Materials and Methods

We retrospectively identified 40 gynecologic-oncology patients who had pre- and posttreatment CT scans in our database from 2005 to 2009. This study was approved by the University of Minnesota Institutional Review Board. Subjects with a baseline CT scan performed prior to treatment and additional posttreatment scans performed at the University of Minnesota Medical Center were included for this study. This was approximately 1 year following up from the start of the treatment. The time of first CT scans and second CT scans were not controlled. They were measured within 2-3 months. Details are summarized in Table 1 .

Ovarian cancer patients who had an initial CT scan at the time of diagnosis, but before any surgery or chemotherapy, and at least 2 additional scans after their initial course of platinum- and taxane-based chemotherapy were included in the study. The primary regimen for 11 of the women was carboplatin (area under the curve = 6) with paclitaxel, 175 mg/m ² intravenously (IV) over 3 hours, given every 21 days for at least 6 cycles.

Because of toxicity, 1 patient received her last 4 treatments with single-agent carboplatin, and another patient received her final 6 cycles (8 total) with carboplatin and docetaxel. One patient received 6 cycles of carboplatin and paclitaxel in the neoadjuvant setting and then received 3 cycles of IV cisplatin (75 mg/m ² ) and intraperitoneal cisplatin and paclitaxel. One ovarian cancer patient was taking alendronate prior to her diagnosis and throughout the series of CT scans. Five women experienced a recurrence of their disease and may have been on salvage chemotherapy at the time of the 12 month CT scan (or last follow-up) in this study.

The majority of endometrial cancer patients received external beam pelvic radiation therapy (EBRT) with a median dose of 45-50 cGy (in 25-28 fractions) and 5-18 Gy additional radiation dose at the vaginal surface using high dose rate (HDR) brachytherapy. Cervical cancer patients received radiation sensitizing chemotherapy with 40 mg/m ² of cisplatin, given once weekly during their external beam radiation. They received extended field EBRT with a median dose of 45-50 cGy (in 25-28 fractions) and 1-2 implants low dose rate (LDR) brachytherapy (20-40 Gy).

The EBRT field and radiation doses were individualized according to tumor size, lymph node involvement, and the patient performance status. Figure 1 is a schematic representation of the EBRT field in anterior-posterior direction used for endometrial and cervical cancer treatment. In general, for endometrial cancer patients (without lymph node metastasis, stages I and II), the upper border extended up to lumbosacral joint (S1-L5) as shown in green boundary. For most cervical cancer patients, the superior border was extended up superiorly to L5-L3. The inferior border was at the bottom of the obturator foramen. The lateral borders were approximately 1.5-2 cm lateral to the inner bony margins of the true pelvis. For the lateral fields, the anterior border included the anterior one-third of the symphysis pubis and the posterior border.

Schematic representation of radiation fields for pelvic irradiation

Figure shows schematic representation of radiation fields for pelvic irradiation up to S1-L5 junction ( inside green boundary ) and extended irradiation field up to L2-L3 junction ( inside green and brown boundary ).

Hui. Bone loss in gynecologic cancer treated with chemotherapy and radiation. Am J Obstet Gynecol 2010 .

CT scans for this study were acquired using a standard, clinical CT protocol without a CT calibration standard to facilitate QCT analysis. Instead, images were calibrated by scanning a CT calibration phantom and QA phantom (Mindways Software, Inc, Austin, TX), and the CT calibration equations derived from the phantom study were assumed to apply to all patient studies. The CT scanner was maintained in accordance with the CT manufacturer’s specifications over the time period of this study.

No anomalies in the CT scanner performance were found in the CT records. Images from picture archiving and communication systems were transferred to a computer for QCT analysis. Images were analyzed using the QCT PRO software (version 4.2; Mindways Software). Calibration equations were derived from the analysis of the QA phantom and CT phantom (Mindways Software) imaged simultaneously. The CT calibration equations were used in turn to estimate vBMD when analyzing the patient data. We analyzed vBMD (milligrams per cubic centimeter) from trabecular regions (L1-L5) of the spine and from the femoral neck (FN) region ( Figure 2 ). L1 and L2 trabecular vBMD was averaged to measure L1-L2 vBMD; L3 and L4 trabecular vBMD was averaged to measure L3-L4 vBMD; L5 vBMD was from L5 trabecular bone only; FN vBMD was calculated from integrated vBMD of trabecular and cortical bone at the FN region of the hip.

QCT measurements

Locations of QCT measurements. Left panel shows trabecular region of L spine in which BMD was measured. Right panel shows femoral neck selection for BMD measurement.

BMD , bone mineral density; QCT , quantitative computed tomography.

Hui. Bone loss in gynecologic cancer treated with chemotherapy and radiation. Am J Obstet Gynecol 2010 .

Statistical Methods

Patients were categorized into 3 treatment groups: chemotherapy (ovarian cancer patients), EBRT plus HDR (endometrial cancer patients), and combination of radiation (EBRT plus LDR) and chemotherapy (cervical cancer patients). We compared vBMD from CT scans performed at baseline (prior to any surgery, chemotherapy, or radiation) and at 2 time points after the completion of treatment. For each subject the BMD level at L1-L2, L3-L4, L5, and the FN was measured on each scan. The percent change in vBMD was computed. Descriptive statistics including mean, SD, and range of the absolute vBMD and percent reduction in vBMD were calculated.

For each scan region, a patient’s percent reduction in vBMD on each of 2 posttreatment scans was evaluated using paired Student t test or Wilcoxon’s signed-rank test. In addition, general linear models for longitudinal data were used to evaluate the effects of age, treatment, scan time, scan region, and their interaction terms on percentage reduction of BMD. The variance-covariance structure was the first-order autoregression, and variance parameters were estimated using restricted maximum likelihood method with Satterthwaite approximation. The final model was determined by a backward selection, and the least squares means for percentage reduction of vBMD were calculated. The P values reported for multiple comparisons were unadjusted. SAS version 9.1 (SAS Institute Inc, Cary, NC) was used. P < .05 indicated statistical significance.