Anal. Chem. 1995,67,3799-3801
Failure of =Si NMR To Detect Increased Blood Silicon Levels in Silicone Gel Breast Implant Recipients Peter Macdonald,*lt Nick Plavac,t Walter Peters,* Stanley Lugowski,@and Dennis Smiths Department of Chemistry, Division of Plastic Surgery, and Centre for Biomaterials, Universiiy of Toronto, Toronto, Ontario, Canada M5S 1A2
We have compared directly the results of atomic absorption (AA) spectroscopy and a 2gSimagic angle spinuing (MAS) nuclear magnetic resonance (NMR) technique reported in the literature by Garrido et al. (Garrido, L; et al. Magn. Reson. Med. 1994, 31, 328-330) for analyzing blood silicon levels in control patients versus patients with silicone gel breast implants. AA spectroscopy yielded blood silicon levels in the nanogram per milliliter range for control patients,while somewhat higher values were found in patients with implants. The 29Si MAS NMR technique applied to the identical blood samples was unable to detect silicon in any of the samples. Sensitivity calculations demonstrate that 29SiMAS NMR should not be expected to detect silicon at the levels determined by AA spectroscopy under the spectroscopic conditions employed and that the concentration of siliconcontaining compounds would need to lo4limes the level detected by AA in order to be detected by this NMR method. The fate of silicone leaked from silicone gel breast implants has become the focus of intense scrutiny. Free silicone can “bleed into the tissues immediately adjacent to and can migrate further via the lymphatic system? The silicones, or polysiloxanes, of which implant materials are composed are generally rather inert chemically. Nevertheless, immunological sensitization to silicone could possibly contribute to the develop ment of autoimmune connective tissue diseases4 A recent report by Ganido and colleagues suggested that biodegradation and chemical modification of silicone accompanies its migration among tissues and organs and that measurable amounts of silicon, on the order of 100 mM total silicon (corresponding to approximately 2.8 mg/mL), were found circulating in the blood of some implant patients5 The studies of Garrido et aL5were conducted using 29Simagic angle spinning (MAS) nuclear magnetic resonance (NMR) spec* Phone: 905-828-3805. Fax: 905-828-5425. E-mail:
[email protected]. utoronto.ca. Department of Chemistry. Division of Plastic Surgery. 8 Centre for Biomaterials. (1) Nelson, G.D. Plast. Reconstr. Suw. 1980,66,969-970. (2) Barker, D. E.; Retsky, M. I.; Schultz, S. Plast. Reconstr. S u q . 1978,61, 836-841. (3) Hausner, R J.; Schoen, F. J.; Pierson, K K Plast. Reconstr. S u q . 1978, 62,381-384. (4) Peters, W.Ann. Plast. S u q , 1995,34, 103-109. (5) Garrido, L.; Weiderer, B.; Jenkins, B. G.; Hulka, C. A; Kopans, D. B. Magn. Reson. Med. 1994,31, 328-330. +
0003-2700/95/0367-3799$9.00/0 0 1995 American Chemical Society
troscopy to assess levels of silicone in blood and to identify any chemical modifications that had occurred. NMR spectroscopy is ideally suited for chemical identification, because there is a rich variety of spectral parameters sensitive to structural and dynamic details and because it is possible to perform the analysis from withii the midst of a complex mixture, such as a biological fluid. However, NMR spectroscopy is not a particularly sensitive technique in comparison to others6 When lower absolute amounts of an analyte are present, sensitivity considerations inevitably restrict the utility of NMR spectroscopy. More recent atomic absorption (AA) spectroscopy measurements of blood silicon levels indicate that total elemental silicon concentrationsfallin the 10-20 ng/mL range for control patients, while statistically significant higher levels (20-40 ng/mL) are found in patients with silicone gel breast implants7 Nevertheless, even the highest levels of silicon determined using AA are 5 orders of magnitude lower than those reported using %i MAS NMR5 In order to address this discrepancy between the findings of the two techniques, we have performed a direct comparison of blood silicon levels determined by AA and 29Si MAS NMR analyses. Levels of blood silicon were determined in both control patients and patients having silicone gel breast implants. The details of the AA and 29Si MAS NMR techniques employed matched as closely as possible those described in the conflicting reports. EXPERIMENTAL SECTION Blood Collection. Blood was collected from seven volunteers having silicone gel breast implants and six control patients without implant exposure. No patients with diabetes or autoimmune disease, or patients taking any known silicatecontainingmedications, were included in the study. The implants had been in place for a median of 13 years (range 7-20 years). Five of the seven patients subsequently demonstrated implant rupture at explantation. All blood samples were taken in a fashion designed to eliminate silicon contamination, using a meticulously clean collection room and “siliconedecontaminated syringes and plasticware as described previ~usly.~ Heparinized blood samples were stored at 4 “C prior to analysis by AA and ?Si NMR spectroscopies. AA Spectroscopy. Silicon measurements were carried out with a Varian Model 875 atomic absorption spectrophotometer (6) Abragam, k Principles of Nuclear Magnetism; Oxford Press: Hong Kong, 1961. (7) Peters, W.; Smith, D.; Lugowski, S.; McHugh, A Ann. Plast. Sue. 1995, 34, 343-347.
Analytical Chemistry, Vol. 67,No. 20,October 15, 1995 3799
Table 1. Blood Silicon Levels in Control versus Implant Patients Determined via Atomic Absorption Spectroscopy
controls WmL) number minimum maximum mean median standard deviation standard error
implants WmL)
6
7
5.30 57.70 25.42 21.75 20.41 8.33
7.10 57.50 27.11 25.31 17.77 6.72
equipped with a graphite furnace Model GTA 95 and autosampler, using procedures and protocols described el~ewhere.~ 29Si MAS NMR. 29Si NMR spectra were recorded under conditions matching as closely as possible those described by Ganido et a1.j Specifically, spectra were obtained using a Varian 9.4 T VXRS 400 NMR spectrometer equipped with a Doty MAS probe tuned to the resonance frequency of 29Si (79.459 MHz). Typically, a 200 pL volume of whole blood was transferred into a standard Doty 7 mm (id.) zirconium MAS rotor, and the 29SiNMR spectrum was acquired using acquisition parameters as follows: slow spinning of the MAS rotor (-1OOO Hz),singlepulse excitation with a 3.5 ps radio-frequency pulse (-45O tip angle), a recycle delay of 10 s, with a sweep width of 33 000 Hz digitized into 2048 data points. Eight thousand such scans were signal-averaged (-22 h total acquisition time). RESULTS AND DISCUSSION AA Spectroscopy. Table 1summarizes the results of the AA spectroscopy measurements of silicon levels in blood samples for the control patients versus those with silicone gel breast implants. For the control group, blood silicon levels were in the nanogram per milliliter range, the mean value being 25.42 ng/mL, in agreement with previous determinations using the same meth~ d o l o g y . ~For the implant group, blood silicon levels were somewhat higher, the mean value being 27.11 ng/mL, again agreeing with previous determinations in a similar, albeit larger, patient group7 29SiMAS NMR Spectroscopy. Figure 1 shows a series of 29Si MAS NMR spectra obtained as described in the Experimental Section. Spectrum A is a positive control acquired using silicon oil @ow Corning 360 Medical Fluid; viscosity 350 cs, average molecular weight 12 OOO). A single broad resonance line is visible (6 = -21 ppm relative to tetramethylsilane, TMS) and is assigned to the dimethylsiloxanes of poly(dimethylsi1oxane) .* Other possible silicon resonances, such as hydrolyzed and methylterminated silicone, or silica, or high-coordinated silicon, are not observed. Either these compounds are present in amounts too small to be detected or their signals are broadened beyond the detection limit under the processing conditions used to generate the 29SiNMR spectra. Of particular note is our use of an exponential multiplication of the free induction decay signal prior to Fourier transformation, corresponding to a line broadening of 400 Hz, exactly as employed by Ganido et a1.j Spectrum B is typical of the blood samples tested here in that the only 29SiNMR resonance that is visible is a broad, featureless (8) Coleman, B. NMR of Newly Accessible Nuclei; Laszlo, P., Ed.; Academic Press: New York, 1983: Vol. 2, p 197.
3800 Analytical Chemistry, Vol. 67,No. 20, October 75, 7995
D l " " r " " ~ " " ~ " " ~ " " " " " " " " ' " ' 1w 108 so -sa
-1-
-I=
-me
P P
Figure I. 29SiMAS NMR spectra: (A) 200 pL of silicon oil, 4 scans, line broadening (Ib) 400 Hz, no data points removed from the FID, and vertical scale (vs) 1500; (B) 200 p L of whole blood from implant patient, 8000 scans, Ib 400 Hz, no data points removed from the FID, and vs 75 000; (C) 200 pL of whole blood from implant patient [same sample as (B)], 8000 scans, Ib 400 Hz, first 10 data points removed from the FID, and vs 75 000; (D) 200,uL of whole blood from implant patient [same sample as (B)], 8000 scans, Ib 400 Hz, first 10 data points removed from the FID, and vs 600 000; selected integration and phasing of particular regions of spectrum as described in the text.
line centered at --lo0 ppm, corresponding to the expected resonance frequency of silicon dioxide (otherwise known as silica). This same broad resonance line is observed in all blood samples, both implants and controls, and even in a blank sample containing only water. It must arise, therefore, from background silica present in the NMR probe itself. A common means of removing such broad background resonances from NMR spectra is to eliminate the first few data points from the free induction decay @ID) prior to Fourier transformation. When we removed the first 10 data points of the FID, exactly according to Garrido et a1.j (Garrido, L., personal communication), we obtained spectrum C shown in Figure 1, from which the broad resonance assigned to background silicon in the probe is now clearly absent. More significant is the absence of any other silicon resonances, whatsoever, under these acquisition and signal processing conditions. The same result is obtained in all blood samples tested, whether from control patients or patients having silicone gel breast implants. We would mention, in addition, an eighth patient, who had subcutaneous silicone injections rather than silicone gel implants. A4 spectroscopy results indicate that blood silicon levels for this patient were in the high end of the range displayed by the silicone gel breast implant group. (The AA results for this eighth patient are not included in the data or the statistical analysis of Table 1.) Even using blood from this eighth patient, we were unable to detect any 29SiNMR signal. One difficulty arising from eliminating the initial data points in an FID is the necessity of correcting for the substantial firstorder phase shift that is an unavoidable side effect of employing such a signal-processing routine. Spectrum D in Figure 1shows that it is possible to generate what appear to be broad resonance signals in the %i NMR spectrum simply by employing the spectral baseline correction and phasing routines which are a standard
part of the data-processing sofhvare offered by every modem NMR spectrometer manufacturer. In the spectrum shown, we chose four particular regions of the spectrum for integration. After selecting the desired range of frequencies for integration, the phase of the signal within that region is adjusted as desired, and the computer “corrects” the intervening regions for apparent baseline anomalies. The resulting phased and baseline-corected spectrum apparently contains resonances withiin the selected frequency ranges. These arise purely as a result of computer manipulation. The difficulty here originates with the fact that there is no “real” signal in the spectrum upon which to base a judgment regarding the correct phase and baseline treatment. Note that we could have chosen any region of the spectrum, performed the same manipulations, and generated similar anomalies. A simple calculation supports our experimental finding that 29SiNMR, under the conditions described by Gamdo et al.5 and reproduced here, is incapable of detecting levels of silicon consistent with those reported by AA spectroscopy. The detection limit is a function of the signal to noise ratio (S/N) in the NMR spectrum. Signal increases directly with the number of NMRsensitive nuclei in the sample coil of the NMR probe and the number of scans that are averaged together. Noise, on the other hand, increases only as the square root of the number of scans. Consequently, S/N increases with the square root of the number of scans. The spectrum of silicon oil in Figure 1has a S/N ratio of 33, obtained with a 200 pL volume and signal averaging just four scans. Using the density of silicon oil (1.05g/mL), the weight fraction of Si in poly(dimethy1siloxane) (37.8 wt %),and the natural abundance of 29Si (4.7%),one calculates that such a volume of silicon oil contains -3.4 mg of 29Si. If the number of scans were increased to SO00 from 4, then the S/N ratio would increase by a factor of (8000/4)1/2to equal -1476. The absolute lower limit of sensitivity corresponds to a S/N ratio of 1. Under conditions identical to those under which the various blood samples were tested, this would equate to approximately (3.4 mg /1476) = 2.3 pg of 29Si/200pL, corresponding to -245 pg/mL total elemental silicon. For signal detection a S/N ratio of 2 or more is desirable, ~
,
~
~
(9) Wu, N.; Peck, T.; Webb, A; Magin, R; Sweedler, J. Anal. Chem. 1994, 66, 3849-3857.
(10) Black, R D.; Early, T. A; Roemer, P. B.; Mueller, 0. M.; MogreCampero, A; Turner, L. G.; Johnson, G.A Science 1993,259, 793.
while for quantification a minimum S/N of 3 is generally considered necessary. Clearly, these %SiNMR sensitivity limits are a factor of 104 higher than the blood silicon levels determined from AA spectroscopy. We would point out that a further calculation shows that to detect such low levels of silicon using the described %i NMR technique would require multiple centuries of signal acquisition time. Even allowing for differences in the sensitivity of one NMR probe versus another, and the approximate nature of the calculation, it seems impossible that %i MAS NMR could detect the silicon levels reported by AA spectroscopy. More appropriate conditions for 29Si NMR spectroscopy of blood samples than those described by Garrido et al.5 are readily conceived. In particular, MAS of liquid samples is a poor means of eliminating 29Si-1H dipolar coupling, one of the main sources of loss of S/N through line broadening in %i NMR spectra. A better approach would be to directly decouple ‘Husing a radiofrequency decoupling field while employing an inverse-gated sequence to avoid deleterious nuclear Overhauser enhancements due to silicon’s spin I = -‘/2. Larger sample sizes and a higher sensitivity solution-state NMR probe would also seem to be a necessary prerequisite to better signal-to-noise ratios. Finally, recent advances in NMR radio-frequency coil designgJO may eventually permit routine detection of such low levels of analyte, but these are not yet widely available. The question remains as to why Ganido et al.5 observed silicon NMR resonances in their blood samples. If the published spectra were due to real signals rather than artifacts from data processing, then contamination from extraneous sources during the collection and/or sample-handling process seems the most likely explanation. ACKNOWLEDGMENT
This research was funded by a Medical Research Council of Canada University Industry Grant, by Dow Coming, and by a Trillium Clinical Research Grant. Received for review May 9, 1995. Accepted August 1, 1995.@
AC950448W @
Abstract published in Advance ACS Abstracts, September 1, 1595.
Analytical Chemistty, Vol. 67, No. 20, October 15, 1995
3801
