Chromaffin cells from the adrenal gland secrete a mixture of compounds that have a strong analgesic effect, especially when administered intrathecally. Many studies in animal models have shown that discordant xenogeneic cell isolates, including chromaffin cells, can survive and have biologic effects when transplanted within a semipermeable membrane capsule.
To evaluate the clinical potential of encapsulated cell therapy, a human-scale implant containing bovine chromaffin cells was developed, characterized, and implanted in the subarachnoid space of seven patients with severe chronic pain not satisfactorily managed with conventional therapies. Patients received no pharmacologic immunosuppression. Cell devices were implanted during minimally invasive surgery, and device design allowed retrieval. All devices were recovered after implant periods of 41 to 176 days.
Postexplant histologic analysis, immunostaining, and secretory function all confirmed survival and biochemical function of the encapsulated cells. Reductions in morphine intake and improvement in pain ratings were observed in several patients.
This study represents the first successful trial of encapsulated xenogeneic cells in humans. The preliminary findings of pain reduction warrant the initiation of a randomized, double-blind phase II study to evaluate the potential efficacy of the procedure.
Key words: Catecholamines. Chromaffin cells. Chronic pain. Encapsulation. Enkephalins.
Several emerging therapeutic strategies rely on transplantation of naturally occurring or genetically modified cells as a continuous source of biologically active cell secretions. [1-3]In one such approach, the implanted cells are surrounded by a biocompatible semipermeable membrane that allows passage of oxygen, nutrients, electrolytes, and cell products but restricts transport of the larger agents of the body's immune defense system. Called "immunoisolation," this technique was first proposed in the mid-1970s as an artificial endocrine pancreas for the treatment of diabetes [5,6]and subsequently has been investigated preclinically for several other diseases or disease complications, including Parkinson disease, [7,8]Alzheimer disease, amyotrophic lateral sclerosis, hemophilia, and intractable chronic pain. The conceptual appeal of immunoisolation rests on the assumptions that the imposition of a barrier between graft and host will eliminate the need for host immunosuppression, with its associated side effects and expense, and also allows the harvesting of cells from nonhuman species, thus providing an unconstrained quantity of graft material. In the case of diabetes, these assumptions have been amply confirmed in rodents, but successful results in canines [13,14]or preliminary human trials [15,16]have been limited to allografts, sometimes with immunosupressed hosts. Central nervous system applications typically require far fewer cells than does diabetes and involve placement of the implant in a less immunoresponsive host site. [17,18]Successful transplantation of encapsulated xenogeneic cells to the central nervous system, without immunosuppression, has been confirmed in primates and sheep ; the study we report here represents the first human trial of encapsulated xenogeneic cells.
Chronic pain is a serious medical problem. More than 75% of patients at the end stage of cancer experience serious pain. Ten to thirty percent of these patients cannot achieve adequate relief with morphine and similar opioid agonists because of opioid insensitivity or development of tolerance, or because side effects such as constipation, cognitive impairment, pruritus, and respiratory depression have become unmanageable or unacceptable. Therapeutic options for patients with inadequate relief with opioids include implantable pumps for regional or continuous delivery of analgesics, neuroablative surgery, and electrostimulation. All of these options may be helpful in certain circumstances but none is fully satisfactory. Adrenal chromaffin cells release several analgesic compounds, including catecholamines, enkephalins, endorphins, neurotensin, and somatostatin, which have strong analgesic effects, especially when administered epidurally or intrathecally. The analgesic pathways of these compounds rely on a strong synergy due to activation of alpha2adrenergic receptors by norepinephrine and epinephrine and of delta opioid receptors by met-enkephalin. [22,23]Sagen [24,25]and others have exploited the analgesic characters of adrenal chromaffin cells products to obtain significant antinociceptive effect in rodents after intrathecal implantation of adrenal allografts. Other investigators have shown similar reductions in pain sensitivity in rodents after implanting cell lines producing one or more of the constituents normally released by chromaffin cells. [27-29]In 1993, Winnie and colleagues reported results of a human trial in which cadaver-sourced adrenal allograft isolates were injected into the intrathecal space of five patients with terminal cancer who were transiently immunosuppressed. The investigators reported pain relief in four of five patients, although the pretransplant level of cell output was not measured and cell survival per se was not documented. 
Intractable chronic pain in patients with terminal cancer appeared to us an appropriate indication for the initial human trial of immunoisolated xenogeneic cell therapy. The risk-to-benefit ratio was judged favorable; a relatively small graft of only a few million cells was required; the intrathecal site is somewhat immunopriviliged (i.e., more tolerant to foreign tissue than the subcutaneous site [17,18]); and issues related to soft-tissue biocompatibility could be minimized by placing the device in the cerebrospinal fluid space. Studies in rodents showed that encapsulated bovine adrenal chromaffin cells implanted by laminectomy produced antinociceptive effects in acute and chronic rodent models of pain and that encapsulated xenogeneic cells maintained analgesic effectiveness for at least 3 months after implantation. A human-scale device was designed, along with special surgical tools for implantation into the human subarachnoid space; techniques were also developed to retrieve the implant. Devices and methods were carefully validated in preclinical studies in sheep, a model chosen because of the similarity of sheep and human spinal anatomy. Results in this model confirmed device biocompatibility, cell viability, and general suitability of the techniques for human implantation. 
An initial phase I protocol was formulated, with safety and cell viability as the primary outcome parameters. Doses were set at the lowest level anticipated to provide antinociception, based on linear scaling from animal results. After considerable review of the difficult question of a control arm in a trial involving surgery, it was decided to include placebos only in later studies, and only after viability, function, and preliminary efficacy had been demonstrated in unblinded studies.
Materials and Methods
Two-week-old calves, obtained from healthy livestock sources in vaccinated herds with no known exposure to bovine spongiform encephalitis, were transported to an animal surgical suite. Adrenal glands were removed using aseptic technique, placed in a double-wrapped sterile container, and brought to a class 100 cell-isolation facility. Chromaffin cells were separated using a previously described enzymatic digestion technique. Yield per gland was 20 to 30 million cells. Recovered cells were stored for 2 days to permit preliminary bacteriology testing. Using aseptic technique, cells were subsequently suspended in a sterile alginate solution (Protan, Norway) and introduced into the lumen of sterile preassemblies (CytoTherapeutics, Providence, RI), which were then closed adhesively to form the implant. As previously described, the completed implant was a flexible tube about 1 mm in diameter, comprising a 5-cm length of sealed hollow-fiber semipermeable membrane, cells suspended in an alginate matrix within the lumen, a titanium connector, and a 20-cm silicone tether. The permeable portion of the implant wall was fabricated from an acrylic copolymer membrane with a nominal molecular weight cutoff of 50,000 Daltons and a smooth external skin. This membrane enclosed approximately 2 million chromaffin cells in a volume of 20 to 30 micro Liter. After assembly, implants were stored in serum-free defined media (PC-1; Hycor, Biomedical Inc., Irvine, CA) in a sterile incubator. Implants were tested individually for catecholamine release 4 days after filling; units with a nicotine-evoked norepinephrine release of less than 2,000 pmol/30 min were not processed further. For devices used in this study, preimplant values +/- standard error of basal epinephrine and norepinephrine release were 134 +/- 30 and 248 +/- 136 pmol/device/30 min; this is equivalent to a combined in vitro catecholamine release of approximately 2 to 3 micro gram [centered dot] day sup -1 [centered dot] device sup -1. The corresponding values for 30 min of nicotine-stimulated release were 2,895 +/- 756 for epinephrine and 3,654 +/- 504 for norepinephrine. Culture media from the implants and from cohorts were tested for sterility before releasing units for implantation. All devices were implanted between 4 and 6 days after encapsulation, except for one (patient 6), for whom the device was held for 12 days before implantation.
Seven patients with severe chronic pain inadequately managed with conventional therapies were enrolled in the study. All protocols were reviewed and approved by the Ethical Committee of the Faculty of Medicine at the University of Lausanne, Switzerland. Patients gave written informed consent after being advised of the risks of the study. Table 1lists patient characteristics and diagnoses. Six had terminal cancer and short life expectancies; a seventh (patient 4) had unrelieved neurogenic pain secondary to thoracotomy and scoliosis with a deformation of the thoracic cage. Four patients (1, 2, 5, and 6) were receiving epidural morphine at the time of implant but were clinically judged to have unfavorable balance between pain relief and side effects. Two patients (3 and 7) were not treated with morphine because its administration induced incapacitating side effects before achieving any pain relief. The seventh (patient 4) was medically unsuited to receive morphine because of his respiratory status. No patients had known spinal cord compression or metastases, severe psychosis, or a major clotting disorder.
After written informed consent was obtained, a medical history, including questions about concomitant medications, was taken. Patients underwent baseline testing to quantify their pain before implantation. Two scales were used: the short McGill Pain Questionnaire and a 10-cm visual analog scale (none = 0; most severe = 10). The devices were surgically implanted in a 15-min, minimally invasive procedure. Under local anesthesia, a 3-cm craniocaudal incision was performed at the L4-L5 level. The subcutaneous tissue was dissected down to the dorsal fascia. The subarachnoid space was punctured with a 25-G Tuohy needle and approximately 10 ml cerebrospinal fluid (CSF) was withdrawn. A guide wire was introduced through the Tuohy needle, which was then retrieved. The ligatum flavum was widened by introducing a dilatator over the guide wire and the dilatator was retrieved. A 4-French cannula was placed into the subarachnoid space. The guide wire was retrieved and the active portion of the device was placed in the CSF among the spinal roots of the cauda equina by sliding it through the cannula. The cannula was then retrieved. The external end of the silicone tether extending out of the implantation track was sutured to the lumbodorsal fascia and completely covered with skin closure. The position of the implant, relative to the spine, is shown in Figure 1, a magnetic resonance image in patient 6 taken 3 weeks after implantation. Patients were not given systemic or local immunosuppression at any time during this trial. In the seven patients in this series, devices were implanted during a 6-month period. After implantation, patients remained in bed overnight (patients 1 to 4) or for 48 h (patients 5 to 7). During the period of follow-up, patients were closely monitored, medication use was recorded, and McGill and visual analogue scale ratings were taken frequently. Periodic lumbar punctures were performed on patients 1 to 4 to measure CSF concentrations of epinephrine and norepinephrine. Devices were retrieved by subcutaneous dissection and withdrawal from patients 1, 2, and 4 after implantation periods of 55, 41, and 85 days, respectively; devices were removed 1 to 9 h after death from the remaining patients after periods of 43 to 172 days (Table 1). All devices were recovered intact, except for one that broke during explantation from patient 2. This breakage occurred within the paravertebral muscle and is believed to be the consequence of removing the device while the patient was in the sitting rather than the usual prone position; the capsule fragment was completely removed without another surgical procedure. Postexplant basal and stimulated catecholamine release was only measured for devices retrieved from patients 1, 4, and 6. All retrieved capsules were examined visually and microscopically at explant. Devices were then divided in half and one portion was embedded in acrylate for histologic analysis; the remaining portion was cryosectioned before immunostaining with methionine enkephalin and tyrosine hydroxylase. Explants from patients 1, 4, and 6 were tested for catecholamine release before histologic and immunochemical analyses.
Patients 1 and 2 died of complications from their underlying malignancies 6 and 3 weeks after device removal. Autopsies, including histologic sectioning of the spinal cord with special attention to the implant site, were performed on patients 1 to 3 and 5 to 7. The presence of viable cells at explant and preliminary suggestions of a antinociceptive effect for patients 1 to 3 were described previously in a preliminary report. 
Overall study results are summarized in Table 1. The implantation procedure was generally uncomplicated and recovery was uneventful. Four patients reported headaches after surgery that were attributed to CSF leakage at the implant site. These resolved either spontaneously or after an epidural blood patch; none continued more than 3 days after surgery. The extent of postoperative headaches appeared to decrease with longer periods of postoperative bed rest and a 20% reduction of the diameter of the insertion cannula.
Four of the patients were receiving epidural morphine at the time of implant and in all four their opioid use decreased during the study (Table 1) with either a modest improvement or no worsening in pain ratings. Morphine administration is plotted in Figure 2for patients 2, 5, and 6. Patient 1 was not included in this plot because epidural bupivacaine and fentanyl therapies were initiated and maintained during the study period and it is unclear whether improvements resulted from these agents or from the implant or both. Consistent with a pattern of increasing level of pain due to progression of his underlying disease, patient 2 began receiving intravenous morphine (155 mg/day) 36 days after device implantation and 5 days before device retrieval. He survived 26 days after device removal; during this period his epidural morphine intake increased from 35 mg per day at explant to 130 mg per day at the time of death.
Three patients were not receiving oral or epidural morphine treatment at the time of implant. Patient 3 was unresponsive to oral morphine but did receive intravenous morphine (36 mg day) from day 21 to day 42 after implantation as treatment for respiratory complications. Patient 4 was medically unsuitable for systemic opioids because of severe chronic respiratory failure and other respiratory problems. Patient 7 was diagnosed as a nonresponder. All three patients showed improvement in McGill Pain ratings (Figure 3); patients 3 and 7 showed a corresponding or greater improvement in their visual analogue scale pain ratings. The change in pain rating for patient 3 preceded the initiation of intravenous morphine therapy but must nevertheless be interpreted with some caution because she also received short-term courses of several different concomitant analgesic therapies, including epidural and subarachnoid bupivacaine and epidural fentanyl. Patient 4 was the only participant in this study who experienced prolonged survival after device retrieval. His pain ratings did not appreciably worsen after device removal, but he did request and received a second implant about 3 months after removal of the first implant.
There was no visible difference between the appearance of the devices as implanted and as retrieved. Microscopic examination after recovery showed that external surfaces of the implants were free of adherent cells, fibrous overcoats, and other signs of acute-phase response or foreign-body reaction. Intracapsular populations of chromaffin cells were identified by histologic analysis in six of seven devices after retrieval. Viable cells in these six devices were also positive for tyrosine hydroxylase and met-enkephalin by immunostaining, thus confirming cell function and implying secretory capacity. Representative histologic analysis and immunostaining is presented in Figure 4. Viability was assessed subjectively as high in all but one device but was not quantitated. The one device not positive for viable cells by histologic analysis was recovered from patient 5, in whom a massive bilateral subdural hematoma with subarachnoid hemorrhage secondary to meningeal metastases confirmed by computed tomographic scan developed before death. Histologic examination showed cells within the capsule that appeared consistent with recent death.
Catecholamine concentrations in lumbar puncture aspirates taken from patients 1 through 4 were inconclusive. Relative to the time of implant, norepinephrine increased in these patients from an initial mean +/- SEM of 1.4 +/- 0.6 pmol/ml to mean values in life of 5.3 +/- 1.1 pmol/ml; corresponding initial and in-life values for epinephrine were 11.0 +/- 1.1 pmol/ml and 9.4 +/- 0.9 pmol/ml, respectively. Catecholamine release from capsules was measured in patients 1 and 4 after retrieval and after death in patient 6. In patients 1 and 4, capsule basal release declined about 50% compared with values at the time of implant; nicotine-stimulated release had declined to nearly zero in patient 1 but increased three times in patient 4. The data for patient 4 are shown in Figure 5. In patient 6, basal release decreased to zero and evoked release decreased to about one third of input values; however, the capsule remained in the patient for 9 h after death before explanation.
Autopsies and histologic sectioning of the spinal cord from patients 1 to 3 and 5 to 7 were unremarkable. No signs of demyelination or meningitis were observed. In general, no adverse findings were reported that could be attributed to the implantation technique or to the presence of the device in the subarachnoid space. In patient 2, whose capsule broke during retrieval, no evidence of living cells, superinfection, granuloma formation, or other untoward sequellae was observed.
The present study supports the observation that discordant xenogeneic cells can survive and function in humans, in the absence of immunosuppression, when separated from the host by an appropriate semipermeable membrane. It also demonstrates the safety of transplanting encapsulated bovine chromaffin cells in the human CSF.
The survival of immunoisolated bovine cells in human hosts was confirmed by histologic, immunocytochemical, and postexplant secretory function analyses. This finding is encouraging for the ongoing development of other therapies involving transplantation of immunoisolated cells to the central nervous system. [7-10]The survival of cells in patient 4 (Figure 4) is particularly significant because he lacked any deterioration of the immune system that might be suspected in patients with terminal cancer. Care must be taken in extrapolating these results to implants in soft tissue, to less immunoprivileged sites, and to devices containing more cells or different xenodonor-host combinations. Nevertheless, the fundamental principle has been shown that nonhuman cells can be transplanted without immunosuppression to humans, where they will survive for at least several months and function inside an immunoisolatory barrier and without apparent danger to the host. The mechanisms of recognition and response within the central nervous system are still poorly understood and it is difficult to predict what impact, if any, this finding will have on neuroimmunology or the broader field of xenotransplantation.
Subject to caveats about placebo effects (as explained in the next paragraph), the reduction in opioid administration and improvements in pain scores after device implantation are promising and suggest that cell therapy may have an important future role in alleviating chronic pain. The nature and extent of this role, however, await further studies that compare the pharmacoeconomics and risk-benefit ratio of this therapy with alternative approaches. Future studies, most likely in animals, will also need to address mechanisms of action, particularly questions of which combinations of chromaffin cell secretions are responsible for the observed effects.
The existing study does not rule out possible placebo effects, either a partial contribution or in toto. Nevertheless, in cases requiring surgical intervention (even minimally invasive surgery), we believe it entirely appropriate to postpone a blinded study until implant function and safety has been confirmed and a strong suggestion of effectiveness is in hand. Subsequent studies involving dose-escalation, noninvasive sham surgeries, and blinded observers are underway. Perhaps fittingly for a hybrid device, this approach represents a midpoint between clinical evaluation of pharmaceuticals, where placebo controls are often included early on, and testing of medical devices, where placebos are rarely used.
Device retrievability proved to be both an important study tool and a considerable safety advantage; whenever possible, this design feature should be included in future immunoisolatory devices. The incidence of CSF leaks immediately after implantation may need to be reduced further, either by improved surgical technique and postoperative management, or by additional changes in device design. The rupture of the device in patient 2 appears due to the particular explant procedure for that patient; no similar problem was observed in any other patients. Nevertheless, device integrity should be monitored carefully in the future. Device sterility is critical and should eventually be assured in USP 14-day protocols rather than in the shorter bacteriological tests relied on in this study. Ideally, concomitant patient medication would not vary during the course of the study, as it did here for patients 1, 3, and 6; as a practical matter, such controls are difficult to impose on the real world of terminal cancer pain triage, and such excursions will need to be carefully recorded and judiciously interpreted. Many additional issues relate to dosage and how to measure it. The variability of preimplant catecholamine release could be further reduced by industrial engineering or implant preselection. Release after explant needs to be measured routinely in subsequent clinical trials, even though this may require explanting and replacing functioning devices. Intrathecal release may vary widely from in vitro results before and after implantation; this question and issues of concentration of various agents in the subarachnoid space may best be resolved in animal models rather than in humans. In the main, however, the performance of these xenogeneic cells in patients without immunosuppression represents a highly encouraging initiation of encapsulated cell therapy in humans.