Injury to the vagina and its adjacent supportive tissues at the time of vaginal childbirth is one of the greatest risk factors for the development of long-term gynecologic complications, including dyspareunia, prolapse, and incontinence.1–8 Although it is generally accepted that the vagina and its supportive connective tissues undergo pronounced adaptations throughout pregnancy in preparation for the passage of the fetus, these adaptations are poorly understood. Because nothing is known of the requisite tissue changes to achieve vaginal delivery, little has been done to develop strategies to prevent maternal birth injury with the exception of recommending primary elective cesarean deliveries.9,10
One of the primary limitations to studying the behavior of the vagina and its supportive tissues in pregnancy and delivery is access to tissue. Although adaptations of maternal tissues could be determined longitudinally by the analysis of serial biopsies obtained before pregnancy, during pregnancy, and in the early and late postpartum period, obtaining such tissue poses an ethical dilemma. In addition, small biopsies provide local data on a specific tissue, but reveal little insight into the mechanism by which a group of tissues functions collectively (ie, the mechanism by which the vagina and its adjacent supportive tissues act interdependently in response to downward distension). Alternatively, animal models provide a well-controlled system in which such ethically challenging mechanistic studies can be addressed in a safe, rigorous manner. We have previously described a testing protocol in the rat in which the biomechanical properties of the vagina and its supportive tissues are tested as a complex, maintaining insertion or attachment points intact.11 By this test, the biomechanical behavior of this tissue complex is defined in response to a specific loading condition (ie, downward distension along the longitudinal axis). Importantly, the test is not meant to recapitulate the process of parturition or a mechanism of prolapse.
It is our general hypothesis that the vagina and its supportive tissues adapt during pregnancy until the time of delivery and if these adaptations are incomplete (inadequate tissue preparation) or exceeded, then maternal injury ensues. In this study, we sought to define specifically the biomechanical adaptations that must take place to afford downward distension of the vagina and supportive tissues at the time of delivery. As such, we employed our previously established testing protocol to define the biomechanical behavior of the rat vagina–supportive tissue complex in pregnancy (mid and late) and at the time of delivery.11 To control for the effects of vaginal delivery, we compared abdominally delivered animals to those who delivered vaginally. Finally, we examined the capacity of the tissues to recover by testing at 4 weeks postpartum. It is our intention to use these studies as a basis for performing future targeted studies in humans. Our goal is to improve our currently limited understanding of maternal tissue adaptations requisite for delivery and eventually the governing mechanisms so as to ultimately provide clinicians with alternative strategies to prevent maternal birth injury.
MATERIALS AND METHODS
A total of 74 female 3-month-old (12 virgin, 62 primigravida) Long Evans rats (Harlan Laboratories, Indianapolis, IN) were used for this study. The weight and number of fetuses (if applicable) were recorded for all rats before killing them. The surgery (abdominal delivery) and killing were performed in accordance with Institutional Animal Care and Use Committee guidelines. Thirteen pregnant animals were killed at mid gestation (day 13), 10 animals in late gestation (day 19–21), eight animals at 0 to 4 hours after vaginal delivery (immediate postvaginal delivery), eight animals 4 weeks after vaginal delivery (late postvaginal delivery), 10 animals 4 hours after abdominal delivery (immediate postabdominal delivery), and 13 animals 4 weeks after abdominal delivery (late postabdominal delivery). Because many of the rats delivered at various times throughout the night, they were checked at 4-hour intervals over a period of 48 hours. Therefore, it is possible that some of the rats were as old as 4 hours postpartum at the time of killing. For this reason, the rats were not immediately killed at the time of cesarean delivery but were allowed to recover for 4 hours so as to equally match them to the vaginal delivery group. We interpreted the 0 to 4 hour postpartum groups as reflecting events that occurred at the time of delivery. The University of Pittsburgh School of Medicine Institutional Animal Care and Use Committees (Institutional Animal Care and Use Committee # 03–003) approved this study.
For rats undergoing abdominal delivery, anesthesia was induced with isoflurane (Abbott Laboratories, North Chicago, IL) at 4 L/min and maintained at 2 L/min. Adequate anesthesia was insured by regular assessments of pain response to a paw pinch. The fur on the lower abdomen was clipped and the rat was prepared with Betadine (povidone-iodine, Purdue Pharma, L.P., Stamford, CT) and draped with a sterile towel. A vertical 2.0-cm skin incision was made with a scalpel, and the abdomen was entered in layers. A uterine horn was identified and gently delivered through the incision, followed by the second horn. A vertical incision was made in the uterus at the union of the two horns and the pup–placental units were delivered. Once all pups and placental tissues were removed, the uterine incision was closed with a 5–0-polydiaxonone suture in a running fashion. The peritoneum and muscle were approximated with 2–0 polydioxanone sutures in a running fashion and the skin was closed with skin clips. Intramuscular butorphanol tartrate (10% vol/vol; Fort Dodge Animal Health, Fort Dodge, IA) was administered (0.5 μL/g body weight) for postoperative pain control. The rats were returned to their individual cages and observed for signs of pain, infection, return to movement, and oral intake. There were no perioperative deaths or infections.
In pregnancy and at delivery, the vagina and its supportive tissues interact to accomplish delivery of the fetus. Thus, for this study, all rats were tested according to a protocol previously developed by us in which the vagina and its supportive tissues are tested as a complex (structural properties), with all of the insertion and attachment points of the complex left intact.11 We have shown in this protocol that the ultimate load corresponds to disruption of level II support (lateral attachments of the vagina to the pelvic sidewall) followed rapidly by loss of level I support (attachments between cervix and upper vagina and spine) and finally level III support (attachments between distal vagina and ischiopubic rami and pubocaudalis). It is important to point out that our ability to perform a controlled multidirectional test in this system is limited, because the pelvis encases the tissues of interest. Technically, because the test is not multidirectional, it does not reproduce the process of parturition or a mechanism of prolapse. In this way, the goal of the test is to quantitate the biomechanical behavior of the vagina–supportive tissue complex in response to one specific loading conditioning, ie, downward distension of the vagina–supportive tissue complex along its longitudinal axis. The data obtained from this test is most relevant to understanding the function of these tissues in response to loads that might be generated by maternal pushing and uterine contractions.
After killing, the rats were dissected down to the level of the pelvis to expose the vagina and supportive tissues. The hind limbs were disarticulated at the acetabulum and the spine was disarticulated above the L1 vertebra as described in Moalli et al11 in 2005. The resulting specimen consisted of the uterus transected horizontally at the level of the cervix, with vagina and all of its supportive tissue attachments intact. The supportive tissues included attachments between the cervix and upper vagina and caudal spine, the paravaginal attachments to the pelvic sidewall, the distal attachments to the ischiopubic rami, and posterior attachments to the pubocaudalis muscle. In addition, the pelvic bones and musculature were left intact. It is noteworthy that this model does not include the suspensory ligaments of the uterine horns. After dissection, the specimens were wrapped in gauze, sealed in plastic bags, and maintained on ice until testing. Before testing, the lumbar spine and the adjacent bony pelvis were aligned on a metal post secured by suture and potted in polymethylmethacrylate. The vagina and supportive tissues were maintained at an appropriate distance from the polymethylmethacrylate and continuously moistened with 0.9% normal saline to prevent damage from the exothermic reaction of the polymethylmethacrylate during hardening and dehydration. This specimen was then mounted in a custom-made cylindrical clamp. A customized soft-tissue clamp was used to secure the distal 5 mm of the vagina. The cylindrical clamp was fixed to the base of a material testing machine (SmartTest EMS, Enduratec, Minnetonka, MN; displacement resolution 0.025 mm), whereas the soft tissue clamp was fixed to a load-cell (SM-1000N, Interface, Scottsdale, AZ; resolution 0.015 N) attached to the crosshead of the machine.
The uniaxial load to failure test simulates downward distension of the vagina–supportive tissue complex by pulling the distal vagina in the direction of its longitudinal axis. Thus, the measured load (force) is that which results as the vagina and supportive tissues collectively accommodate and then resist distension as they attempt to maintain normal anatomic relationships. The resulting load–distension curve describes the overall performance of the vagina–supportive tissue complex as it accommodates and then resists downward distension. Four characteristics are derived from the curve: linear stiffness (N/mm), ultimate load at failure (N), maximal distension (mm), and energy absorbed to failure (N-mm). A representative curve is shown in Figure 1.
Before each test, the specimens were preloaded to 0.15 N, and preconditioned at 25 mm/min between 0 and 2 mm of distension for 10 cycles. This corresponded to cycling within the toe region (ie, initial low stiffness region) of the load–distension curve. A uniaxial (load applied along the longitudinal axis of the vagina) load to failure test was performed immediately after preconditioning at the same distension rate. A computer, interfaced to the load cell, collected corresponding data points for load and distension every 0.02 seconds using software provided by Enduratec. Failure modes were determined after each test based on visual inspection of the sample. By this test, failure occurred within the vagina–supportive tissue complex and not at the tissue clamps.
Based on preliminary data (virgin compared with midpregnant data), eight rats were required in each group to achieve 80% power to detect a 38% difference in distension at the 0.05 significance level. Eight rats per stage provided greater than 80% power to detect a 27% and 50% difference in stiffness and ultimate load, respectively. Variance estimates were calculated based on preliminary data and using nQuery Advisor 4.0 (Statistical Solutions, Saugus, MA). For each sample, corresponding load and distension data were imported into Excel for analysis (Excel, Microsoft Corp, Redmond, WA). The resulting load–distension curve was used to estimate the biomechanical properties of the vagina–supportive tissue complex. The properties included linear stiffness (N/mm), ultimate load at failure (N), maximal distension (mm), and energy absorbed to failure (N-mm). All load–distension curves generated in this study were shaped similarly with defined toe, linear, and failure regions. Linear stiffness was defined as the steepest positive slope measured over a 1 mm interval of distension for each specimen. It is a measure of the specimen’s ability to resist distension, ie, maintain normal anatomic relationships. Ultimate load and maximal distension define the point of failure on the load–distension curve, which in our study was observed to correspond with disruption of level II support followed by failure of levels I and III. Thus, ultimate load at failure is a measure of the maximal sustainable force of the vagina–supportive tissue complex or maximal resistance to distension, and was defined as the highest load on the load–distension curve. Maximal distension was the distension that corresponded with the ultimate load and describes the distance the vagina could be pulled before tissue disruption. Energy absorbed to failure was defined as the area under the load–distension curve up to failure; this measured the work or effort the specimen exhibited to resist distension. The testing protocol was previously established for testing the rat vagina and its supportive tissues.11 Statistical analyses were performed using SPSS 13.0.1 statistical software (SPSS Inc., Chicago, IL). Normality was assessed using graphic displays of the data, as well as skewness and kurtosis statistics. Since there was not a significant departure from normality, structural properties were analyzed using one-way analysis of variance. Post hoc pair-wise comparisons (all groups compared with the virgin animals) were made using Dunnett’s multiple comparison t tests.
No gross damage to the vagina and supportive tissues was seen at the time of delivery or at dissection in any specimen. The biomechanical properties of the vaginal supportive tissue complex obtained for the virgin 3-month-old rats were similar to previous data reported (Moalli et al11 in 2005). The weights of the rats and number of pups at the time of delivery in each group are shown in Table 1. In all of the rats, the ultimate load at failure corresponded to disruption of level II support (lateral attachments of the vagina to the pelvic sidewall), followed rapidly by loss of level I support (attachments between cervix and upper vagina and spine) and, finally, at level III support (attachments between distal vagina and ischiopubic rami and pubocaudalis) similar to our previous study. During the testing, there was no grip slippage at the clamps and the vaginal wall grossly remained intact.
In comparison with virgin rats, the pregnant rats displayed a decrease in linear stiffness (P<.001) and a decrease in ultimate load at failure (P<.001) but no difference in maximal distension or energy absorbed (P >.999). Comparison of mid and late pregnancy values demonstrated no significant differences in linear stiffness, ultimate load at failure, maximal distension, or energy absorbed to failure (Table 2).
Relative to virgins, the biomechanical properties of the vagina–supportive tissue complex of rats killed at the time of vaginal delivery had a lower linear stiffness (P<.001) and a decrease in ultimate load at failure (P<.001). In addition, the vaginally delivered specimens distended more than virgin, midpregnant, and late pregnant specimens before failure (P<.001). Finally, the vaginally delivered group had a higher energy absorbed to failure compared with the virgin, midpregnant, and late pregnant groups (P=.01).
The animals killed at the time of abdominal delivery displayed similar results to the vaginal delivery group, with a lower linear stiffness (P<.001) and a lower ultimate load at failure (P<.001) compared with virgin animals. Also similar to the vaginal delivery group, these specimens distended more than virgins, midpregnant, and late pregnant rats (P<.001). Abdominally delivered specimens had a higher energy absorbed compared with virgins (P=.05). No statistically significant differences in the biomechanical properties were detected between specimens killed at the time of vaginal delivery and at the time of abdominal delivery, indicating that route of delivery did not affect any biomechanical changes.
Four weeks postvaginal delivery and 4 weeks postabdominal delivery, specimens had recovered to a higher linear stiffness (P<.001) and ultimate load at failure (P<.001) compared with pregnant rats (middle and late gestation) and rats killed at the time of vaginal delivery and abdominal delivery. Interestingly, at 4 weeks, the ultimate load at failure and the energy absorbed until failure in the abdominal delivery rats was higher than the 4-week vaginal delivery rats and virgin rats (Table 2). However, no significant differences were demonstrated in the other biomechanical properties (linear stiffness, maximal distension, and energy absorbed to failure) between these two groups.
The goal of this study was to improve our understanding of the adaptations or changes in tissue behavior that allow the vagina–supportive tissue complex to accommodate labor and the passage of a fetus. The most important finding was that the biomechanical behavior of the vagina–supportive tissue complex is indeed highly dynamic, undergoing profound changes in pregnancy and at the time of delivery, most likely as an adaptation that affords tissue distension in the process of parturition. We found that pregnancy decreases the linear stiffness and ultimate load at failure of the vagina–supportive tissue complex as early as mid gestation, persisting through late gestation and delivery. Route of delivery, abdominal or vaginal, did not further influence these characteristics. However, the maximal distension, a measure that quantifies the distance the vagina–supportive tissue complex was pulled before failure, was significantly increased at the time of both abdominal and vaginal delivery. Finally, we found that, in contrast to a vaginal delivery, the long-term recovery of the tissue after abdominal delivery resulted in specimens with a higher ultimate load at failure.
Linear stiffness is defined as the slope of the load-distension curve that describes the ability of the specimen to resist distension. As described, linear stiffness is an index of distensibility. A decrease in stiffness translates into increased distensibility. The decrease in linear stiffness during pregnancy and at the time of delivery makes intuitive sense, because it is likely part of an adaptive process that allows the vagina and supportive tissues to distend downward during labor without tissue disruption. This is further supported by the increase in maximal distension for the pregnant and immediate postdelivery rats. Thus, these data indicate that the tissues of the vagina–supportive tissue complex become more distensible in preparation for delivery and can distend to a much greater degree at the time of delivery. These measures, in theory, would simultaneously facilitate the passage of the fetus while protecting the mother from birth injury. The absence of a difference in these characteristics in rats delivered abdominally and vaginally confirm that the changes affording increased distensibility of the vagina–supportive tissue complex represent a true adaptation and were not induced by the mechanical pressures associated with delivery.
On the other hand, the adaptive changes that allow increased distention seem to develop at the expense of the ultimate load at failure of the vagina–supportive tissue complex, because it decreases in conjunction with the decrease in stiffness and increase in maximal distension. The ultimate load is defined as the maximal load that was sustained by the vagina–supportive tissue complex before failure. The resulting more distensible but weaker tissues therefore may be at an increased risk for injury during vaginal delivery should adaptive mechanisms be exceeded. The complete recovery of all biomechanical characteristics to prepregnancy values 4 weeks after abdominal delivery indicates that the rat vagina–supportive tissue complex has effectively evolved to a point at which it can accommodate multiple deliveries over a relatively short period with minimal deleterious effects. It is unlikely that human maternal tissues would demonstrate a similar recovery. We believe that injury to the vagina–supportive tissue complex at the time of delivery results in tissue with long-term inferior biomechanical properties and may provide the basis for the development of long-term problems associated with maternal birth injury, such as dyspareunia, incontinence, and prolapse. The changes observed in biomechanical properties of the vagina–supportive tissue complex 4 weeks after abdominal delivery need further evaluation. The increased ultimate load at failure and energy absorbed could be explained by scar tissue development as sequelae of surgery. However, the decrease in linear stiffness and increase in maximal distension are the opposite of what we would expect to occur with scar tissue. Future animal studies aimed at inducing injury by exceeding adaptations are needed to provide further insight into mechanisms of injury.
In this study, we performed all abdominal deliveries before the onset of labor. This raises the question of whether we would have seen the superior biomechanical properties at 4 weeks had the abdominal delivery been achieved during labor instead of before its onset. The increased ultimate load at failure is potentially due to surgery-induced scarring that may lead to adhesion formation in the pelvis between the vagina and uterus and surrounding structures. If scarring results in the superior biomechanical properties at 4 weeks, this outcome may be obtained whether the abdominal delivery is performed before or after the onset of labor. We are currently investigating the mechanism of the apparent conferring of superior biomechanical properties of the rats that deliver abdominally.
The precise mechanism mediating tissue adaptations in pregnancy and at the time of delivery are not known; however, it is likely that pregnancy-related hormones play an important role. For example, passive immunization with an antirelaxin antibody to neutralize endogenous circulating relaxin at days 12–22 of gestation caused a disruption of labor and birth in rats.12,13 Although the cervices from antibody treated rats were much smaller and less extensible than those obtained from control rats, the vagina was also profoundly affected. Indeed, the vaginas of antibody treated animals did not undergo the normal increases in vaginal wet weight, length, diameter, and inner surface area observed in the pregnant controls.13–15 In humans, nulliparous pregnant women have increased total vaginal length and increased laxity in the anterior and posterior compartments relative to nonpregnant nulliparous women.16 It is likely that this laxity as measured by the Pelvic Organ Prolapse Quantification17 examination represents adaptations that facilitate distension at the time of delivery. We predict that women with insufficient adaptations are more likely to incur injury at the time of delivery. This concept is supported by a study from Dietz et al18 demonstrating that women with the least amount of vaginal laxity in pregnancy were the most likely to be injured by vaginal delivery. Thus, future studies interrogating the roles of individual pregnancy hormones in mediating pregnancy adaptations as well as the effect of subtle differences in the expression of these hormones in humans and rats may provide insight into the mechanism by which adaptations of the vagina–supportive tissue complex are achieved.
Our model is very sensitive in detecting subtle changes in the biomechanical properties or the compensatory adaptations that occur in the vagina and supportive tissues during pregnancy, delivery, and recovery. All members of this complex (ie, vagina, paravaginal attachments to the pelvic sidewall, uterosacral ligaments, perineal membrane, etc) interact to accomplish delivery of a fetus, and in the nonpregnant state these structures act in concert to resist prolapse of the pelvic organs and inversion of the vagina. Nevertheless, studies aimed at specifically defining the mechanical properties of these individual tissues will eventually be necessary to correlate biomechanical adaptations directly with changes in tissue morphology and composition as well as to build models to investigate tissue responses to loading conditions applied during parturition. Ultimately, we will perform such studies to define the contributions of the individual components of the complex to the behavior of the entire structure.
In conclusion, pregnancy and delivery, either vaginal or abdominal, resulted in changes to the biomechanical properties of the vagina and supportive tissue complex. The increase in the maximal distension and decrease in the linear stiffness may be adaptive changes to afford downward distension of the tissues without disruption. However, the decrease in the ultimate load at failure may predispose the tissue complex to injury as the mechanical demand of pregnancy and delivery exceeds adaptive changes. Further studies are indicated to determine whether the recovery and compensatory changes that occur during and after pregnancy and delivery in the rodent can be overwhelmed and lead to permanent changes to the structural properties of the vagina–supportive tissue complex. Lin et al19 have successfully created a birth injury model for stress urinary incontinence in rodents that has shown permanent neurologic and histologic changes in the urethra as a result of the simulated birth trauma. Creation of a model that exceeds adaptations and induces long-term injury to the vagina–supportive tissue complex may shed light on the potential effect of birth injuries on this tissue complex, as it has for stress urinary incontinence. With more than 3 million vaginal deliveries in the United States annually,20 new insights into the factors that regulate maternal tissue adaptations and events that lead to maternal injury may provide clinicians with objective criteria on which to base decisions regarding the institution of preventive measures to avoid birth injury, to determine the mode of delivery in the event that adaptations are likely to be exceeded, and to initiate treatment after injury, all of which would improve the lives of thousands of women each year.
1. DeLancey JO, Kearney R, Chou Q, Speights S, Binno S. The appearance of levator ani muscle abnormalities in magnetic resonance images after vaginal delivery. Obstet Gynecol 2003;101:46–53.
2. Allen RE, Hosker GL, Smith AR, Warrell DW. Pelvic floor damage and childbirth: a neurophysiological study. Br J Obstet Gynecol 1990;97:770–9.
3. Dietz HP, Bennett MJ. The effect of childbirth on pelvic organ mobility. Obstet Gynecol 2003;102:223–8.
4. Signorello LB, Harlow BL, Chekos AK, Repke JT. Postpartum sexual functioning and its relationship to perineal trauma: a retrospective cohort study of primiparous women. Am J Obstet Gynecol 2001;184:881–8.
5. Albers L, Garcia J, Renfrew M, McCandlish R, Elbourne D. Distribution of genital tract trauma in childbirth and related postnatal pain. Birth 1999;26:11–7.
6. Sultan AH, Kamm MA, Hudson CN, Thomas JM, Bartram CI. Anal-sphincter disruption during vaginal delivery. N Engl J Med 1993;329:1905–11.
7. Peyrat L, Haillot O, Bruyere F, Boutin JM, Bertrand P, Lanson Y. Prevalence and risk factors of urinary incontinence in young and middle-aged women. BJU Int 2002;89:61–6.
8. Goldberg RP, Abramov Y, Botros S, Miller JJ, Gandhi S, Nickolov A, et al. Delivery mode is a major environmental determinant of stress urinary incontinence: results of the Evanston-Northwestern Twin Sisters Study. Am J Obstet Gynecol 2005;193:2149–53.
9. Feldman GB, Freiman JA. Prophylactic cesarean at term? N Engl J Med 1985;312:1264–7.
10. Nygaard I, Cruikshank DP. Should all women be offered elective cesarean delivery? Obstet Gynecol 2003;102:217–9.
11. Moalli PA, Howden NS, Lowder JL, Navarro J, Debes KM, Abramowitch SD, et al. A rat model to study the structural properties of the vagina and its supportive tissues. Am J Obstet Gynecol 2005;192:80-8.
12. Guico-Lamm ML, Sherwood OD. Monoclonal antibodies specific for rat relaxin. II. Passive immunization with monoclonal antibodies throughout the second half of pregnancy disrupts birth in intact rats. Endocrinology 1988;123:2479–85.
13. Hwang JJ, Sherwood OD. Monoclonal antibodies specific for rat relaxin. III. Passive immunization with monoclonal antibodies throughout the second half of pregnancy reduces cervical growth and extensibility in intact rats. Endocrinology 1988;123:2486–90.
14. Hwang JJ, Shanks RD, Sherwood OD. Monoclonal antibodies specific for rat relaxin. IV. Passive immunization with monoclonal antibodies during the ante partum period reduces cervical growth and extensibility, disrupts birth, and reduces up survival in intact rats. Endocrinology 1989;125:260–6.
15. Zhao S, Sherwood OD. Monoclonal antibodies specific for rat relaxin. X. Endogenous relaxin induces changes in the histological characteristics of the rat vagina during the second half of pregnancy. Endocrinology 1998;139:4726–34.
16. O’Boyle AL, Woodman PJ, O’Boyle JD, Davis GD, Swift SE. Pelvic organ support in nulliparous pregnant and nonpregnant women: a case control study. Am J Obstet Gynecol 2002;187:99–102.
17. Bump RC, Mattiasson A, Bo K, Brubaker LP, DeLancey JO, Klarskoy P, et al. The standardization of terminology of female pelvic organ prolapse and pelvic floor dysfunction. Am J Obstet Gynecol 1996;175:10–7.
18. Dietz HP, Steensma AB. Which women are most affected by delivery-related changes in pelvic organ mobility? Eur J Obstet Gynecol Reprod Biol 2003;11:15–8.
19. Lin AS, Carrier S, Morgan DM, Lue TF. Effect of simulated birth trauma on the urinary continence mechanism in the rat. Urology 1998;52:143–51.
20. Ventura SJ, Martin JA, Curtin SC, Menacker F, Hamilton BE. Births: final data for 1999. National vital statistics reports; vol 49 no. 1. Hyattsville (MD): National Center for Health Statistics; 2001. Available at: http://www.cdc.gov/nchs/data/nvsr/nvsr49/nvsr49_01.pdf
. Retrieved October 25, 2006.
Figure. No caption available.