Population-level data on antenatal screening for proteinuria; India, Mozambique, Nigeria, Pakistan

Abstract Objective To estimate the prevalence and prognosis of proteinuria at enrolment in the 27 intervention clusters of the Community-Level Interventions for Pre-eclampsia cluster randomized trials. Methods We identified pregnant women eligible for inclusion in the trials in their communities in four countries (2013–2017). We included women who delivered by trial end and received an intervention antenatal care visit. The intervention was a community health worker providing supplementary hypertension-oriented care, including proteinuria assessment by visual assessment of urinary dipstick at the first visit and all subsequent visits when hypertension was detected. In a multilevel regression model, we compared baseline prevalence of proteinuria (≥ 1+ or ≥ 2+) across countries. We compared the incidence of subsequent complications by baseline proteinuria. Findings Baseline proteinuria was detected in less than 5% of eligible pregnancies in each country (India: 234/6120; Mozambique: 94/4234; Nigeria: 286/7004; Pakistan: 315/10 885), almost always with normotension (India: 225/234; Mozambique: 93/94; Nigeria: 241/286; Pakistan: 264/315). There was no consistent relationship between baseline proteinuria (either ≥ 1+ or ≥ 2+) and progression to hypertension, maternal mortality or morbidity, birth at < 37 weeks, caesarean section delivery or perinatal mortality or morbidity. If proteinuria testing were restricted to women with hypertension, we projected annual cost savings of 153 223 981 United States dollars (US$) in India, US$ 9 055 286 in Mozambique, US$ 53 181 933 in Nigeria and US$ 38 828 746 in Pakistan. Conclusion Our findings question the recommendations to routinely evaluate proteinuria at first assessment in pregnancy. Restricting proteinuria testing to pregnant women with hypertension has the potential to save resources.


Introduction
Hypertensive disorders of pregnancy are a leading cause of maternal and perinatal death and disability worldwide. As such, antenatal care is devoted in large part to the detection of pregnancy hypertension and in particular pre-eclampsia. Pre-eclampsia is the most dangerous form of pregnancy hypertension, being responsible for approximately one quarter of maternal deaths and serious near-miss morbidities. 1 Pre-eclampsia most commonly manifests as hypertension and proteinuria, so the World Health Organization (WHO) recommends and considers essential the measurement of blood pressure and proteinuria at each antenatal care contact. 2 While antenatal proteinuria testing for pregnancies at 20 or more weeks gestation has the potential to detect the proteinuria of pre-eclampsia, such testing at any gestational age might reveal underlying chronic kidney disease, which is itself associated with adverse outcomes.
The value of proteinuria testing at antenatal care contacts for pregnant women without high blood pressure has been questioned, however. First, WHO, in a discussion of asymptomatic bacteriuria, endorsed the widely-held view that dipstick proteinuria testing for pre-eclampsia has low diagnostic accuracy. 2 Second, proteinuria testing may be less specific in very hot climates or during dry seasons when women may become dehydrated. Third, it is rare for women to present with proteinuria before the hypertension of pre-eclampsia. 3 Fourth, proteinuria screening may impede progress towards group antenatal care given the need for privacy and toilet facilities. Finally, devoting resources to routine proteinuria screening has been questioned when most antenatal care contacts will not be associated with proteinuria. 4 The Community-Level Interventions for Pre-eclampsia trials were cluster randomized controlled trials of community health worker (CHW)-based diagnosis and initial management of women with hypertension in pregnancy. The trials took place in four low-or lower-middle-income settings in India, Mozambique, Nigeria and Pakistan. 5 Proteinuria testing was performed at baseline for all women and at all subsequent visits if elevated blood pressure were found. We report the Research Antenatal screening for proteinuria, India, Mozambique, Nigeria and Pakistan Laura A Magee et al.
incidence of baseline proteinuria assessed at the first visit and the relationship between baseline proteinuria and hypertension, preterm birth, caesarean section delivery, and maternal and perinatal mortality and morbidity.

Methods
In this exploratory secondary analysis, we included data from the 27 intervention clusters of the Community-Level Interventions for Pre-eclampsia cluster randomized controlled trials (NCT01911494). 5 The trials comprised primary rural clusters: six clusters in and 2590-Obs-ERC-13, Pakistan). The protocol is published 5 and included in the authors' data repository, 6 along with the statistical analysis plan and STROBE (Strengthening the Reporting of Observational studies in Epidemiology) checklist. 6

Trial design
We enrolled women aged 12-49 years in the trials after they had confirmed their pregnancy and given informed consent. The intervention was carried out in a community setting and consisted of community engagement and a clinical assessment with initial treatments and referrals to health-care facilities provided by a CHW. The CHW was guided by a mobile health application based on the miniPIERS Pre-eclampsia Integrated Estimate of RiSk predictive model in hypertensive pregnancy. The application, running on mobile devices (tablet computers), provided step-by-step guidance on assessment and decision support for triage, transport and treatment of women with hypertension or emergency medical conditions. [7][8][9] In the trial protocol, visits were recommended for women at least every 4 weeks before birth. CHWs measured the women's blood pressure at every intervention visit in a standardized fashion using a device validated for use in pregnancy and pre-eclampsia (3AS1-2 ® semi-automatic blood pressure monitor, Microlife, Clearwater, United States of America). 10 The CHWs also carried out proteinuria screening for all women at the first intervention visit and at subsequent visits only if hypertension were detected. Proteinuria screening was carried out by visual assessment of urinary dipsticks. Women in control clusters received usual care, consisting of blood pressure measurement (using the device available) and proteinuria testing at each antenatal care contact, according to WHO guidelines. 2 The primary outcome measure was a composite of maternal, fetal and newborn mortality and serious morbidity, such as eclampsia or pulmonary oedema. Maternal mortality was measured to 6 weeks and neonatal mortality to 28 days after birth. Hypertension was a systolic blood pressure ≥ 140 mmHg or a diastolic blood pressure ≥ 90 mmHg. Proteinuria was defined in two ways, as a urinary dipstick result of ≥ 1+ or ≥ 2+, according to manufacturer's instructions.
Surveillance data were collected by a separate team, by household survey (quarterly in Pakistan and 6-monthly in Mozambique and Nigeria) or a research registry (India). 11 In Nigeria, trial surveillance was suspended and the trial closed after the pilot phase because of challenges with data collection. In all countries in the trials the data were entered directly onto the mobile devices by the CHWs.
Data entered on mobile devices were synchronized and stored on the Research Electronic Data Capture servers. We transferred de-identified data from the trial (intervention clusters) and surveillance data (all clusters) to the University of British Columbia Community-Level Interventions for Pre-eclampsia coordinating centre. Data management protocols ensured security (encryption), tracking (user identification numbers and audit trails) and synchronization between devices within the cluster and with the server.

Analysis
For this analysis, we included pregnancies in intervention clusters in which the woman had received at least one mobile application-guided (intervention) visit and delivered by the trial end. Women in control clusters did not receive interven-tion visits, by definition. We excluded pregnancies in which the woman was still on follow-up and undelivered to avoid underestimation of hypertension and adverse outcomes.
Our analyses included those pregnancies with complete information for variables of interest. We treated intervention clusters in each country as one cohort for our primary analysis comparing proteinuria prevalence at booking, that is, the first intervention visit of the study. We summarized continuous data by median and interquartile range and categorical data by number and proportion. Four-way between-country comparisons were made by χ 2 -test for categorical variables, or Kruskal-Wallis test for continuous variables, as appropriate. When comparisons were significant, we made pairwise comparisons by χ 2 -test and Wilcoxon rank sum test, as appropriate, to ascertain differences among countries.
We used logistic regression adjusted for country to compare proteinuria prevalence at the first intervention visit among countries. To explore whether proteinuria was related to pregnancy outcomes within a country, we matched controls to each woman with proteinuria according to individual characteristics: maternal age, parity, basic education, gestational age at enrolment (usually by last menstrual period), cluster and distance to facility. 12 This analysis was not possible in Nigeria due to the absence of outcome data. We pooled data and calculated the overall odds ratios (ORs) for outcomes. We computed confidence intervals (CIs) for each outcome via bootstrapping, through 1000 iterations of the entire matching process. We did this to quantify variability in matching as there were many possible control matches for each case, and to prevent the results being dependent on which match was chosen.
We estimated the financial implications of a strategy of testing proteinuria only for women with hypertension. We estimated the number of antenatal care visits at which proteinuria testing would be avoided by using national estimates of annual number of births (India: 24 229 725; Mozambique: 1 085 797; Nigeria: 7 329 535; Pakistan: 5 945 845). 13 We also estimated the incidence of normotension in pregnancy from population-based estimates of pregnancy hypertension from the  14 and an eight-visit antenatal care contact model as per WHO guidelines. 2 We calculated the cost of supplies used for proteinuria testing from the budgetary statements of the trials (2013-2017) in United States dollars (US$), inflated to 2019 US$. 15 We made four additional sensitivity analyses. We explored whether between-country differences in proteinuria were affected by women's baseline characteristics: age and parity (in all countries), maternal basic education (except in Nigeria) and gestational age at the first intervention visit. We also explored the effect of multiple pregnancies by excluding women with more than one pregnancy (in all countries), and the effect of antiretroviral therapy by excluding women with human immunodeficiency virus infection (HIV; in Mozambique). Finally, we estimated the financial implications of a strategy of testing proteinuria only for women with hypertension based on the former four-visit model.
We performed analyses using R programming software, version 3.3.2 (R Foundation for Statistical Computing, Vienna, Austria). 16 P < 0.05 was considered statistically significant.
The baseline characteristics of the included pregnancies differed across countries (Table 1). At enrolment in the trials, women in India and Mozambique were slightly younger than those in Pakistan and Nigeria. About one-third of women were nulliparous, except in Pakistan where the proportion was closer to one-fifth. Women in India were enrolled earlier, at the end of the first trimester, compared with early (Nigeria and Pakistan) or late (Mozambique) second trimester in the other countries. Levels of maternal education were low, particularly in Pakistan. The prevalence of HIV-positivity was 23.0% (972/4234) in Mozambique, the only country where this was measured, and the majority of HIV-positive women were on antiretroviral therapy. There were few women with multiple pregnancies, although more in Mozambique. Women in all countries delivered at a median of 39 weeks, although slightly earlier in Pakistan. The trials' data did not include information about smoking, body mass index or prior pre-eclampsia (due to low health literacy of the mothers). In all countries, women who received an intervention visit(s) were similar at baseline to those who did not, although gestational age at enrolment in the trials was 2-4 weeks earlier (data repository); 6 information was unavailable in Nigeria.
The frequency and quality of intervention antenatal visits has been previously reported. 14 In brief, most intervention visits began 2-4 weeks after enrolment in the trials (except in Nigeria where there was a larger delay) and at < 20 weeks in most pregnancies in India, just under half in Pakistan, and a distinct minority in Mozambique and Nigeria (Table 2). This resulted in more intervention visits in India than in the other countries. Proteinuria screening was undertaken at the first intervention visit for more than 90.0% of pregnancies in each country, and at subsequent antenatal visits for more than 90.0% of pregnancies with hypertension detected, except in Nigeria where it was 80.7% ( Note: Data on enrolment were derived from all women who had an intervention antenatal care visit, as trial surveillance data were not available. Maternal basic education was defined as 8 years or more of schooling (India), achievement of grade 5 or above (Mozambique) and 5 years or more of schooling (Pakistan). c Trial surveillance data were not available for Nigeria. d Questions about HIV and antiretroviral therapy were not asked in India and Pakistan. Notes: n is the total number of pregnancies included in this analysis. At intervention visits women received a clinical assessment by a community health worker guided by a mobile health application, including dipstick proteinuria assessment at the first community-level visit and all subsequent visits when hypertension was detected.   n is the total number of pregnancies included in this analysis. At intervention visits women received a clinical assessment by a community health worker guided by a mobile health application. The trial protocol specified that proteinuria should be measured at the first intervention visit, and then at subsequent visits at which the woman was hypertensive. In Nigeria, proteinuria was measured at many subsequent visits regardless of blood pressure status ( Table 3). The prevalence of proteinuria ≥ 1+ was significantly different across countries (P < 0.001); it was highest in Nigeria, followed by India and then Pakistan, with the lowest prevalence in Mozambique. The ratio of pregnancies with 1+ proteinuria (as opposed to ≥ 2+) varied substantially from 5:1 in Pakistan to 1:1 in India (Table 3). However, in sensitivity analyses, adjustment for maternal characteristics and gestational age at the first intervention visit revealed that all countries had a lower prevalence of baseline proteinuria compared with India (Table 3).
In most of the pregnancies with proteinuria, the women were normotensive at the first intervention visit in each of the four countries (India: 225/234; Mozambique: 93/94; Nigeria: 241/286; Pakistan: 264/315; Table 3). Among women who were normotensive at their first visit, those with proteinuria had similar pregnancy outcomes, progression to hypertension, and maternal and perinatal mortality and morbidity compared with those without proteinuria (Table 4 and data repository). 6 While the 95% CIs around the OR for these events were wide, the percentages were similar, with no consistent patterns of increasing proportions of pregnancies with adverse outcomes with increasing proteinuria (data repository). 6 The results were similar when women with multiple pregnancies or those known to be HIV-positive were excluded in sensitivity analyses (data repository). 6 We estimated that at the national level in the trial countries, there would be large numbers of antenatal care visits at which proteinuria testing would not be undertaken if restricted to women with hypertension (annually number

Discussion
In almost 30 000 pregnancies from 27 intervention clusters in sub-Saharan Africa and South Asia, we demonstrated a very low prevalence of dipstick proteinuria by visual assessment soon after antenatal care booking. The prevalence of proteinuria in each country was related to maternal characteristics and gestational age at booking. Few women with baseline proteinuria had hypertension at the first intervention visit and among those with blood pressure in the normal range, there was no compelling relationship between baseline proteinuria and adverse pregnancy outcomes.
Previous studies of proteinuria screening have focused on its role in Table 3. pre-eclampsia diagnosis in women with hypertension, rather than the added value of proteinuria screening when blood pressure is normal. Even WHO regards routine baseline assessment and ongoing surveillance of proteinuria testing as good practice without the need for evidence review. 2 Our findings suggest otherwise.

Prevalence of proteinuria at the first antenatal intervention visit in the Community-Level Interventions for Pre-eclampsia trial and relationship with adverse outcomes, 2013-2017
Our measure of baseline proteinuria may reflect chronic kidney disease, pre-eclampsia (for women assessed at gestational age ≥ 20 weeks) or another process, such as dehydration, vaginal contamination, urinary tract infection, exercise or orthostatic proteinuria. Chronic kidney disease complicates up to 3% of pregnancies. 17,18 Our results are consistent with this estimate, although we have inferred the possibility of chronic kidney disease from baseline proteinuria, rather than the accepted diagnostic tests in pregnancy that are not performed routinely anywhere (serum creatinine, quantitative proteinuria evaluation and urinalysis). Estimated glomerular filtration rate is inaccurate in pregnancy. 19,20 Although chronic kidney disease in pregnancy is mostly mild, it is still associated with adverse pregnancy outcomes. 21 As we were unable to demonstrate a relationship between baseline dipstick proteinuria and adverse pregnancy outcomes among almost 30 000 pregnancies, it may be that the relationship between chronic kidney disease and adverse outcomes is different in our study settings. More likely, dipstick proteinuria has such low diagnostic accuracy for true proteinuria and underlying chronic kidney disease that it is not a useful test. 2 We were unable to estimate the incidence of new proteinuria without hypertension, which has been documented by others to be low (0.5% to 1.9%) in unselected pregnancies. 3,22 Most of these women (75.3%) did not develop hypertension indicative of pre-eclampsia. In our trials, the number-needed-toscreen by blood pressure measurement to detect pregnancy hypertension is 10 women (80 visits). In contrast, the number-needed-to-screen by dipstick proteinuria to detect proteinuria that will progress to pre-eclampsia is at least 213-769 women (1704-6152 visits) as these women will of course also be screened with blood pressure measure-   14 gestational proteinuria of 0.5-1.9% in unselected pregnancies, 3,22 progression of gestational proteinuria to pre-eclampsia of 24.7% of pregnancies, 3,22 and an eight-contact antenatal care model 2 in which all women undergo blood pressure measurement and proteinuria screening at each contact.
We believe the practice of proteinuria screening among normotensive women in pregnancy should be questioned. First, there is the high volume and cost of testing to detect one woman with isolated proteinuria and normal blood pressure. Second, our findings suggest there is no compelling relationship between isolated baseline proteinuria and adverse outcomes. Third, after the booking visit, there is no evidence that subsequent antenatal care contacts with proteinuria testing (versus those without) result in superior outcomes. Finally, more than 80% of proteinuria occurs in the third trimester. 23 We have demonstrated that screening for proteinuria only in the presence of hypertension could be undertaken to inform a diagnosis of pre-eclampsia. This practice would not increase the incidence of pre-eclampsia, as most women in our trials presented with non-severe, gestational hypertension. 14 Importantly, however, this practice could be associated with large cost savings for health systems in low-resource settings. 14 Although the cost of each proteinuria screen is low (about US$ 1), the use of dipstick screening at each antenatal care contact for each woman results in a large cumulative sum. In line with the Choosing Wisely movement, it is reasonable to ask whether those funds could be used in other ways to optimize outcomes. 4 The strengths of our study include evaluation of a large number of women in four low-resource sub-Saharan and south Asian countries. We estimated baseline proteinuria prevalence independent of hypertension, and considered the utility of this measurement using common definitions of dipstick proteinuria.
Some limitations of our study include having proteinuria measurements from most but not all women, from bookings only in community care and not at subsequent antenatal care visits by women with normotension, and from measurements by CHWs. We do not know how many women initially tested negative for proteinuria but later developed isolated proteinuria. However, it is unlikely that we missed earlier presentations of pre-eclampsia, as the incidence of gestational hypertension in our trials (6.5-8.4%) 14 was as high as in settings where antenatal care contacts are frequent. Second, we had data only on basic maternal characteristics for adjusted analysis of proteinuria prevalence. As is typical in the settings where we carried out the trials, no reliable information was available on women's past history of chronic hypertension or renal disease to differentiate prior renal disease from pre-eclampsia (if booking occurred after 20 weeks). We had no information on the type of antiretroviral therapy taken by women with HIV (although exclusion of HIV-positive women left the results unchanged). We also had no direct measure of dehydration or any information on whether women were dehydrated due to manual labour occupations or a lack of toilet facilities. Third, while the 95% CIs were wide around our outcome estimates, we must question the importance of an effect that cannot be demonstrated among more than 30 000 women. The relationship between baseline proteinuria and adverse outcomes would only have been strengthened by the fact that more than 50% of women in Pakistan and most women in Mozambique booked after 20 weeks, as their baseline proteinuria could have reflected pre-eclampsia. Finally, we illustrated potential health system cost savings of a strategy of proteinuria testing only in pregnant women with hypertension. However, we did not undertake a formal cost-consequences analysis and we acknowledge that government bulk purchase of testing supplies may lower costs. Also, while our calculations were based on an eight-visit antenatal care model, we estimate that a four-visit model would still be associated with substantial numbers of proteinuria assessments avoided in normotensive pregnancy and cost savings.
In conclusion, our findings do not support the usefulness of proteinuria screening at the first assessment in pregnancy. This practice should be reevaluated and robust health economic studies undertaken, to avoid unnecessary tests and treatments that fail to add value to care, consume resources and may cause harm through follow-up investigation and worry for women. 4