BACKGROUND AND OBJECTIVE: Cyclophosphamide is a teratogenic medication used in the treatment of adolescents with autoimmune disorders. This adolescent population is sexually active, does not receive adequate contraceptive care, and is at risk for unintended pregnancy. We undertook a quality improvement initiative to improve rates of pregnancy screening before intravenous cyclophosphamide administration in our adolescent girl patients.
METHODS: Data were collected from the electronic medical record. The primary outcome was completion of a urine pregnancy test before intravenous cyclophosphamide infusion in girls aged 12 to 21 years between July 2011 and June 2015. Data were reviewed quarterly and an iterative quality improvement approach was used. Interventions included staff education, electronic order set updates, and a Maintenance of Certification project. Interrupted time series analysis and multivariable mixed effects logistic regression were used to evaluate trends over time and to adjust for potential confounders.
RESULTS: Thirty girls received 153 cyclophosphamide infusions during the study. Pregnancy testing before medication administration increased from 25% to 100% by study completion. Infusions in the last time period were significantly more likely to be accompanied by a pregnancy test versus those in the first time period (odds ratio: 17.7; 95% confidence interval [CI]: 3.1–101.6) after adjustment for patient age, managing service, infusion setting, and insurance type.
CONCLUSIONS: Our institution achieved a significant increase in standard pregnancy screening in adolescent girls receiving intravenous cyclophosphamide. The interventions most valuable in increasing screening rates were updating electronic order sets, educating staff, and physician engagement in the Maintenance of Certification program.
- CI —
- confidence interval
- EMR —
- electronic medical record
- FDA —
- US Food and Drug Administration
- HCG —
- human chorionic gonadotropin
- ITS —
- interrupted time series
- MOC —
- Maintenance of Certification
- OR —
- odds ratio
- PHIS —
- Pediatric Hospital Information System
- REMS —
- Risk Evaluation and Mitigation Strategy
High-dose intravenous cyclophosphamide is used in the treatment of autoimmune diseases, such as systemic lupus erythematosus, systemic sclerosis, and primary vasculitis. Although efficacious, cyclophosphamide is teratogenic and has been rated “category D” by the US Food and Drug Administration (FDA) for use in pregnancy. This rating indicates substantial evidence of fetal risk, although use may be considered in situations where no acceptable alternatives exist.1,2 In 2009, the American College of Rheumatology published a set of quality indicators for care of adult patients with systemic lupus erythematosus undergoing cyclophosphamide treatment, which recommend discussion of the teratogenicity risks and need for appropriate contraception in women aged 18 to 45 years before medication initiation.3,4 Although the specific issue of pregnancy testing before cyclophosphamide administration is not addressed, such screening is a logical step to avoid unintentional fetal exposure and precedents exist nationally for other teratogenic medications.5 There are no existing guidelines around teratogenicity avoidance for this medication in pediatrics, yet many adolescents receiving cyclophosphamide are of childbearing age.6
In the United States in 2010, an estimated 40% of girls aged 15 to 18 years were sexually active. Approximately 24 out of 1000 teenaged girls became pregnant in 2014 and up to 80% of teen pregnancies were unplanned.7–9 Adolescents with rheumatologic diseases are reported to be sexually active at rates comparable to peers, and may not use adequate contraception.10,11 Despite this, routine pregnancy screening before use of cyclophosphamide and other teratogenic medications is not yet a care standard in the pediatric poluation.10,12,13 Studies examining teratogenicity discussions and routine pregnancy screening have indicated that these practices have been difficult to implement.14–19
Seattle Children’s Hospital is a 323-bed tertiary pediatric teaching hospital in Seattle, Washington. In April 2012, we created a clinical pathway for cyclophosphamide administration in nonmalignant conditions to ensure safe administration, including hydration, clinical monitoring, standard laboratories, Pneumocystis prophylaxis, and osteoporosis prevention. This pathway is primarily used for rheumatology and nephrology patients. This quality improvement effort was developed under the auspices of our Clinical Effectiveness program, whose structure has been previously described.20–22 Pathway issues are reviewed at least quarterly, including outcomes monitoring. This structure for improvement was in place when the sentinel event of an unintended pregnancy was discovered in a young woman after an outpatient rheumatology clinic visit in August 2012.
After this sentinel event, we sought to improve the rate of pregnancy screening before medication administration among young women ≥12 years receiving teratogenic medications, such as cyclophosphamide. We describe our experience using an iterative quality improvement approach. We also sought to benchmark national rates of pregnancy screening before cyclophosphamide administration. Our hope is to improve awareness of the avoidable risks posed by teratogenic medications in this population and to share effective approaches to improving preexposure pregnancy screening before administering these medications.
At our institution, administration of intravenous cyclophosphamide for autoimmune disorders occurs both in the inpatient setting and in an outpatient infusion center. The most common ordering services are rheumatology and nephrology, including ∼20 ordering physicians and nurse practitioners. Inpatients are managed by a team approach, where trainees, including residents and fellows, may place orders. For outpatient infusions and planned admissions, orders are entered into the electronic medical record (EMR; Cerner Millennium) before infusion by a clinician. This clinician may be different from the physician overseeing patient care on the following day. Because of its narrow therapeutic index and potential for toxicity, cyclophosphamide orders must be reviewed and cosigned by an attending physician.
After the sentinel event, the Cyclophosphamide Pathway team analyzed key drivers (Fig 1) and constructed interventions to prevent unintended fetal exposure to intravenous cyclophosphamide among girl patients of reproductive age receiving this medication for nonmalignant diagnoses.
Intervention 1: Provider Education and REMS protocol
Providers within the rheumatology division developed a division-wide Risk Evaluation and Mitigation Strategy (REMS) protocol for adolescent girl rheumatology patients ages ≥12 years being prescribed high-risk teratogenic medications, including cyclophosphamide. This protocol encompasses patient education, contracting for safe use of medications, referral for contraceptive counseling, and consent/assent for urine pregnancy screening. We developed standard scripts for providers to use when (1) discussing the need for testing with patients and caregivers, as well as (2) reporting positive results to the patients in accordance with state and legal privacy requirements regarding adolescent sexual health information (provider script available in Supplemental Information). Implementation of the program was conducted through a series of staff training events from September 2012 through May 2013.
Intervention 2: Electronic Order Set Update and Maintenance of Certification Improvement Project
When surveillance revealed that pregnancy screening was still inadequate after our first set of interventions, additional interventions were performed. The cyclophosphamide EMR order sets were revised to include orders for urine human chorionic gonadotropin (HCG) testing, with a note to ordering physicians to consider pregnancy testing in eligible patients. A comment in the cyclophosphamide order instructed nursing staff to check pregnancy test results before its administration. These changes were implemented in February 2014. Shortly thereafter, the rheumatologist leading the Cyclophosphamide Pathway team initiated a division-wide Maintenance of Certification (MOC) Part IV quality improvement project with the specific aim of increasing compliance with pregnancy screening in girl patients ≥12 years of age receiving cyclophosphamide to 90% by October 2014. The MOC project involved physician engagement through in-person meetings to review aggregate quarterly data for pregnancy screening rates for patients on the pathway as well as to brainstorm potential solutions to barriers to adherence to the screening recommendation. Provider-specific performance data were not provided to individual physicians nor to the group at large and there were no formal negative consequences for lack of progress detected.
The primary outcome measure was completion of a urine HCG pregnancy test within 1 week before cyclophosphamide infusion. We included 12- to 21-year-old girl patients receiving cyclophosphamide infusions in the inpatient or outpatient setting for a diagnosis other than malignancy from July 1, 2011 through June 30, 2015. We excluded patients on the hematology/oncology, bone marrow transplant, and solid organ transplant services because these patients were not eligible for inclusion in the clinical pathway for nonmalignant use of cyclophosphamide. We performed several additional analyses to assess the efficacy of our screening program: we queried the electronic data warehouse for any patients who had a positive pregnancy test documented and cyclophosphamide administration at any time in the patient’s EMR as well as for patients with a cancelled cyclophosphamide order and a positive pregnancy test documented.
Laboratory and medication administration data from Cerner and hospital administrative data from Epic Hyperspace were accessed via our enterprise data warehouse, and analyzed retrospectively. Core, outcome, and process-related measures were tracked using run charts. Measures for each encounter were categorized into 3 time periods: period 1: pathway launch up to REMS protocol launch (7/1/2011–5/22/2013); period 2: REMS protocol/provider education (5/23/2013–2/24/2014); and period 3: EMR revision and MOC (2/25/2014–6/30/2015).
To provide context for our local experience, we also examined rates of pregnancy screening for adolescent girls receiving cyclophosphamide nationally using the Pediatric Hospital Information System (PHIS) administrative database. PHIS hospitals include 44 children's hospitals in the United States, representing the most demanding standards of pediatric care.23 Participating institutions contribute administrative data including patient demographics, diagnoses, and charges as well as detailed data for billed services, such as pharmacy, clinical services, and laboratory tests for certain institutions. At the time of data submission, records are de-identified and data are subjected to reliability and validity checks before being included in the database. We queried PHIS for inpatient girls who were 12 to 21 years old at time of admission with intravenous cyclophosphamide administration between July 1, 2011 and June 30, 2015, excluding hematology/oncology, bone marrow transplant, and solid organ transplant services. Outpatient cyclophosphamide encounters could not be included due to insufficient data available on associated laboratory tests. The numerator was eligible cyclophosphamide encounters with laboratory tests for HCG documented.
Interrupted time series (ITS) analysis was used to evaluate trends over time in the proportion of infusions that had a preinfusion pregnancy test ordered. Linear regressions were estimated for each time period using the outcome of the proportion of infusions with a test ordered at each quarter. To assess differences between periods, the slope parameters of the regression equations were compared, as well as the difference in the ending estimated value of 1 time period and beginning intercept of another time period. Wald tests were used to make comparisons of the regression-estimated slope and intercept parameters. Multivariable mixed-effects logistic regression was used to estimate the odds of a pregnancy test being ordered before each unique medication infusion for all eligible patients in each time period. Covariates were specified a priori and included patient age, inpatient versus outpatient treatment, ordering provider specialty, and insurance type. Time period and covariates were treated as fixed effects, with a random effect for participants to account for clustering among patients with repeated infusions. The rates of pregnancy testing in the PHIS database were compared across the 3 Seattle study periods using the Pearson’s χ2 test of independence.
All analyses were conducted by using Stata version 12 (Stata Corp, College Station, TX).
In 2008, our institution implemented a robust mandatory pregnancy screening process for adolescent girls before selected radiographic studies and surgical procedures. However, no institutional standards existed with regards to administration of teratogenic medications. For consistency, equity, and autonomy, the pathway team adopted the same standards as our existing institutional policy, to screen all girls ≥12 years regardless of reported sexual activity or onset of menses. This study was approved by the Seattle Children’s institutional review board.
During the study period, our hospital administered cyclophosphamide to 69 unique patients over 237 separate encounters. Within this population, we identified 30 girl patients aged ≥12 years who received 153 cyclophosphamide infusions in the inpatient (67%) and outpatient (33%) settings and were included in our analysis. Mean subject age in this cohort was 14.5 years (range: 12–19 years). Cyclophosphamide infusion orders were placed in the EMR by a group of 30 unique providers during this time.
Patient characteristics are reported in aggregate for the study in Table 1. Certain patient and encounter characteristics chosen as logistic regression covariates are shown by time period (Table 2). Patients could be included in multiple time periods or have multiple encounters within a single time period given the recurring nature of cyclophosphamide infusions. Overall, patient and encounter characteristics remained stable over the 3 time periods.
The primary outcome of pregnancy testing before infusion increased from 25% in the baseline time period to 100% at study completion. One positive pregnancy test occurred in this population within the overall study time period. Chart review revealed that the positive result was detected in the outpatient setting 4 months after the patient’s last cyclophosphamide infusion. A negative pregnancy test had been recorded immediately before the last cyclophosphamide dose.
Results of the interrupted time series analysis are provided in Fig 2. By this analysis, the difference between period 2 and period 1 did not meet statistical significance for trend over time (change in slope = –2.6; 95% confidence interval [CI]: –16.9 to 11.7) or percentage level (change in intercept = 21.5; 95% CI: –2.2 to 45.2). Comparing period 3 to the baseline period 1 without the intervening period did show a statistically significant change in both level (change in intercept = 34.2; 95% CI: 15.4–53.0) and trend over time (change in slope = 9.5; 95% CI, 3.5–15.5).
When comparing the odds of receiving a pregnancy test at each infusion, infusions in the last time period were significantly more likely to be accompanied by a pregnancy test than those in the first time period (odds ratio [OR]: 17.7; 95% CI: 3.1–101.6), after adjusting for patient age, managing service, infusion setting, and insurance type. There was a trend toward increased odds of receiving a pregnancy test in period 2 compared with period 1(OR: 4.1; 95% CI: 0.9–18.9) and in period 3 compared with period 2 (OR: 4.3; 95% CI: 0.7–25.7).
As a rough national comparison, the PHIS data showed a rate of pregnancy testing associated with an inpatient cyclophosphamide infusion of 68.1% (912 of 1340 encounters; 784 unique patients) for the entire study period. This rate did not demonstrate statistically significant change over time when compared across the Seattle study time periods (P = .262).
Our institution achieved a clinically and statistically significant increase in standard pregnancy screening in adolescent girls receiving intravenous cyclophosphamide. This change impacted a small total patient population in our institution but demonstrated reproducibility over a larger number of at-risk events in a short time frame after implementing changes. In particular, we were curious if our intervention prevented the administration of any doses of cyclophosphamide based on the finding of a positive pregnancy test. Fortunately, no positive pregnancy tests were identified within our cohort of patients during the course of their cyclophosphamide regimen. However, we did have one patient become pregnant shortly after completing cyclophosphamide infusions, thus underscoring the importance of implementing this reliable screening regimen.
The concept of preventing teratogenic exposures using REMS protocols in certain populations is well established. Previous experience with FDA-mandated REMS programs comes from pharmacologic agents, such as isotretinoin and more recently mycophenolate mofetil in 2012.24–26 There has been debate about the efficacy of these strategies in actually preventing pregnancies among enrolled patients, but nevertheless they represent an ethically sound approach to avoidance of fetal exposure in cases where a substantial risk of teratogenicity exists.5,27,28 Awareness of the need for pregnancy screening and prevention among girls receiving teratogenic medications is increasing in the pediatric community.12 However, a recent retrospective study demonstrated low rates (28.6%) of documented contraceptive counseling and prescribing among adolescent girls receiving teratogenic medications, highlighting the importance of improving contraceptive care as well as testing for unintended pregnancy in this population.19
Considering the PHIS hospitals system alone, at least 784 girls ≥12 years of age received 1340 cyclophosphamide infusions over the past 4 years. Assuming national rates of sexual activity and pregnancy among this population would suggest a rough estimate of 20 potential pregnancies in this cohort alone. Because almost 32% of these encounters had no recorded preexposure pregnancy screening, it is probable that multiple fetuses were at risk for exposure to the teratogenic effects of cyclophosphamide. This risk is magnified in the adolescent population as a whole given that cyclophosphamide is only one of many teratogenic medications used in treating young girl patients.
When implementing any screening initiative, it is important to consider the benefits of screening relative to the costs of the program. In terms of measurable costs, the average cost of a single urine pregnancy test at our institution during this time period was 16 US 2014 dollars. Using this estimate, the cost of urine pregnancy testing alone for the entire PHIS cohort would have been ∼$21 440. There are additional costs related to program implementation as well as costs associated with additional time required for care, however, these are difficult to quantitate and a formal cost-benefit analysis is beyond the scope of this work. However, the magnitude of the cost of universal pregnancy screening in adolescent girls seems tolerable when compared with personal and societal costs associated with the birth of even 1 child with cyclophosphamide embryopathy, which includes the potential for multiple congenital anomalies and severe developmental delay.1,2
Despite increasing awareness of the concerns regarding use of teratogenic medications in the adolescent patient population, clinicians have struggled with implementation of standard pregnancy screening protocols. Our institution was spurred into action by a patient-related sentinel event. However, neither increased provider awareness of the importance of pregnancy screening in our population nor an isolated educational intervention were sufficient to achieve reliable compliance with uniform screening. Our processes and reflections show that it took a series of interventions over time to achieve more consistent testing. One of the steps that appeared most effective was use of an automated process, such as the inclusion of a pregnancy testing order and nursing communication in the EMR surrounding pregnancy testing before cyclophosphamide administration. This finding is consistent with results of previous quality improvement studies and suggests that interventions involving hardwiring of behaviors results in greater overall compliance than education alone.29,30
A potentially novel intervention that resulted in behavior change was engagement of faculty in the change process through the MOC program. Because the order set updates and the MOC project occurred concurrently, it is difficult to discern whether one or both of these interventions were ultimately responsible for the sustained improvement. However, the potential role for MOC projects to provide a robust mechanism to facilitate institutional change and improve patient safety deserves further study.
Our study had several limitations. It is possible that factors external to our intervention could have impacted the uptake of pregnancy screening over the time period of our study. We attempted to control for this possibility by using ITS analysis. ITS analysis typically requires 8 quarters of observations per time period to provide sufficient power for comparisons.31 Our second time period did not meet this threshold number of observation points with the result that our data were underpowered to detect a difference between this time period and the other 2 study time periods. We cannot entirely rule out the presence of additional confounders that may have contributed to a general increased awareness of the importance of pregnancy screening among patients receiving teratogenic medications. However, there was no significant increase in pregnancy screening rates before cyclophosphamide administration in the PHIS data over this same time. Limitations of the PHIS analysis include lack of proximal timing of the pregnancy tests and cyclophosphamide administration to know the pregnancy test was completed before drug administration as well as potential under-ascertainment of pregnancy tests if they were performed outside the PHIS system. The PHIS population differed significantly from our internal population in terms of inpatient versus outpatient inclusion, age distribution and race/ethnicity, therefore, rigorous statistical comparison with our internal data was not feasible.
In this naturalistic study, participants changed over time and subjects could be included in multiple different time periods or multiple times within a single time period. We addressed this limitation by adjusting for differences in patient characteristics, such as age, and accounting for repeated measures in our final model. New physicians joined our practice at different points during the period of observation and may not have participated in the provider education events. However, staff turnover would have most likely reduced uptake in pregnancy screening and dampened improvement, which was not observed.
The findings of this study are applicable to other settings where girls of reproductive age are treated with potentially teratogenic medications. Availability of EMRs and standard orders is now widespread and can be used to pair pregnancy testing with medication ordering. In our experience, more invasive clinical decision support (such as a programmed hard-stop on pharmacy dispensation until completion of screening) was not required. However, our effort was supported by robust institutional structures to facilitate physician-led quality improvement work, such as the Clinical Effectiveness and MOC programs, and may not be reproducible at locations in which such infrastructure or safety culture is lacking.
Given that pregnancy is largely an unanticipated event at pediatric centers, reliable systems to detect pregnancy before administration of teratogenic medications are often lacking. We were able to demonstrate that through a variety of interventions, including provider education and EMR clinical decision support, routine pregnancy screening before cyclophosphamide administration in girls of child-bearing age can be achieved. Additional work is needed to replicate these approaches in other settings and with other potentially teratogenic medications to improve the safety of care for our adolescent girl population.
We thank Kristi Klee, clinical nurse specialist; Asa Herman, project leader; Susan Klawansky, medical librarian; and Susan Stanford and Susanne Spencer, Knowledge Management analysts of Seattle Children’s Cyclophosphamide Pathway team whose efforts made this work possible. We thank Mark Del Beccaro, Bob Sawin and Cara Bailey, sponsors; Darren Migita, Medical Director; and Kathy Mullin, Director of Seattle Children’s Clinical Effectiveness program for their support. We thank the Seattle Children’s Hospital MOC committee, Joel Tieder, MOC chair, and Jim Stout, MOC project coach, for their advice and assistance.
- Accepted June 9, 2016.
- Address correspondence to Kristen Hayward, MD, Seattle Children’s Hospital, MA.7.110, 4800 Sand Point Way NE, Seattle, WA 98105. E-mail:
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: Statistical support for this project was funded by the Seattle Children’s Clinical Effectiveness department.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- Gillis JZ,
- Panopalis P,
- Schmajuk G,
- Ramsey-Goldman R,
- Yazdany J
- Kost K,
- Henshaw S; Guttmacher Institute
- Packham JC,
- Hall MA
- Tincani A,
- Nuzzo M,
- Lojacono A,
- Cattalini M,
- Meini A
- Quinzanos I,
- Davis L,
- Keniston A, et al
- Leu MG,
- Morelli SA,
- Chung OY,
- Radford S
- Thompson KL,
- Leu MG,
- Drummond KL,
- Popalisky J,
- Spencer SM,
- Lenssen PM
- Childrens Health Corporation of America
- Main C,
- Moxham T,
- Wyatt JC,
- Kay J,
- Anderson R,
- Stein K
- Studer Q.
- Copyright © 2016 by the American Academy of Pediatrics