Reduction of Indocyanine Green–Associated Photosensitizing Toxicityin Retinal Pigment Epithelium by Sodium Elimination

Objective  To determine if eliminating sodium affects indocyanine green (ICG) photosensitizingtoxicity and uptake in cultured human retinal pigment epithelial (RPE) cells.

Methods  Cultured human RPE cells were exposed to ICG (2.5 mg/mL) in balancedsalt solution and sodium-free balanced salt solution for 2 minutes. Afterwards,ICG was removed, and the cells were irradiated with a light beam (4 ×10⁴ lux) for 40 minutes. Toxicity was monitored using light microscopy,calcein AM–ethidium homodimer 1 staining, trypan blue exclusion test,and 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazoliumviability assay. Indocyanine green uptake was measured by optical absorptionat 790 nm.

Results  Photoreactive changes occurred in RPE cells exposed to ICG and light.These changes included cell shrinkage, cell death, pyknotic nuclei, reducedviability, and reduced mitochondrial dehydrogenase activity. These changeswere less severe when ICG was dissolved in sodium-free balanced salt solution.In addition, ICG uptake was reduced when the solvent was sodium-free balancedsalt solution.

Conclusion  Indocyanine green and intense light exposure in RPE cells caused photosensitizingtoxicity that was reduced when sodium in the solvent was eliminated and replacedwith other cations.

Clinical Relevance  Eliminating sodium from the solvent reduced ICG uptake into RPE andits associated photosensitizing toxicity. This reconstitution method of ICGmay be helpful for safer intravitreal ICG use in macular hole surgery.

Peeling of the internal limiting membrane (ILM) to eliminate tangentialtraction force improves the anatomical closure rate and functional outcomesafter macular hole surgery,^1-4 althoughsome investigators hold a different opinion.^5-7 However,if ILM removal is attempted during macular hole surgery, its visualizationmay be difficult. Staining of the ILM with indocyanine green (ICG) to enhanceits visibility^8-13 hasbeen proposed. The procedure involves direct application of ICG to the innersurface of the retina in the macular area. No adverse effects of ICG werereported in these studies. The concentration of ICG used in these reportsranged from 0.6 to 5.0 mg/mL, and ICG was left in the vitreous cavity from30 seconds to 5 minutes.

Other investigators, however, have described ICG-associated adverseeffects. For example, Gandorfer et al¹⁴ reportedless favorable visual outcomes when 5 mg/mL of ICG was used and drained immediatelyafter injection. Similarly, Engelbrecht et al¹⁵ observeda high incidence of unusual atrophic retinal pigment epithelial (RPE) changesat the site of the previous macular hole and its surrounding subretinal fluidafter using 1 mg/mL of ICG for 0.5 to 2.5 minutes. The median preoperativebest-corrected visual acuity was 20/200, while the median postoperative best-correctedvisual acuity was 20/400. The macular hole was closed in 86% of eyes. Haritoglouand coworkers¹⁶ reported no statistically significantimprovement in postoperative visual acuity after ICG-assisted ILM peeling.In an in vitro study, Sippy et al¹⁷ reporteda significant reduction in mitochondrial dehydrogenase activity in culturedRPE cells exposed to ICG alone or ICG plus light. In a recent study,¹⁸ ICG plus intense light exposure caused cell cyclearrest and apoptosis in cultured human RPE. Because intraoperative endoilluminationwith a high level of light intensity is used during macular hole surgery withICG-assisted ILM peeling, photosensitizing toxicity to RPE is possible.

Indocyanine green is an organic anion with amphophilic (ie, hydrophilicand hydrophobic) properties.¹⁹ It is most oftenadministered intravenously. After intravenous injection, it is rapidly clearedfrom the circulation by the liver and eliminated in bile. The uptake of organicanions by hepatocytes involves sodium-dependent and independent transportsystems.²⁰ Indocyanine green is also takenup by keratinocytes²¹ and cultured aortic endothelialcells.²² In an attempt to reduce ICG uptakeinto RPE cells and its associated photosensitizing toxicity, we replaced sodiumin the solvent with other cations.

Human RPE cells (ARPE-19) were obtained (American Type Culture Collection,Manassas, Va). This cell line is not transformed and has structural and functionalproperties characteristic of RPE cells in vivo.²³ TheRPE cells were cultured in DMEM/F12 medium (1:1) containing 10% fetal bovineserum (Invitrogen Corp, Grand Island, NY). The following substances were added:0.01 g/L of transferrin, 0.01 g/L of insulin, 0.91 g/L of sodium bicarbonate,100 U/mL of penicillin G potassium, 0.1 mg/mL of streptomycin sulfate, 5.0mg/mL of gentamicin sulfate, 3.58 g/L of HEPES, and 1.75 g/L of D-glucose.The cells were cultured at 37°C in 5% carbon dioxide.

Indocyanine green was prepared by completely dissolving 25.0 mg of sterileICG powder (containing iodide) (Daiichi Pharmaceutical, Tokyo, Japan) in 0.5mL of sterile distilled water. To achieve a final ICG concentration of 2.5mg/mL (289.8 mOsm/kg), 9.5 mL of balanced salt solution (BSS Plus; Alcon Laboratories,Fort Worth, Tex) was added.

The sodium-free BSS solution was prepared in the laboratory with thesame constituents as BSS, except that sodium chloride, sodium phosphate dibasic,and sodium bicarbonate were replaced with 17.06 g/L of choline chloride, 0.52g/L of potassium phosphate dibasic, and 2.50 g/L of potassium bicarbonate,respectively, to achieve the same molar concentration of the replaced chemicals.The pH was adjusted to 7.4 with hydrochloric acid/potassium hydroxide (potassiumhydroxide was used instead of sodium hydroxide to keep the solution sodium-free).The solution was then sterilized by using a disposable sterilizing filter(0.22-mm-diameter polyethersulfone bottle top filter; Corning Costar, Corning,NY). The osmolarity of this solution was 296 mOsm/kg, as measured by an auto-osmometer(Osmostat, model OM-6020; Daiichi Kagaku Co, Kyoto, Japan). Indocyanine greenwas also dissolved in sodium-free BSS to achieve a final concentration of2.5 mg/mL (281.2 mOsm/kg).

Retinal pigment epithelial cells were seeded into the wells of chamberslides or culture plates in appropriate density and volume. Retinal pigmentepithelial cells were grown to total or 70% confluence (depending on the assay)for ICG treatment and light illumination. After removal of culture medium,the cells were rinsed with prewarmed (37°C) BSS or sodium-free BSS (thesame as the ICG solvent used in ICG preparation). Indocyanine green (2.5 mg/mL,prewarmed to 37°C) in BSS or sodium-free BSS was then added and incubatedfor 2 minutes at 37°C in the dark. Balanced salt solution or sodium-freeBSS alone was added to the control wells. The RPE cells were then washed withBSS or sodium-free BSS. The cells were then placed in BSS and subjected tointense light exposure of a surgical microscope. The light intensity was adjustedso that the illuminance at the level of RPE cells was 4 × 10⁴ lux(comparable to the illuminance from an endoillumination probe, as measuredby a lux meter; Lutron, Coopersburg, Pa). The cells were exposed to the lightfor 40 minutes. After illumination, the RPE cells were washed with BSS, culturemedium was added, and the cultures were maintained at 37°C with 5% carbondioxide in the dark for predetermined periods.

Retinal pigment epithelial cells were cultured in 2-well chamber slides.Each well was seeded with 1.0 mL of cell suspension (10⁵ cells/mL).After reaching total confluence, RPE cells were treated with ICG and light.Phase-contrast microscopy was used to observe the morphologic changes. Forcalcein AM–ethidium homodimer 1 staining, the culture medium was removedat appropriate periods after the ICG and light treatment. The samples wererinsed with PBS and then incubated with a solution containing 2-µM calceinAM and 4-µM ethidium homodimer 1 (Molecular Probes, Eugene, Ore) for45 minutes. The cells were evaluated under an epifluorescence microscope.

Cells (5 × 10⁴ in 0.5 mL) were seeded into each wellof the 24-well plates. After reaching 70% confluence (to avoid contact inhibitionduring the observation period of 48 hours), the cells were treated with ICGincubation followed by illumination. The number of viable cells was determinedat 2, 6, 12, 24, and 48 hours after the treatment. The RPE cells were trypsinizedand stained with trypan blue (0.2%). Cells that excluded trypan blue wereconsidered viable and counted with a hemocytometer.

Cell viability was assessed by MTS colorimetric assay (3-[4,5-dimethylthiazol-2-yl]-5-[3-carboxymethoxyphenyl]-2-[4-sulfophenyl]-2H-tetrazolium)(Promega Corporation, Madison, Wis). This quantitative assay detects livingbut not dead cells.²⁴ The absorbance at 490nm (test wavelength) and at 650 nm (reference wavelength) was measured usingan enzyme-linked immunosorbent assay microplate reader (VERSAmax; MolecularDevices, Sunnyvale, Calif); wells containing culture medium but no cells servedas blanks. In this experiment, 10⁴ cells in 100 µL of culturemedium were seeded into each well of a 96-well plate. After achieving 70%confluence, the cells were treated with ICG and light. After incubation for2, 6, 12, 24, and 48 hours, the wells were washed, and 100 µL of culturemedium and 20 µL of MTS were added to each well. After a 2-hour incubationat 37°C, the absorbance at 490 nm was measured.

The ICG uptake was measured by using the optical absorption of ICG at790 nm.²¹ The absorption spectrum of ICG in0.1N sodium hydroxide (the medium used to lyse the RPE cells) was measuredwith a spectrophotometer (DU800 UV/Visible; Beckman Coulter, Fullerton, Calif)to ensure a high absorption coefficient at 790 nm in different ICG concentrations.For the ICG uptake assay, 1.5 × 10⁵ cells in 1.5 mL of culturemedium were seeded into each well of the 6-well culture plates. After reachingtotal confluence, the cells were rinsed 3 times with prewarmed (37°C)BSS or sodium-free BSS (the same as the ICG solvent used in ICG preparation).Indocyanine green (2.5 mg/mL, prewarmed to 37°C) in BSS or sodium-freeBSS was then added and incubated for 2 minutes at 37°C in the dark. Thereactions were then terminated by aspiration of the ICG solution, followedby 4 rapid washings of the RPE cells with ice-cold BSS or sodium-free BSS.This washing procedure removed more than 99% of extracellular substrate, withminimal loss of cell-associated substrate.²⁵ Thecells in each well were then lysed with 0.8 mL of 0.1N sodium hydroxide. Theoptical absorption of the lysate containing ICG was measured at 790 nm. Theactual ICG concentration in the cell lysate was calculated using a calibrationcurve. To create the calibration curve, appropriate amounts of ICG were dissolvedin 0.1N sodium hydroxide to make solutions of known ICG concentrations. Theoptical absorption at 790 nm for each ICG concentration was measured, andthe calibration curve was created by linear regression analysis.

Cultured human RPE cells showed morphologic changes 2 hours after treatmentwith 2.5 mg/mL of ICG in BSS for 2 minutes and subsequent exposure to lightfor 40 minutes. These cells assumed a heterogeneous appearance. Some cellsbecame round. Some cells were reduced in size (Figure 1A). Some cells detached from the bottom of the culture welland floated in the medium, leaving space between the cells that attached onthe bottom of the culture well. The extent of these changes increased 6 hoursafter treatment, and more cells assumed a shrunken appearance (Figure 1B). After exposure to 2.5 mg/mL of ICG in sodium-free BSSfor 2 minutes followed by illumination for 40 minutes, some cells became round2 hours after the treatments. Many cells were somewhat reduced in size, anda few cells detached from the bottom of the culture well, producing gaps betweenthe cells (which were totally confluent before treatment). However, thesechanges were less severe than previously observed (Figure 1C). These changes were slightly more evident 6 hours aftertreatment; more cells became round (Figure1D). However, the changes observed after exposure to ICG in sodium-freeBSS were less severe than those produced by ICG in BSS. As a control, RPEcells exposed to light showed no morphologic abnormality (Figure 1E).

To investigate ICG-associated photosensitizing toxicity, RPE cells werestained with 2 µM of calcein AM and 4 µM of ethidium homodimer1. Live cells are identified by the presence of ubiquitous intracellular esteraseactivity, which converts the nonfluorescent, cell-permeant calcein AM to green-fluorescentcalcein, which is retained within live cells. Ethidium homodimer 1 entersthe cells with compromised membranes and, on binding to nucleic acid, providesa red fluorescence in dead cells.²⁶

Some cells had shrunken and red-fluorescent nuclei with reduced intracytoplasmicgreen fluorescence 2 hours after treatment with 2.5 mg/mL of ICG in BSS for2 minutes and subsequent exposure to light for 40 minutes. Their cell membraneintegrity was compromised, and intracytoplasmic esterase activity was reduced.These cells were dead. Other cells showed intense green fluorescence in thecytoplasm, indicating they were viable (Figure2A). More RPE cells showed red-fluorescent nuclei 6 hours aftertreatment. The green fluorescence was faint or absent in the cytoplasm ofthese cells containing red-fluorescent nuclei (Figure 2B).

A smaller fraction of cells treated with 2.5 mg/mL of ICG in sodium-freeBSS for 2 minutes followed by illumination for 40 minutes contained condensedand red-fluorescent nuclei, compared with cells treated with ICG in BSS. Mostcells remained green-fluorescent 2 hours after treatment (Figure 2C). Six hours after treatment, there was a slight increasein the number of cells with red-fluorescent nuclei (Figure 2D). However, the ratio of cells with red-fluorescent nucleiwas much less compared with treatment with ICG dissolved in BSS. Retinal pigmentepithelial cells exposed only to illumination for 40 minutes showed intensegreen fluorescence in the cytoplasm (Figure2E), as did cells exposed only to 2.5 mg/mL of ICG (in BSS or sodium-freeBSS) for 2 minutes (data not shown).

The viable cell number of RPE cells was reduced by treatment with 2.5mg/mL of ICG (in BSS and sodium-free BSS) for 2 minutes plus intense lightillumination for 40 minutes (Figure 3A).It was reduced to a greater extent when 2.5 mg/mL of ICG was dissolved inBSS than in sodium-free BSS. After light exposure, the number of viable cellsin the ICG BSS group was one fourth to one third of that in the control group(no ICG and no light exposure). In contrast, this ratio was around three fourthsfor the ICG sodium-free BSS group. Cell viability was significantly reducedin the ICG BSS group, compared with the ICG sodium-free BSS group, at eachtime point. Viability was not affected by illumination treatment for 40 minutesonly or incubation in 2.5 mg/mL of ICG (in BSS or sodium-free BSS) for 2 minuteswithout illumination.

Enzymatic integrity in cultured human RPE cells was evaluated with anMTS colorimetric assay. The MTS tetrazolium compound is reduced to a coloredformazan product by a nicotinamide adenine dinucleotide phosphate–dependentor nicotinamide adenine dinucleotide–dependent dehydrogenase in metabolicallyactive cells.²⁷ The formazan was quantitatedwith an enzyme-linked immunosorbent assay microplate reader at 490 nm. Indocyaninegreen (2.5 mg/mL) incubation for 2 minutes followed by light treatment for40 minutes significantly reduced dehydrogenase activity in the ICG BSS andICG sodium-free BSS groups (Figure 3B).Treatment with ICG in BSS produced a greater photosensitizing toxic effectthan ICG in sodium-free BSS. Exposure to 2.5 mg/mL of ICG for 2 minutes orlight for 40 minutes alone did not affect enzymatic activity.

Indocyanine green dissolved in 0.1N sodium hydroxide had high absorptionat 790 nm at different ICG concentrations (Figure 4). We measured the amount of ICG uptake by RPE cells indifferent solvents to elucidate the mechanism by which sodium affected ICG-associatedphotosensitizing toxicity. A calibration curve was created to correlate theICG concentration with optical absorption at 790 nm. The mean ± SEICG uptake by RPE was 1.81 ± 0.24 µg/10⁵ cells forRPE cells incubated with 2.5 mg/mL of ICG dissolved in BSS for 2 minutes at37°C. This value was significantly decreased (P =.01) to a mean ± SE of 1.05 ± 0.17 µg/10⁵ cellsfor RPE cells incubated for 2 minutes with 2.5 mg/mL of ICG dissolved in sodium-freeBSS (Figure 5).

In our study, incubation in 2.5 mg/mL of ICG (in BSS) for 2 minutesfollowed by 40 minutes of illumination produced toxicity in cultured humanRPE cells. This ICG-associated photosensitizing toxicity was reduced by replacingsodium in the ICG solvent with other cations. Reduced toxicity was associatedwith decreased ICG uptake into RPE cells.

The cause of the postoperative atrophic RPE changes¹⁵ andthe poorer visual outcome^14-16 afterICG-assisted ILM peeling is not yet clear. These observations may be due tocytotoxic (including photosensitizing toxicity) effects of ICG to RPE cells.When ICG is used to stain the ILM during macular hole surgery, ICG has directaccess to the bare RPE at the base of the macular hole. Although ICG is appliedfor brief periods up to 5 minutes^5-13,15,16 andthen washed out, ICG can be taken up by the RPE cells during this application,as we demonstrated. Furthermore, residual ICG may remain in the eye for prolongedperiods that can extend up to 8 months.^28-31 Endoilluminationwith a high level of light intensity is used during macular hole surgery,and photosensitizing damage to RPE cells can occur.

Although ICG is a commonly used dye with a long history of safety andlow toxicity,^32,33 it is mostoften administered intravenously. Indocyanine green is an organic anion thatis taken up by liver cells after intravenous injection and rapidly clearedfrom the circulation via bile secretion. The clearance of ICG in blood isbiphasic, with a rapid initial phase (half-life, 3-4 minutes) and a secondaryphase (half-life, >1 hour).³⁴ Because the barriersurrounding the vitreous compartment is tight,³⁵ thepharmacokinetics of intravenous and intravitreal ICG are probably different.In fact, scanning laser ophthalmoscopy revealed ICG fluorescence in the posteriorpole of patients with macular holes, and it remained for several months postoperatively.^28-31 Besides,the ICG uptake rate by RPE was higher than that in other cells, such as skinkeratinocyte, as evidenced by comparing our results of ICG uptake assay withthose of others.²¹ Therefore, the safety ofICG, as assessed from intravenous studies,³² maynot be relevant to intravitreal applications.

In this study, removing sodium from the solvent reduced ICG-associatedphotosensitizing toxicity, and the underlying mechanism may involve reduceduptake of ICG by RPE. The cellular uptake of organic anions, such as ICG,involves sodium-dependent and sodium-independent transport systems.²⁰ Because ICG can be taken up by RPE and lead to photosensitizingdamage under intense light exposure, we replaced sodium in the solvent withother cations, and this change reduced ICG uptake into RPE cells and its associatedphotosensitizing damage. The actual mechanism through which ICG is taken upby RPE cells remains unclear. Recently, an organic anion transporter, organicanion transporting polypeptide 2, was found at the apical region of the ratRPE.³⁶ This transporter mediated the transportof many endogenous and exogenous amphophilic compounds in a sodium-independentmanner.²⁰ However, in our study, ICG uptakeby RPE cells was partially sodium-dependent. One possible explanation is thatanother transporter, which is sodium-dependent, is present on RPE cells andis responsible for the sodium-dependent portion of ICG transport. Anotherpossibility is that the ICG uptake is coupled to a secondary active sodium-involvingcotransporter, which is driven by a sodium gradient, just like the cholateuptake in hepatocyte.²⁵

It was shown recently that there was a difference in light absorptionqualities when ICG was dissolved in different solvents.³⁷ Toinvestigate whether this difference accounted for the reduced photosensitizingtoxicity of ICG sodium-free BSS solution observed in this study, we measuredthe emission spectrum of the light source with an optical spectrum analyzer(Figure 6) and the absorption spectraof ICG BSS and ICG sodium-free BSS with a spectrophotometer (Figure 7). There was a significant overlap of the emission spectrumof the light source and the absorption spectrum of the ICG solutions (in BSSand sodium-free BSS) between 570 and 800 nm. In addition, ICG sodium-freeBSS had a slightly higher absorption coefficient than ICG BSS at wavelengthsbetween 570 and 800 nm when the ICG concentrations were identical. This indicatedthat ICG sodium-free BSS had higher absorption of the light source used inthis study than ICG BSS, if ICG concentrations were the same. Conversely,ICG sodium-free BSS showed less photosensitizing toxic effects on RPE cells,as demonstrated in this study. Therefore, the reduction in photosensitizingdamage of ICG sodium-free BSS compared with ICG BSS was not due to the differencein the absorption qualities of these 2 solutions.

In this study, RPE cells were illuminated for 40 minutes. We chose thisillumination duration because it demonstrated the distinction of the photosensitizingtoxic effect on RPE cells between ICG BSS and ICG sodium-free BSS. In fact,we observed this difference in the morphologic structure of cells and cellnuclei when the light exposure duration was 25 minutes, but to a lesser extentand in a smaller number of cells. The results in this study emphasize thatICG dissolved in sodium-free BSS had a higher safety margin than ICG dissolvedin BSS in terms of photosensitizing toxicity. That is, using sodium-free BSSas a solvent for ICG reduced the photosensitizing damage in RPE cells, evenwhen the light exposure duration was prolonged to 40 minutes. Clinically,removing the ILM is a technically challenging procedure, especially for inexperiencedsurgeons. It is not uncommon for an inexperienced surgeon to take as longas 40 minutes to complete the ILM peeling procedure, even when ICG stainingis used. In addition, there are other procedures to perform after ILM removalis completed. These include fluid-air exchange, air-gas exchange with sulfurhexafluoride or perfluoropropane, and closure of the sclerotomy and peritomy.All these procedures are performed under light illumination (from an endoilluminationprobe or surgical microscope).

Cytotoxicity of ICG may contribute to the clinically observed RPE changesafter ICG-assisted ILM peeling, because ICG has been shown to decrease mitochondrialenzyme activity in RPE cells in vitro¹⁷ andcause mitochondrial toxicity.³⁸ In addition,intravitreally administered ICG may be retained in the eye for a prolongedperiod.^28-31 Prolongedexposure to ICG, even at a low concentration, can lead to cytotoxicity inRPE cells.³⁹ However, because sodium removalcan reduce ICG uptake, we believe that sodium removal would also reduce thecytotoxicity induced by prolonged ICG exposure.

It has been suggested that the toxic effects of ICG on RPE may be relatedto the hypoosmolarity of the solvent.⁴⁰ Inour study, the osmolarities of the solutions containing 2.5 mg/mL of ICG were289.8 mOsm/kg in BSS and 281.9 mOsm/kg in sodium-free BSS. To examine theeffects of a hypoosmotic medium on RPE cells, we diluted BSS and sodium-freeBSS with sterile distilled water to osmolarities of 289.8 mOsm/kg and 281.9mOsm/kg, respectively. Exposure to diluted BSS or sodium-free BSS alone orfollowed by light exposure caused no alteration in RPE cell morphologic structure,calcein AM–ethidium homodimer 1 staining, or viable cell number counting(data not shown). Therefore, a change in osmolarity is probably not responsiblefor the ICG-induced RPE toxicity observed in vitro and clinically.⁴¹

A recent study⁴² describes adverse effectson functional outcome of ICG-assisted macular pucker surgery, in which therewere more cellular debris and inner processes of Müller cells in theremoved specimens after ICG application during surgery. Epiretinal cells hadruptured and lost their cellular integrity. In macular pucker surgery, thereis no direct contact of the RPE and the ICG solution. Therefore, ICG may betoxic to retinal cells other than RPE cells, and this toxicity may also contributeto the less favorable functional outcome observed after ICG-assisted macularhole surgery.

The sodium-free BSS used in this study was prepared in the laboratory.To our knowledge, it is not available commercially. There are potential concernsabout quality control and safety for this laboratory-prepared solution. Theseconcerns include stability of osmolarity and pH, sterility, and shelf lifeof the solution. These concerns should be fully addressed before this solutioncould be used in human surgery.

Indocyanine green staining of the ILM facilitates its identificationand removal during macular hole surgery. Indocyanine green staining can bea valuable tool; however, no standardized procedure that specifies reconstitutionmethod, concentration, volume, and incubation time exists for the intravitrealuse of ICG. In this study, we demonstrated that replacement of sodium in thesolvent reduces the ICG-associated photosensitizing damage on cultured humanRPE. Further in vivo studies are required to examine if this reconstitutionmethod can provide a safer intravitreal ICG use in macular hole surgery.

Corresponding author and reprints: Ray J.-F. Tsai, MD, Departmentof Ophthalmology, Taipei Medical University Hospital, Second Floor, Building350, Section 4, Cheng Kung Road, Taipei 114, Taiwan (e-mail address: raytsai@ms4.hinet.net).

Submitted for publication April 15, 2003; final revision received October8, 2003; accepted October 29, 2003.

Dr Ho had full access to all the data in the study and takes responsibilityfor the integrity of the data and the accuracy of the data analysis.