1.

Introduction

Optical technologies provide a promising new approach to real-time, in vivo diagnosis of epithelial precancers, in both the screening and diagnostic settings. A number of groups have examined novel optical approaches, including fluorescence spectroscopy,^{1 2 3 4 5} reflectance spectroscopy,^{6 7 8} Raman spectroscopy,⁹ confocal imaging,^{10 11} and optical coherence tomography.¹² Of these, the largest clinical experience has been reported using fluorescence spectroscopy.

Fluorescence spectroscopy for the screening and diagnosis of cervical precancer has demonstrated promising results. Clinical trials show a sensitivity and specificity of 86 and 74 in the diagnostic colposcopy clinic¹³ and 75 and 80 in the screening setting.¹⁴ In both diagnostic and screening settings, peak fluorescence intensity varies by more than an order of magnitude from one patient to another. In an attempt to further improve the sensitivity and specificity of real-time diagnosis, it is necessary to understand and control for the sources of this inter-patient variability. Figure 1 shows how fluorescence intensity can vary in squamous normal tissue of the cervix from one patient to another. Plotted here is the fluorescence intensity at 340 nm excitation, 440 nm emission from a previous series of 50 squamous normal sites from 25 patients.¹³ Patients are listed in the order of increasing age, and the age of the patients ranges from 18 to 60. In general, the intra-patient variation is less than the inter-patient variation. However, the biological basis for these variations is not well understood. Previous work¹⁵ has shown that race and smoking do not account for these variations, while age and menopausal status may play a role. In Figure 1, patients 23, 24, and 25 show a large increase in fluorescence intensity compared to the rest of the patients. It is interesting to note that the patients have ages 49, 50, and 60, respectively, and that patients 24 and 25 are postmenopausal. We are conducting large clinical trials in the diagnostic and screening settings to examine the influence of age on fluorescence spectroscopy. However, it is unknown whether the phase in the menstrual cycle may also play a role.

Figure 1

Fluorescence intensity measured in vivo from squamous epithelium of the cervix at 340 nm excitation, 440 nm emission from 25 patients from a previous study. Multiple sites from a single patient are linked with lines. The patients are listed in the order of increasing age.

A more thorough understanding of the correlation between the menstrual cycle and fluorescence spectroscopy of the cervix is necessary to explore the inter-patient variation. The goal of the clinical study described in this paper was to examine the variations in the fluorescence spectra throughout the menstrual cycle and to determine the range of effect the menstrual cycle had on the fluorescence of the cervix.

2.

Material and Methods

3.

Results

Ten patients enrolled in and completed the study. The patients ranged in age from 26 to 45. All ten patients were pre-menopausal as determined by FSH levels, and nine of ten had ovulatory bleeding by medical history and by estradiol and progesterone levels. One patient (patient 2) had an anovulatory cycle. Nine out of the ten patients had normal Pap smears. One of the ten patients (patient 10), who had a previous history of normal Pap smears, had stage 1 cervical intraepithelial neoplasia (CIN 1)/HPV-associated atypia. Eight of the ten patients participated in a second study and underwent colposcopically directed biopsies on two of the measured sites at the end of the trial. One of these patients displayed colposcopic abnormality and biopsies of all three measured sites were obtained at the end of the trial. Out of the eight patients, seven patients had normal cervical biopsies. The patient who had CIN 1/HPV-associated atypia on her Pap smear had the diagnosis of low grade SIL confirmed in all three biopsies. Each patient underwent fluorescence EEM measurements daily of three sites. For three patients (patient 2, 6, 10), one site was columnar epithelium and two sites were squamous epithelium. For seven other patients, all three sites were squamous epithelium.

Figure 2 shows a typical fluorescence EEM from a normal squamous site on cycle day 7. The fluorescence EEM is plotted as a topographical map, with excitation wavelength on the ordinate and emission wavelength on the abscissa. Contour lines connect points of equal fluorescence intensity. Several excitation-emission maxima are present; the peak at 350 nm excitation, 450 nm emission is consistent with emission of the cofactor NADH as well as collagen cross-links. A shoulder at 370 nm excitation, 525 nm emission is consistent with emission of the cofactor FAD. The peak at 450 nm excitation, 525 nm emission is consistent with the cofactor FAD as well as structural protein fluorescence. Fluorescence of endogenous porphyrins is present with excitation maxima at 410 nm and emission maxima at 630 nm. Tissue vascularity can influence fluorescence spectra, when hemoglobin absorbs fluorescent light at 420, 540, and 580 nm, producing valleys in the EEMs parallel to the excitation and emission wavelength axes.

Figure 2

A typical fluorescence EEM from a squamous normal site on the cervix. NADH/CC, FAD/SP, FAD, P represent peak positions of NADH/cross-linked collagen, FAD/structural protein, FAD and porphyrin, respectively. Dotted lines represent positions for hemoglobin absorption. Scale bar for the intensity values is shown on the right.

Figure 3 shows how the fluorescence intensity at 340 nm excitation, 440 nm emission varies in ten patients daily throughout the cycle. On average, the fluorescence intensity varies by approximately a factor of 2 within each patient throughout the menstrual cycle. This is much less than the inter-patient variation observed in Figure 1, which reaches a factor of 10. Furthermore, visual examination of the raw data in Figure 3 does not show any obvious correlation between fluorescence intensity and cycle date within a single patient, although the expected level of interpatient variation was present.

Figure 3

Fluorescence intensity from squamous epithelial cells of the cervix at 340 nm excitation, 440 nm emission daily throughout the cycle in all ten patients. The plots are the mean fluorescence intensity of the three sites measured each day for each patient, and the error bars represent the maximum and the minimum values of the three sites. Blue, green, and red dots represent menstrual, proliferative, and secretory phases of the cycle, respectively. Note that data are plotted against the day of the cycle, which is not necessarily the same as the measurement day.

We explored data reduction strategies, which break the data down into linearly independent factors, and then examined whether variation in any of these individual factors correlated with cycle date. Data reduction, using PCA, was carried out in two excitation wavelength regions: 340–380 and 410–470 nm excitation. In the area at 340–380 nm excitation the strongest correlations were observed. In particular, the scores from the second principal component were visually correlated with the day of the cycle for many of the patients. In one patient (patient 9), this correlation was especially strong, indicating that the corresponding principal component described a portion of the EEM that was most strongly correlated with variations in the cycle. For this patient, the second principal component accounted for approximately 16 of the total variance. Using this second principal component, we calculated the corresponding principal component score at each cycle day for each patient. We then explored visual correlation with the day of the cycle. Results are shown in Figure 4. To facilitate the visualization of data, their sequence has been reordered. Data from all the patients are presented from the first day of the menstrual phase to the last day of the measurement. Then, the data from the first day of the measurement to the day before the first day of the menstrual phase is concatenated at the end.

Figure 4

Principal component scores daily throughout the cycle in all ten patients. The plots are the mean of the three sites at each day, and the error bars indicate the minimum and maximum values among the three sites. Blue, green, and red dots represent menstrual, proliferative, and secretory phases, respectively. Dotted vertical line indicates the day of ovulation (day 14 of the menstrual cycle). Note that data are plotted against the day of the cycle, not necessarily the same as the measurement day.

In order to investigate inherent patterns in the plots, regression analysis was performed using a third order linear fit to the average plot. The linear fit is shown as dark gray curves in each plot of Figure 4. The R² value, which measures the goodness of each fit, is shown in Table 1. In general, the second principal component score is lowest during the menstrual phase of the cycle, increases throughout the proliferative phase of the cycle, and plateaus during the secretory phase.

Table 1

Figure 5 shows the second principal component of patient 9 multiplied by the principal component score at each day of the cycle. For simplicity, the figure shows the principal component corresponding to the emission spectrum at 360 nm excitation. The mean emission spectrum at 360 nm excitation, which is scaled by a factor of 10, is also shown. Note that there are two opposing peaks in the principal component: one peak is in the emission wavelength region from 450 to 480 nm and the other is in the emission wavelength region above 500 nm. The first wavelength region is consistent with the fluorescence peak of the cofactor NADH at 360 nm excitation and 450 nm emission, while the second is consistent with the peak of the cofactor FAD at 360 nm excitation and 525 nm emission.

Figure 5

The second principal component weighted by the principal component scores of patient 9 at each day in the cycle. The principal component corresponding to emission spectrum at 360 nm excitation is shown. Blue, green, and red plots represent menstrual, proliferative, and secretory phases, respectively. Mean emission spectrum at 360 nm excitation, scaled by 10, is also shown in black.

Figure 6

Redox ratio daily throughout the cycle in all ten patients. The plots are the mean of the three sites at each day, and the error bars indicate the minimum and maximum values among the three sites. Blue, green, and red dots represent menstrual, proliferative, and secretory phases, respectively. Dotted vertical line indicates the day of ovulation (day 14 of the menstrual cycle). Note that data are plotted against the day of the cycle, not necessarily the same as the measurement day.

The significance of the fluorescence from the cofactors NADH and FAD in this second principal component suggests that a simpler ratio of fluorescence intensities, known as the redox ratio, may correlate with the day of the cycle. The redox ratio, R, is the ratio of fluorescence intensity at 370 nm excitation, 530 nm emission (FAD) to the sum of intensities at 370 nm excitation, 530 nm emission (NADH) and 370 nm excitation, 450 nm emission (FAD)²¹

Ramanujam²⁰ reports a sensitivity and specificity of 88 and 68, respectively, using a diagnostic classification algorithm to discriminate SIL from squamous normal cervical tissues. Another measure of algorithm performance is the area under the receiver operator characteristic (ROC) curve.²² Using data from Ref. 20, Utzinger calculated an area under the curve of 0.71 for this algorithm.²³ When the algorithm developed in the previous study was applied to the data from this study and the prior probability of disease was adjusted to be 10.2, consistent with the data from this study, specificity was 94.3; this is higher than reported by Ramanujam. There are insufficient data points from biopsy proven SIL to calculate a reliable sensitivity from this study. However, if we assume that all measurements from sites with biopsy proven SIL at the end of the study contained SIL throughout the entire study, we could calculate the area under the ROC curve. We found an area of 0.64, which is comparable to that found previously.²³

4.

Comment

The results of the PCA (Figure 4) and the analysis of the redox ratio (Figure 6) imply that the metabolic activities of the epithelial cells of the cervix are relatively high during the menstrual and proliferative phases of the cycle and that they are relatively low during the secretory phase.²⁴ These results are consistent with other findings. Ferenczy²⁵ states that in general estradiol 17B stimulates epithelial proliferation and maturation, whereas progesterone prohibits upper midzone maturation in the squamous epithelial cells of the cervix. He also states that estrogen receptors show a relative increase during the follicular phase compared to the luteal phase, suggesting that cell proliferation increases during the follicular phase and decreases during the luteal phase. Gorodeski^{26 27} investigated the effect of sex steroids on the formation of ectocervical epithelial envelopes, which are a layer of cross-linked protein that forms beneath the plasma membrane during ectocervical cell terminal differentiation. He states that estrogen increases ectocervical epithelial cell differentiation, whereas progestin decreases envelope formation in the ectocervical epithelial cells.

Fluorescence spectroscopy for the screening and diagnosis of cervical precancer has already demonstrated promising results, with sensitivities and specificities in the 75–90 range. It should be noted that variations in fluorescence intensity throughout the cycle do not account for the tenfold variation in fluorescence intensity observed between patients. Within a single patient, fluorescence intensity varies by approximately a factor of 2 throughout the cycle. Only 8–16 of this variance appears to be correlated with day in cycle. When the data from the present study are applied to a diagnostic classification algorithm, specificity is well above that reported in literature. Thus, in developing algorithms for cervical precancer based on fluorescence spectroscopy, it does not appear necessary to control for cycle date. Patients can be measured at any point in their cycle and the same diagnostic algorithms can be used.

Real-time diagnosis of cervical cancer using fluorescence spectroscopy would provide a great advance in the day-to-day practice of medicine. For developed countries, optical technologies will translate into more cost-effective care by reducing the number of visits and unnecessary biopsies. In developing countries, optical technologies could enable screening programs to be implemented by less trained health care providers.