Use of a Checklist to Rule out Pregnancy: a Systematic Review☆ ⁎ Naomi K

Contraception 87 (2013) 661–665

Review article Use of a checklist to rule out pregnancy: a systematic review☆ ⁎ Naomi K. Tepper , Polly A. Marchbanks, Kathryn M. Curtis Division of Reproductive Health, Centers for Disease Control and Prevention, Atlanta, GA 30341, USA Received 6 July 2012; revised 27 July 2012; accepted 7 August 2012

Abstract

Background: Safe initiation of contraceptive methods requires that pregnancy be excluded. The World Health Organization has developed a list of criteria to assess pregnancy status. This review was conducted to evaluate the evidence regarding these criteria in excluding pregnancy. Study Design: The PubMed database was searched from database inception through March 2012 for all peer-reviewed articles in any language concerning the performance of a pregnancy checklist compared to urine pregnancy tests. The quality of each study was assessed using the United States Preventive Services Task Force grading system. Results: Four analyses of data from three studies met inclusion criteria as direct evidence. All were diagnostic accuracy studies of fair quality that evaluated the performance of a pregnancy checklist compared with urine pregnancy test to rule out pregnancy. The performance of the checklist varied, with sensitivity ranging from 55–100% and specificity ranging from 39–89%. The negative predictive value was consistent across studies at 99–100%. Conclusion: All four analyses demonstrated high (99–100%) negative predictive value for the pregnancy checklist. Published by Elsevier Inc.

Keywords: Pregnancy checklist; Pregnancy test; Contraception; Systematic review

1. Introduction following criteria: has not had intercourse since last normal menses, has been correctly and consistently using a reliable Safe initiation of contraception relies on accurately method of contraception, is within the first 7 days after excluding pregnancy. Pregnancy tests are reliable, inexpen- normal menses, is within 4 weeks postpartum for non- sive and easy to administer. However, they may not always lactating women, is within the first 7 days postabortion or be available or affordable, particularly in low resource miscarriage or is fully or nearly fully breastfeeding, settings, and they may not detect very early pregnancies. One amenorrheic, and less than 6 months postpartum [2]. These alternative to a pregnancy test is physical examination; criteria have been developed into a pregnancy checklist for however, this is not always feasible and is not reliable until use by family planning providers in excluding pregnancy [3]. 8–10 weeks following the last menstrual period [1]. The The US Centers for Disease Control and Prevention is World Health Organization's (WHO) Selected Practice currently undergoing a formal process to adapt the WHO Recommendations for Contraceptive Use (SPR) provides SPR for use in the United States. The objective of this guidance on the safe and effective use of contraceptive systematic review is to examine the evidence regarding the methods [1]. The WHO SPR states that certain criteria can be accuracy of the listed criteria to exclude pregnancy. used to assess pregnancy status. A provider can be reasonably sure that a woman is not pregnant if she has no signs or symptoms of pregnancy and meets any of the 2. Methods

☆ In order to assess the accuracy of a pregnancy checklist, Disclaimer: The findings and conclusions in this article are those of we searched the PubMed database for all peer-reviewed the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention. articles in any language published from database inception ⁎ Corresponding author. Tel.: +1 770 488 6506; fax: +1 770 488 6391. through March 2012 using the search term “pregnancy E-mail address: [email protected] (N.K. Tepper). checklist.” Reference lists from articles identified by the

0010-7824/$ – see front matter. Published by Elsevier Inc. http://dx.doi.org/10.1016/j.contraception.2012.08.007 662

Table 1 Evidence for accuracy of a pregnancy checklist to exclude pregnancy Author, year, Study design Population Results Strengths Weaknesses Quality location, source of support Stanback et al. [7], 1999 Diagnostic Convenience sample of 1852 non-menstruating Checklist Dipstick test Large sample size Did not provide Fair Kenya accuracy family-planning clients: Used standardized independent information Pregnant Not pregnant Supported by USAID study 59% postpartum and lactating tool to assess on accuracy of urine 37% between menses Pregnant 14 209 pregnancy status pregnancy test

4% recent abortions Not pregnant 8 1621 661 (2013) 87 Contraception / al. et Tepper N.K. All women received commercial dipstick Sensitivity=64% pregnancy test Specificity=89% Positive predictive value=6% Negative predictive value=99% Prevalence of pregnancy=1% Stanback et al. [5], 2006 Diagnostic Secondary analysis of previously collected With signs and symptoms of pregnancy (uterine Large sample size Small numbers of women Fair Kenya accuracy data (Stanback et al. [7], 1999) to assess mass, nausea, abdominal pain, spitting, vomiting, Used standardized reporting signs and symptoms Supported by USAID study signs and symptoms of pregnancy engorged breasts, missed periods, “giddiness”, tool to assess of pregnancy “laziness”): pregnancy status Did not provide independent Checklist Dipstick test information on accuracy of urine pregnancy test Pregnant Not pregnant Pregnant 14 209 Not pregnant 8 1621 Sensitivity=64% (95% CI 44–84%) – Specificity=89% (95% CI 87–90%) 665 Positive predictive value=6% (95% CI 3–9%) Negative predictive value=99% (95% CI 99–100%) Prevalence of pregnancy=1% Prevalence of signs and symptoms=1.5%

Without signs and symptoms of pregnancy: Dipstick test Pregnant Not pregnant Pregnancy not ruled out 12 191 Pregnancy ruled out 10 1639 Sensitivity=55% (95% CI 34–75%) Specificity=90% (95% CI 88–91%) Positive predictive value=6% (95% CI 3–9%) Negative predictive value=99% (99–100%) N.K. Tepper et al. / Contraception 87 (2013) 661–665 663 search strategy and review articles were hand-searched in

Fair Fair order to identify additional articles. We did not attempt to identify abstracts from scientific conferences. 2.1. Study selection

Studies were included which assessed the validity of a pregnancy checklist compared with a reference standard. Studies were excluded if they utilized a pregnancy checklist but did not report its accuracy.

Small sample size Did not provide independent information on accuracy of urine pregnancy test Did not provide independent information on accuracy of urine pregnancy test 2.2. Study quality assessment

The evidence was summarized and systematically assessed by all authors. The quality of each individual piece of evidence was assessed using the United States Preventive Services Task Force grading system for diagnostic accuracy studies [4]. ional Development; WHO, World Health Organization. tool to assess pregnancy status demonstrated that study had adequate power Used standardized tool to assess pregnancy status 2.3. Data synthesis

100%) Summary measures of association were not computed due – 8%) 8.6%)

– – to heterogeneity among studies with respect to subject characteristics. 42.4%) 98.4%) – – 100%) 66%) – –

Pregnant Not pregnant Pregnant Not pregnant 3. Results

The PubMed search strategy for articles assessing a

99.8%) pregnancy checklist identified 453 articles, of which 4 – analyses from 3 studies met inclusion criteria for this review (Table 1) [5–8]. All were diagnostic accuracy studies, comparing a pregnancy checklist to urine pregnancy testing, Specificity=60% (95% CI 54 Positive predictive value=3% (95% CINegative 1 predictive value=100% (95% CI 98 Prevalence of pregnancy=1% Positive predictive value=6% (90% CINegative 4 predictive value=99% (90% CI 96.9 Prevalence of pregnancy=4% PregnantNot pregnantSensitivity=100% (95% CI 29 0 3 157 103 Pregnancy not ruled outPregnancy ruled out 20Sensitivity=90.9% (90% CI 74.1 Specificity=38.7% (90% CI 2 35.1 314 198 Checklist Dipstick test Sample size calculation Checklist Dipstick test Used standardized which was considered to be the reference standard. All used a pregnancy checklist that was based on either the current WHO criteria [1] or a predecessor to that criteria [2].

350) One study, conducted in Kenya, used a convenience b sample of seven family-planning clinics [7]. From these clinics, 1852 non-menstruating women seeking contraception were enrolled in the study; 59% were postpartum and lactating, 37% were between menses, and 4% had recent abortions. All women were administered the checklist and a commercial dipstick pregnancy test. The prevalence of pregnancy (those detected by the pregnancy test) was 1%. 200, WHO stage IV regardless of

b The performance of the checklist in this study was calculated to be the following: sensitivity 64%, specificity 89%, Convenience sample of 535 non-menstruating women from 20 ART sites (1 woman excluded duepregnancy to test absence results) of Inclusion criteria: all women initiating(CD4 ART CD4, or WHO stage III and CD4 Exclusion criteria: self-report of pregnancy, currently menstruating, WHO stage I,or II, III with no CD4All count women received urine dipstick pregnancy test 263 non-menstruating women desiring oral contraceptives Postpartum women excluded All women received urine dipstick pregnancy test positive predictive value 6%, and negative predictive value more than 99%. A further analysis of the study from Kenya was conducted in order to examine the value of assessing signs and Diagnostic accuracy study Diagnostic accuracy study symptoms of pregnancy along with the other criteria in the checklist [5]. From the same 1852 enrolled women, the

, 2008 following signs and symptoms were recorded in 27 women , 2010 [6] (4 of whom were pregnant): uterine mass, nausea, abdominal [8] pain, spitting, vomiting, engorged breasts, missed periods for 2 months, “giddiness,” and “laziness.” The performance of the checklist when the information on pregnancy signs and Zambia Supported by FHI/Zambia Prevention Care and Treatment Partnership and USAID Nicaragua Supported by USAID Torpey et al. Abbreviations: ART, antiretroviral therapy; CI, confidence interval; FHI, Family Health International; USAID, United States Agency for Internat Stanback et al. symptoms was included did not markedly differ from the 664 N.K. Tepper et al. / Contraception 87 (2013) 661–665 performance of the checklist without pregnancy signs and The body of evidence was of fair quality and was limited symptoms. Certain signs and symptoms were more specific by a small number of studies reflecting a narrow sample of to pregnancy in this study. Both women in whom uterine population. Sample sizes were relatively large, although only masses were detected were pregnant. However, among one study completed a power calculation that determined the women who reported engorged breasts (N=2) and missed sample size was adequate [8]. One study was restricted to periods (N=1), none were pregnant. HIV-infected women, which may limit its generalizability to Another study, conducted in Nicaragua, assessed the the broader population of women seeking family planning performance of a pregnancy checklist among 263 women services. None of the studies reported specifically which being enrolled for a randomized trial of oral contraceptives pregnancy test was used and therefore the performance of the [6]. Non-menstruating women desiring oral contraceptives reference standard cannot be assessed from these studies. were administered a pregnancy checklist and a urine The benefits of a pregnancy checklist should be pregnancy test. Postpartum women were excluded. The considered in the context of performance of the reference prevalence of pregnancy (those detected by the pregnancy standard. Studies included in this systematic review used test) was 1%. The performance of the checklist in this study urine pregnancy tests as the reference standard, although was calculated to be the following: sensitivity 100% (95% these also have inherent limitations. Pregnancy detection confidence interval [CI], 29–100%), specificity 60% (95% rates can vary widely based on the sensitivity of the test and CI, 54–66%), positive predictive value 3% (95% CI, 1–8%), the timing of testing with respect to missed menses [9]. One and negative predictive value 100% (95% CI, 98–100%). review found pregnancy detection rates as low as 60% The third study, conducted in Zambia, was undertaken to among over-the-counter tests and 40% among point of care assess the performance of a pregnancy checklist among HIV- tests when pregnancy testing was performed on the day of infected women initiating antiretroviral therapy (ART) [8].A missed menses [9]. Another study found that, among 6 convenience sample of 20 sites was used to enroll 535 different pregnancy tests, the detection rates ranged from women (one woman was subsequently excluded for lack of 55% to 100% on the day of missed menses. By 3 days pregnancy test results). Eligibility criteria included those following missed menses, detection rates ranged from 75% criteria for ART initiation: CD4 cell count less than 200, to 100% [10]. Another study found that the estimated day of WHO Stage IV regardless of CD4 count, or WHO Stage III implantation occurred after the first day of the next expected and CD4 count less than 350. Women were excluded who period in 10% of clinical pregnancies, thus making self-reported pregnancy, were currently menstruating, or pregnancy detection impossible at the time of missed menses who were WHO Stage I, II or III with no CD4 count. The [11]. Several studies have found that an additional 11 days prevalence of pregnancy (those detected by the pregnancy past the day of expected menses were needed to detect 100% test) was 4%. The performance of the checklist in this study of pregnancies [12]. The pregnancy checklist may have was calculated to be the following: sensitivity 90.9% (90% particular utility early in pregnancy, before the time when CI, 74.1–98.4%), specificity 38.7% (90% CI, 35.1–42.4%), urine pregnancy tests are most accurate. positive predictive value 6.0% (90% CI, 4.0–8.6%), and Performance of pregnancy tests is also dependent on who negative predictive value 99.0% (90% CI, 96.9–99.8%). administers the test. Although most urine pregnancy tests are reported to be 99% accurate, the true accuracy can vary from 50% to 75%, depending on who administers the test [13].A 4. Discussion meta-analysis examining the accuracy of home pregnancy tests found that the sensitivity varied from 52% to 100% and Three diagnostic accuracy studies assessed the validity of the specificity varied from 52% to 100% [14]. However, the a pregnancy checklist when compared to the reference sensitivity and specificity are 97% to 100% when performed standard urine pregnancy test. The performance of the in the laboratory [15]. checklist varied in different studies, with sensitivity ranging Due to its ability to correctly rule out pregnancy, the from 55–100% and specificity ranging from 39–89%. The pregnancy checklist may have implications for expanding negative predictive value, however, was consistent across access to contraception, particularly if pregnancy tests are studies at 99–100%. The authors of these papers stated that not available or feasible. A study was conducted in the high negative predictive value is perhaps the most Guatemala, Mali and Senegal to determine whether the important statistic, because it means that the checklist will pregnancy checklist improved women's access to contra- correctly rule out women who are not pregnant. The ceptive services by reducing the proportion of women implication is that women who are not pregnant will not be denied contraception because they were not menstruating denied contraception because a pregnancy test might not be [16]. In Guatemala and Senegal, the proportion of new available or feasible. In the one study that assessed the added clients denied their desired method of contraception utility of signs and symptoms of pregnancy, these criteria because of menstrual status decreased significantly after were not found to substantially improve the performance of introducing the pregnancy checklist. No significant change the pregnancy checklist, although the number of women with was observed in Mali; however, the proportion denied signs and symptoms was small [5]. services at baseline was low (5%). N.K. Tepper et al. / Contraception 87 (2013) 661–665 665 5. Conclusions rule out pregnancy? J Fam Plann Reprod Health Care 2006;32: 27–9. [6] Stanback J, Nanda K, Ramirez Y, Rountree W, Cameron SB. Four analyses of three fair quality diagnostic accuracy Validation of a job aid to rule out pregnancy among family planning studies evaluated the performance of a pregnancy checklist clients in Nicaragua. Rev Panam Salud Publica 2008;23:116–8. compared with urine pregnancy test to rule out pregnancy. [7] Stanback J, Qureshi Z, Sekadde-Kigondu C, Gonzalez B, Nutley T. The performance of the checklist varied, with sensitivity Checklist for ruling out pregnancy among family-planning clients in ranging from 55% to 100% and specificity ranging from primary care. Lancet 1999;354:566. [8] Torpey K, Mwenda L, Kabaso M, et al. Excluding pregnancy among 39% to 89%. The negative predictive value was consistent women initiating antiretroviral therapy: efficacy of a family planning across studies at 99% to 100%. job aid. BMC Public Health 2010;10:249. [9] Cole LA. Human chorionic gonadotropin tests. Expert Rev Mol Diagn References 2009;9:721–47. [10] Cole LA. The utility of six over-the-counter (home) pregnancy tests. [1] World Health Organization. Selected Practice Recommendations for Clin Chem Lab Med 2011;49:1317–22. Contraceptive Use. Second edition. Geneva: World Health Organiza- [11] Wilcox AJ, Baird DD, Dunson D, McChesney R, Weinberg CR. tion; 2004. Available at http://www.who.int/reproductivehealth/ Natural limits of pregnancy testing in relation to the expected publications/family_planning/9241562846index/en/index.html. menstrual period. JAMA 2001;286:1759–61. Accessed April 11, 2012. [12] Cervinski MA, Gronowski AM. Reproductive-endocrine point-of-care [2] Technical Guidance Working Group. Recommendations for updating testing: current status and limitations. Clin Chem Lab Med 2010;48: selected practices in contraceptive use: results of a technical meeting, 935–42. volume 1. Washington DC: United States Agency for International [13] Eichner SF, Timpe EM. Urinary-based ovulation and pregnancy: Development; 1994. point-of-care testing. Ann Pharmacother 2004;38:325–31. [3] Family Health International. Checklist for how to be reasonably sure a [14] Bastian LA, Nanda K, Hasselblad V, Simel DL. Diagnostic efficiency client is not pregnant. 2008. Available at: http://www.fhi360.org/NR/ of home pregnancy test kits. A meta-analysis. Arch Fam Med rdonlyres/eg7priv4amrttz2ejss2dds5qzjdhhf3wtq6e3waqemin3itekw7 1998;7:465–9. mwm6h73yfugw5vl7qimnmpqszh/PregnancyChecklistenrh1.pdf. [15] Bastian LA, Piscitelli JT. Is this patient pregnant? Can you reliably rule Accessed February 21, 2012. in or rule out early pregnancy by clinical examination? JAMA [4] Harris RP, Helfand M, Woolf SH, et al. Current methods of the US 1997;278:586–91. Preventive Services Task Force: a review of the process. Am J Prev [16] Stanback J, Diabate F, Dieng T, de Morales TD, Cummings S, Med 2001;20(3 suppl):21–35. Traore M. Ruling out pregnancy among family planning clients: the [5] Stanback J, Nakintu N, Qureshi Z, Nasution M. Does assessment of impact of a checklist in three countries. Stud Fam Plann 2005;36: signs and symptoms add to the predictive value of an algorithm to 311–5.