1 and C2 were identical and only differed in three of 337 positions from the sequence corresponding to patient C3. As a result, the phylogenetic tree of the NS5B region (Figure 2) showed that patients C1, C2 and C3 clearly clustered in a monophyletic group among sequences belonging to HCV subtype 1b (bootstrap support 77%). The analysis of the NS5B region clearly defined the extent of the outbreak. As expected in a well-defined outbreak, the genetic distances between sequences belonging to patients C1, C2 and C3 were substantially shorter than those between them and local reference sequences from epidemiologically unrelated patients (Figure 2). In addition, another well-supported cluster also appeared among the reference sequences probably indicating local epidemiological links between three isolates in the past. Monophyletic clustering of sequences in the phylogenetic tree of the conserved NS5B region enabled us to define the patients involved in the outbreak. In order to further characterize the respective HCV viral populations, the highly variable E1-E2 region of HCV was analysed.

Thus, sequences from 74 cloned fragments of the E1-E2 region, encompassing the hypervariable region (HVR), were obtained from patients C1, C2 and C3 (26, 28 and 20 sequences respectively). In order to avoid redundancy, only unique sequences within each patient were used in phylogenetic analysis. Therefore, a total of 38 sequences (21, 15 and 2 unique sequences from patients C1, C2 and C3, respectively) were included in the phylogenetic analysis.

These E1-E2 cloned sequences were analysed along with 35 local reference sequences from epidemiologically unrelated patients. The resulting phylogenetic tree, in which only different sequences within patient were included, is shown in Figure 3. Sequences of clones from patients involved in the outbreak gave rise to a well-supported monophyletic group (bootstrap support 80%). Most of the 21 different cloned sequences obtained from patient C1 were unique and showed relatively short distances from each other. However, two divergent sequences were also present among the isolated cloned sequences. Patient C2 provided 15 different cloned sequences, some of which were repeatedly sampled between 2 and 8 times. All these sequences mixed with those from the main cluster of patient C1 sequences in the phylogenetic tree (Figure 3). Moreover, two sequences from patient C1 were also detected in patient C2 (one was sampled three times and the other eight times). Finally, patient C3 presented two unique cloned sequences which grouped in a differentiated cluster (bootstrap support 100%) also located among sequences of the main cluster of patient C1 sequences, with one sequence sampled 19 times and the other only once (Figure 3). In congruence with the phylogenetic tree of the NS5B region, genetic distances between E1-E2 sequences from patients C1, C2 and C3 were much shorter than those between these and the reference sequences.

The outbreak pattern in phylogenetic trees, i. e. monophyletic cluster of outbreak-related sequences, is obvious in both NS5B and E1-E2 regions. However, due to the larger number of mutations accumulated in the more variable E1-E2 region, genetic distances in this portion of the genome can provide additional information on how viral populations have emerged in patients involved in the outbreak. Two additional well-supported clusters appeared among the reference sequences for the E1-E2 region (Figure 3), probably indicating local epidemiological links involving two and three isolates, respectively.

Intrapatient genetic variability for the E1-E2 region estimates (Table 2) showed the highest variability for patient C1 viral population in all parameters (haplotype diversity, number of polymorphic sites, total number of mutations, nucleotide diversity and average number of nucleotide differences between pairs of sequences) followed by lower values for patient C2 and the lowest genetic variability corresponding to patient C3. For instance, nucleotide diversity (π) of patient C1 viral population was one and two orders of magnitude larger than those of patients C2 and C3, respectively. Therefore, these estimates showed a pronounced decrease in genetic variability from patient C1 to patient C3 (C1>C2 > C3). In a reconstructed scenario, first, a viral inoculum from patient C1 was in contact with patient C2 and later with patient C3. An unidentified factor could have reduced the remaining viral population after infecting patient C2, but left enough viral load to cause infection in patient C3.

The newly HCV positivity of the two index patients after undergoing colonoscopy in a private endoscopy unit on 19^th December, 2005 lead to a retrospective survey to study a possible nosocomial transmission. First, a wide case-finding investigation included reviewing HCV-positive and negative records from all local Microbiology Units between January 2004 and February 2006 (eight hospitals, including the one under study) of patients that had undergone endoscopy procedures between July 2005 and February 2006 at the endoscopy unit. A second, more limited and detailed search involved all patients who had attended the endoscopy unit one week before and after the date considered. This investigation covered review of clinical records for 39 patients matching temporal criteria. All these patients, except three due to exitus, were actively contacted, informed about the possibility of exposure to HCV and tested for anti-HCV. The 36 patients completed a detailed epidemiological questionnaire on risk factors for HCV infection including history of transfusion, intravenous drug use, piercing, tattooing, acupuncture and suspicion of nosocomial transmission. In order to rule out delayed seroconversion, the six patients that had undergone colonoscopy or gastroscopy on 19^th December, 2005 and tested negative for HCV in February 2006, were retested 6 months later. All of them were found to be anti-HCV negative at that time. Healthcare staff involved in performing endoscopy (one gastroenterologist, one anesthesiologist, seven nurses and one nurse's aide) had negative HCV status in February 2006.

An epidemiological investigation to determine risk procedures leading to HCV transmission during endoscopies was also carried out. This included a description of the facilities, review of clinical practices and procedures performed between admission and discharge, description of staff member qualifications and assessment of disinfection procedures and devices (i.e. observation of endoscopic procedures and endoscopic reprocessing, and observation of procedures regarding intravenous line insertion for anaesthetic purposes). All healthcare workers were interviewed about infection-control practices.

Serum samples were obtained from four HCV-infected patients (C1, C2, C3 and C4) referred to the endoscopy unit. Samples were taken on February 2006 and stored frozen at -70°C until processed. Viral load of specimens was not measured.

Viral RNA was obtained from 200 μl of serum from each sample using a High Pure Viral RNA Kit (Roche Diagnostics GmbH, Mannheim, Germany). Reverse transcriptions (RT) and PCRs were performed as described in Bracho et al. (2005) [23]. Amplified fragments were 337 and 532-nucleotide-long for NS5B and E1-E2 regions, respectively. Oligonucleotides used for amplification and direct sequencing of the NS5B region were 5'-TATGATACYCGCTGYTTYGACTC-3' (sense), 5'-GTACCTRGTCATAGCCTCCGTGAA-3' (antisense), and for the amplification of the E1-E2 region, 5'-CGCATGGCYTGGGAYATGAT-3' (sense), 5'-GGGGTGAARCARTAYACYGG-3' (antisense), and hemi-nested 5'-GGGATATGATRATGAAYTGGTC-3' (sense). A single amplified product for each region was observed after electrophoresis on a 1.4% agarose gel stained with ethidium bromide. Amplified products of the NS5B and E1-E2 regions were purified with High Pure PCR Product Purification Kit (Roche Diagnostics GmbH, Mannheim, Germany). The NS5B gene sequences were obtained directly with the ABI PRISM BigDye^® Terminator v3.1 Cycle Sequencing Kit in an ABI 3730 (Applied Biosystems Foster City, CA) automated sequencer and amplification oligonucleotides as described in Bracho et al. (2005).

Amplified fragments of the E1-E2 region, encompassing HVR1, HVR2 and HVR3 [22], were directly cloned in EcoRV-digested pBluescript II SK (+) phagemid (Stratagene, La Jolla, CA). Plasmid DNA was purified with a High Pure Plasmid Isolation kit (Roche). Recombinant plasmids were sequenced with KS and SK oligonucleotides (Stratagene) as described above.

Partial NS5B sequences were genotyped using HCV BLAST search at Los Alamos HCV Sequence Database [31]. NS5B sequences from patients C1, C2, and C3 were analyzed along with a panel of 28 additional local sequences from epidemiologically unrelated isolates [23], including NS5B reference sequences for HCV subtype 1b. Accession numbers for the NS5B reference sequences (EMBL) are shown in Figure 1. Cloned sequences of the E1-E2 region, corresponding to the three patients infected with HCV subtype 1b, were analyzed along with 35 reference sequences of HCV subtype 1b. These E1-E2 reference sequences were also obtained from patients included in the above study [23]. The accession numbers for E1-E2 region reference sequences are shown in Figure 2. Prior to phylogenetic analysis of the E1-E2 region, unique cloned sequences within patient were selected using DAMBE software [32]. For both HCV genome regions, multiple sequence alignments were obtained using ClustalX 2.0.10 [33]. Phylogenetic trees were constructed by maximum likelihood with PhyML v3.0 [34] using the nucleotide substitution model that best fitted the data according to corrected Akaike Information Criterion (AICc) implemented in jModeltest 1.0.1 [35]. The evolutionary model that best explained the data in the NS5B region corresponded to TIM2 model distance [35], with a gamma distribution accounting for heterogeneity in evolutionary rates among sites (shape parameter = 1.034) and a proportion of invariable sites (I = 0.630). The evolutionary model that best explained the data for the E1-E2 region corresponded to Tamura-Nei [36], with a gamma distribution accounting for the heterogeneity in evolutionary rates among sites (shape parameter = 0.737) and a proportion of invariable sites (I = 0.273). In addition, a parallel analysis was carried out with E1-E2 cloned sequences from local reference patients instead of the corresponding consensus sequences, but the maximum likelihood tree yielded similar results (data not shown). Robustness of the clades derived in the phylogenetic trees was assessed by bootstrap analysis using 1000 pseudo-replicates as implemented in PhyML. Genetic variability estimates for the complete set of E1-E2 cloned sequences derived from patients C1, C2 and C3 were obtained with DnaSP v5 [37].

The HCV sequences obtained in this study have been deposited in GenBank under accession numbers [EMBL:FM881903 to EMBL:FM881906] for NS5B region sequences and [EMBL:FM955149 to EMBL:FM955222] for clones of the E1-E2 region.