Pregnancy-Associated Changes in Pharmacokinetics: A Systematic Review

Background Women are commonly prescribed a variety of medications during pregnancy. As most organ systems are affected by the substantial anatomical and physiological changes that occur during pregnancy, it is expected that pharmacokinetics (PK) (absorption, distribution, metabolism, and excretion of drugs) would also be affected in ways that may necessitate changes in dosing schedules. The objective of this study was to systematically identify existing clinically relevant evidence on PK changes during pregnancy. Methods and Findings Systematic searches were conducted in MEDLINE (Ovid), Embase (Ovid), Cochrane Central Register of Controlled Trials (Ovid), and Web of Science (Thomson Reuters), from database inception to August 31, 2015. An update of the search from September 1, 2015, to May 20, 2016, was performed, and relevant data were added to the present review. No language or date restrictions were applied. All publications of clinical PK studies involving a group of pregnant women with a comparison to nonpregnant participants or nonpregnant population data were eligible to be included in this review. A total of 198 studies involving 121 different medications fulfilled the inclusion criteria. In these studies, commonly investigated drug classes included antiretrovirals (54 studies), antiepileptic drugs (27 studies), antibiotics (23 studies), antimalarial drugs (22 studies), and cardiovascular drugs (17 studies). Overall, pregnancy-associated changes in PK parameters were often observed as consistent findings among many studies, particularly enhanced drug elimination and decreased exposure to total drugs (bound and unbound to plasma proteins) at a given dose. However, associated alterations in clinical responses and outcomes, or lack thereof, remain largely unknown. Conclusion This systematic review of pregnancy-associated PK changes identifies a significant gap between the accumulating knowledge of PK changes in pregnant women and our understanding of their clinical impact for both mother and fetus. It is essential for clinicians to be aware of these unique pregnancy-related changes in PK, and to critically examine their clinical implications.


Introduction
Women frequently take a variety of medications during pregnancy, including prescription, over-the-counter (OTC), and herbal agents [1,2]. During the last three decades the average number of medications (prescription and nonprescription) used per woman in North America during the first trimester increased by 60% from 1.6 to 2.6 [3]. More recently, from 2006 to 2008, over 80% of women reported using at least one medication during the first trimester, and over 90% reported using at least one medication at any point during their pregnancy [3]. Other studies have demonstrated increased rates of use of various OTC medications in the first, second, or third trimester of pregnancy compared to the prepregnancy period [4]. While some studies have found that the proportion of women receiving at least one prescription medicine increases from the first to third trimester of pregnancy [5,6], others have found that rates of prescription drug use are highest in the first trimester of pregnancy [1,7]. The most common medications used in pregnancy are nonprescription or OTC medications [4]. A longitudinal study aimed at identifying the medications that are most often consumed during pregnancy demonstrated that 95.8% of participants took prescription medications, 92.6% self-medicated with OTC medications, and 45.2% used herbal medications [2].
Most organ systems are affected by substantial anatomical and physiological changes during pregnancy. Such pregnancy-related changes are observed in decreased gastrointestinal motility and increased gastric pH (impacting absorption), increased total body water and plasma volume and decreased concentrations of drug-binding proteins (affecting the apparent volume of distribution and, in some cases, clearance rates), increased glomerular filtration rate (increasing renal clearance), and altered activity of drug-metabolizing enzymes in the liver (affecting hepatic clearance). Overall, these changes in physiological indices take place progressively during gestation (reviewed in [8] and [9]). The increases in cardiac output, total body water, fat compartment, and glomerular filtration rate, together with the decrease in plasma albumin concentration and altered activity of drug-metabolizing enzymes, are all reported to peak during the third trimester (reviewed in [8] and [10]). Table 1 presents typical pregnancy-related changes in organ function leading to altered pharmacokinetics (PK) [10][11][12][13][14][15][16]. Changes during pregnancy in drug metabolism by cytochrome P450 isoenzymes (i.e., CYP3A4, CYP2D6, CYP2C9, CYP1A2, and CYP2C19) and by uridine 5 0 -diphospho-glucuronosyltransferase (UGT) isoenzymes (i.e., UGT1A4 and UGT2B7) have also been demonstrated (Table 2) [10,[17][18][19][20].
For some drug classes, a large number of PK clinical trials during pregnancy are available in the literature [29][30][31][32][33][34]. A recent review noted that, since 2008, about a third of these trials investigated drugs used in the treatment of acute labor and delivery issues, another third investigated drugs used in infectious disease treatment during pregnancy, and the remaining third investigated drugs used for various antepartum indications [35]. However, for the large majority of drugs used during pregnancy, there is little or no information available regarding PK changes or dosage requirements during pregnancy [35]. Moreover, it is often unclear if observed PK changes lead to alterations in drug efficacy and/or adverse effect profiles. Given the complexity of the field, the lack of clear understanding of the clinical significance of PK changes, and renewed recognition of the need to rationalize drug therapy for pregnant and lactating women, it is imperative to systematically examine existing data on PK changes in pregnancy and their potential clinical impact. The objective of this study was to systematically identify all existing evidence of PK changes during pregnancy in the context of clinical significance. We hypothesized that known physiological changes occurring during pregnancy and associated PK alterations could consequently be translated into changes in dosing guidelines.

Methods
This research involved a structured review of the literature, according to the PRISMA guidelines [36] (S1 Checklist).

Search Strategy
Searches were conducted in MEDLINE (Ovid), Embase (Ovid), Cochrane Central Register of Controlled Trials (Ovid), and Web of Science (Thomson Reuters) from database inception to August 31, 2015 (S1 Table). An update of the search from September 1, 2015, to May 20, 2016, was performed, and relevant data were retrieved and added to the review (S2 Table). Text words and, where applicable, database subject heading fields (e.g., MeSH) were used for the following concepts: pregnancy AND pharmacokinetics OR dosing OR clearance OR distribution OR absorption OR metabolism OR excretion OR Cmax OR Tmax OR Ctrough OR AUC OR Vd OR t1/2 OR protein binding AND specific study types (randomized controlled trial, nonrandomized controlled clinical trial, cohort study, case-control study, or case series). Truncation symbols were used with the text words, when appropriate, to capture variations in spelling and word endings. Subsequently, we reviewed the identified studies and examined their references to identify further potential articles. Information available from relevant conferences was also reviewed. No publication date, language, or location restrictions were applied.

Study Selection
In order to locate all published literature, we established a set of criteria to define types of studies to be reviewed. Inclusion criteria were as follows: (1) the study reported dosing data or at least one PK parameter of interest in pregnant women; (2) a comparison of the dosing data or PK parameter between pregnant and nonpregnant women was done; and (3) the data are described in the form of a peer-reviewed randomized controlled trial, non-randomized controlled clinical trial, cohort study, case-control study, or case series. The review did not cover animal studies, case reports, or studies containing no original research or data. Retrieved articles were inspected by two independent reviewers (G. P. and T. L.) to determine whether they met the inclusion criteria. In cases where the eligibility of the study was unclear, it was reviewed by a third independent reviewer (G. K.). The full texts were retrieved and read in full.

Data Extraction
The data extractors (G. P. and T. L.) reviewed each of the included studies independently and extracted data according to the predetermined guidelines, using a predesigned data extraction form. When needed, authors of the included studies were contacted for missing data; however, none of the authors who were contacted for more information responded. Data from studies presented in multiple publications were identified to avoid duplications and were reported as a single study, with all other relevant publications listed.

Data Presentation and Analysis
Results of the literature search. The results from each step of the review process are documented in a PRISMA flow diagram (Fig 1), with an overall summary of the number and types of articles included in the review.
When more than one study reported the same PK parameter(s) for the same drug, these parameters were examined for consistency in the change direction (i.e., decrease, increase, or no change). When study data were presented by trimester, the PK parameters obtained during the third trimester were selected for this study because the majority of the pregnancy-associated physiological changes peak during the third trimester. Drugs were divided into two major categories according to between-study agreement of directions of statistically significant changes in PK parameters. If statistically significant pregnancy-associated changes in PK parameters were in the same direction (e.g., increase in clearance and decrease in volume of distribution) among the studies for all reported PK parameters, we categorized the drug as "consistent. " On the other hand, a drug was categorized as "inconsistent" if at least one study reported a statistically significant change in a PK parameter in the opposite direction (e.g., increased Cl in one study and decreased Cl in the other). The potential source of inconsistency is speculated on and addressed in the Discussion. Note that the definition of the categories described above is based on statistically significant changes of PK parameters, but statistically non-significant changes are also presented, for completeness. In addition, if only one study showed a statistically significant PK parameter change for a drug, the drug was included in the "consistent" category for simplicity of the data presentation, even though the PK parameters were reported in only one study.
Quality assessment. The quality of each accepted article was assessed using the ClinPK checklist [37] for assessing methodological quality in clinical PK studies ( Table 3).
No discrepancies exist between the original protocol and the final data analyses.

Literature Retrieval
The search strategy for the comprehensive systematic review retrieved 9,562 articles, and after removing duplicates, the first screen on title and abstract was performed on 7,163 articles ( Fig  1). For 6,935 of these, the title or abstract clearly indicated that the topic of the article was not relevant to the review question or did not satisfy one of the inclusion criteria. The remaining 228 articles were screened using the full text, applying the full set of eligibility criteria. After applying the eligibility criteria, 202 articles containing comparisons of PK parameters of different drugs between pregnant and nonpregnant women were eligible for inclusion. Twenty-six studies were excluded because they didn't report PK parameters, didn't include a comparison group, or were either review papers or case reports (S3 Table). Following review, four further articles were excluded because they duplicated the same outcome domain, in the same cohort, as another article. The remaining 198 articles were included in the data extraction for the comprehensive systematic review. Twenty-two additional articles were identified using a monthly update search between September 1, 2015, and May 20, 2016. Hence, this review article summarizes the results of a total of 198 studies, involving 121 different medications, reporting comparisons of different PK parameters and dosing data between pregnant and nonpregnant cohorts. Reviewed studies were found to vary widely in both design and quality (S4 Table). There were some differences in the stages of pregnancy in which the women were investigated; while most of the studies provided third trimester results, others reported results from both the second and the third trimesters together [38][39][40][41][42]  treatment and vertical transmission prevention, 27 studies focused on antiepileptic drugs, 17 studies focused on drugs related to cardiovascular disorders, and nine studies focused on drugs for endocrine disorders. An additional eight studies investigated antidepressants and anxiolytic drugs, five other studies focused on drugs involved in addiction management, and two studies described drugs treating immunological conditions. In comparison, 84 studies addressed drugs used in the treatment of acute issues during pregnancy; among them, 23 studies addressed antibiotics, 22 studies addressed antimalarial medications, 13 studies addressed analgesics or anesthetic drugs, and eight studies addressed antithrombotic drugs in pregnancy. Fifty-one studies The abstract minimally includes the name of the drug(s) studied, the route of administration, the population in whom it was studied, and the results of the primary objective and major clinical pharmacokinetic findings.

3
Pharmacokinetic data (i.e., absorption, distribution, metabolism, excretion) that [are] known and relevant to the drugs being studied [are] described. 4 An explanation of the study rationale is provided.

5
Specific objectives or hypotheses [are] provided.

6
Eligibility criteria of study participants are described. 7 Co-administration (or lack thereof) of study drug(s) with other potentially interacting drugs or food within this study is described. 8 Drug preparation and administration characteristics including dose, route, formulation, infusion duration (if applicable), and frequency are described.

9
Body fluid or tissue sampling (timing, frequency, and storage) for quantitative drug measurement is described.

10
Validation of quantitative bioanalytical methods used in the study [is] referenced or described if applicable.

11
Pharmacokinetic modeling methods and software used are described, including assumptions made regarding the number of compartments and order of kinetics (zero, first, or mixed order).

12
For population pharmacokinetic studies, covariates incorporated into pharmacokinetic models are identified and described.

13
Formulas for calculated variables (such as creatinine clearance, body surface area, AUC, and adjusted body weight) are provided or referenced.
14 The specific body weight used in drug dosing and pharmacokinetic calculations [is] reported (i.e., ideal body weight versus actual body weight versus adjusted body weight).

15
Statistical methods including software used are described.

16
Study withdrawals or subjects lost to follow-up (or lack thereof) are reported.

17
Quantification of missing or excluded data is provided if applicable. 18 All relevant variables that may explain inter-and intra-patient pharmacokinetic variability (including: age, sex, end-organ function, ethnicity, weight or BMI, health status or severity of illness, and pertinent co-morbidities) are provided with appropriate measures of variance.
investigated more than one drug. Among the antiretroviral class, all studies but one presented women living with HIV infection who were treated with more than one antiretroviral medication. Eleven of 22 studies investigating antimalarial drugs described more than one drug given to the same patient population. Four of 27 studies investigating antiepileptic drugs described more than one drug given to the same patient population. Other drug classes that reported results of pregnant women taking more than one drug included antibiotics (four studies), anesthesia and analgesia drugs (one study), and antiemetics (one study).

Reported Pharmacokinetics Parameters
PK parameters of interest as defined by our search terms were the following: elimination halflife (t 1/2 ), clearance (Cl), C max , C trough , concentration-to-dose ratio (C/D ratio), area under the curve (AUC), volume of distribution (V d ), and protein binding (i.e., free fraction). The majority of the studies reported on various combinations of some of these PK parameters of interest ( Table 4). The most frequently reported PK parameter was Cl, followed by AUC, t 1/2 , and C max with 116, 103, 88, and 87 counts, respectively. In most of the studies that focused on the free fraction of a drug in plasma, the free fraction was the only PK parameter reported in the study.
While Cl and AUC were the most frequently reported parameters, both parameters were reported for only 46% of the drugs. Whereas more than half of the drugs (53%) were described with both the Cl and the t 1/2 , only 16% of the drugs included C trough , C max , and AUC. The latter group mostly consisted of antiretroviral drugs. C max and AUC were described together for 30% of the drugs.

Pharmacokinetics Parameters That Are Vital for Dosing Decision Support in Pregnancy
We clustered the different PK parameters into three groups. (1) Distribution parameters are V d and percent of free fraction. V d defines how widely the drug is spread in the body. Larger V d causes lower peak plasma concentration (C max ) and also longer elimination half-life. Percent free fraction represents the fraction (percent) of the drug in plasma that is unbound to plasma proteins and, therefore, likely to be pharmacologically active. (2) Exposure parameters are C max , C trough , AUC, C/D ratio. These represent indices of plasma drug concentrations. C max and C trough are the highest and lowest levels within a dosing interval, respectively. AUC is literally the area bounded by the drug concentration-time curve and the x-axis, equivalent to an average drug concentration over time. C/D ratio is the dose-standardized drug concentration in plasma or serum at a given time. By and large, these parameters signify drug exposure levels at a given time point or on average, thereby potentially serving as a surrogate for drug effects.
(3) Elimination parameters are t 1/2 and clearance. Half-life is related to the velocity of a drug's disappearance from plasma/serum. Clearance is an index of drug elimination capacity: higher clearance results in a smaller AUC and a shorter elimination half-life, reducing drug exposure levels.
Tables 5-18 provide information regarding changes in PK parameters (weight-standardized values, if available) during pregnancy compared to the nonpregnant state, assorted by drug classes and the data agreement definitions provided above. In these tables, non-significant results are shown together with statistically significant results (in bold). When a certain PK parameter was reported by several studies, the median value and the range in parentheses are provided. The quality column represents the quality score that was assigned to the study, according to the ClinPK Statement checklist. If the drug was investigated in more than one study, the quality column presents the average quality score of all the studies. Among the frequently investigated drug classes (antibiotics, antidepressants, antiepileptics, cardiovascular drugs, antiretrovirals, and antimalarials), studies have demonstrated enhanced elimination together with a decrease in exposure in pregnancy, indicating decreased availability of the drugs in pregnant women compared to nonpregnant women so far as total drug levels (bound plus unbound) are concerned.  Table 19 shows drugs for which all the studies (36) reported no statistically significant PK differences between pregnant and nonpregnant women. Most of the drugs presented in Table 19 were only investigated in one study, while sertraline, propranolol, quinine folic acid and vitamin D3 were each presented in two publications. For sertraline, statistically non-significant decreases in the exposure parameters were reported [70,217]. In the case of propranolol, mean elimination half-life in pregnancy was shorter in both studies, but the exposure parameter (AUC) changes were not consistent; non-significant increase in the AUC [218] versus nonsignificant decrease in AUC [219]. Consistent but non-significant increase in Cl was reported for quinine [189,[220][221][222]. Plasma folate concentrations showed no statistically significant changes [221,222], but conflicting change directions were seen in the mean values, depending on the dose [222]. Similarly, vitamin D3 showed conflicting change directions in exposure parameters, which were statistically non-significant [223,224]. Sixty of the total 218 PK observations (27.5%) reported changes in either the elimination parameters or exposure parameters. Seven PK observations (3.2%) did not report either exposure or elimination parameters. Among the 116 PK observations reporting changes in both elimination and exposure, 79.3% (92) demonstrated increased elimination together with decreased exposure in pregnant women compared to the nonpregnant population.

Discussion
In this first systematic review, to our knowledge, of pregnancy-associated PK changes, we were able to obtain a clear overview of the landscape of the field. Now that trends of pregnancy PK change have been mapped in major drug categories and responsible metabolism or transport pathways, existing knowledge gaps critical for patient management can be addressed by the combined efforts of regulatory agencies, academia, and industry. As many women presently delay childbearing to an older age [243] and the frequency of medical conditions seen during pregnancy among older women is dramatically greater than that of younger women [244], the Pharmacokinetic Changes During Pregnancy results of this review raise the question of whether there are sufficient data to manage these health issues appropriately during pregnancy.
Recently, the most commonly used medications in the first trimester were reported [245]. Results from 5,381 mothers identified 54 different medications used in the first trimester by at least 0.5% of pregnant women. The most commonly used prescription medications reported fell into the categories of antibiotics, analgesics, antiemetics, antidiabetic medications, and antidepressants. Among those 54 most commonly used medications, only a few had adequate data available to assess PK characteristics and dosing recommendation during pregnancy, as demonstrated by our present study results. Although our study strived to identify all available studies describing PK changes occurring in pregnancy, the total number of these studies was relatively small. Widespread exclusion of pregnant women from clinical studies is most probably the major reason for this limitation.
Changes such as increased clearance, reduced half-life, and reduced AUC in pregnancy have been described for many drugs. These PK alterations generally lead to lower drug concentrations in plasma, decreasing maternal target exposure to drug molecules. However, whether these PK changes compromise efficacy is not necessarily certain. Indeed, the total (unbound plus bound fractions) serum concentration of a drug does not necessarily reflect its activity, as lowered plasma albumin concentration during pregnancy may increase free "active" drug concentrations, depending on the PK characteristics of the drug. Moreover, the impact of maternal dose modifications on fetal exposure requires careful planning.
Published data were inconsistent for several medications, preventing this review from defining a certain direction in PK changes. These conflicting results were seen among the antimalarial drugs (pyrimethamine [199,200], sulfadoxine [199,200], and dihydroartemisinin (DHA) [192][193][194]197,198]), antithrombotic drugs (unfractionated heparin [113,114] and low-molecular-weight heparin [46,114-117]), and other drugs (ampicillin [67,68] and doxorubicin [205,216]). We will discuss these drugs in detail in the following section. Also, we confirmed that the current understanding of pregnancy-associated decrease in CYP1A2 and CYP2C19 activities is not based on large studies. These findings require further validation before making clinical recommendations.
For patients who are indicated to undergo routine therapeutic drug monitoring for clinical decision making and dose titration, pregnancy may be a challenging period in which serum drug levels may decrease below the target value despite adequate adherence by patients to their regimen. As we discussed above, decrease in drug exposure levels (e.g., reduction in serum 2nd-3rd Lamivudine Nevirapine [169][170][171] Sulfadoxine [ concentrations and AUC) in pregnancy may not necessarily alter clinical outcomes. The decision to change dosing schedules in patients based on therapeutic drug monitoring and/or knowledge of PK changes in pregnancy should be associated with critical assessment of the risks of therapeutic failure and adverse effects. Fifty-one studies included in our review investigated more than one drug. Among the antiretroviral class, all studies but one presented women with HIV infection who were treated with more than one antiretroviral medication. The only study that examined a single antiretroviral

2nd-3rd
Mefloquine [190][191][192] 3 32/53 Piperaquine [193][194][195]  drug is also the earliest study from this class, investigating zidovudine during pregnancy (published in 1993) [231]. The authors noted that in those 51 studies, no drug that interfered with absorption, elimination, distribution, etc., was included. In addition, as per Health Canada, the US Centers for Disease Control and Prevention, and the World Health Organization, antiretroviral therapy, when indicated, includes at least three agents. Therefore, it is most natural to have multiple drugs on board when conducting a PK study in HIV-positive cohorts.

Clinical Outcome Data
The focus of the present systematic review is on PK data in pregnancy as a first step toward improving drug therapy in this orphan population. Although clinical outcomes were not reported in many of these PK studies, we identified several studies with such information.
For lamotrigine and indinavir, pregnancy-related changes in the clinical endpoints were in agreement with the observed PK changes [88,148] [70,202]. A study investigating cardiovascular drugs that reported clinical outcomes did not demonstrate significant positive clinical correlations [127]. The three drug groups that provided the richest evidence regarding clinical correlation were the antiretrovirals, antimalarials, and antiepileptics. In the case of antiretrovirals, all studies had showed decreased drug exposure in pregnancy due to PK changes. While most of these studies reported  [192][193][194]197,198] [182], others demonstrated a positive correlation between the decreased exposure and poor clinical outcome, reporting an increase in treatment failure or a decrease in post-treatment prophylactic effect [181,195]. Our review has highlighted those medications that have relatively consistent PK change directions in pregnancy. This collection of PK data could prove to be a decision support base for future attempts to tailor medication prescription for pregnant women to achieve target serum concentrations; however, one must take into account that many studies often report undiminished drug efficacy despite the aforementioned pregnancy-associated PK changes [132,135,138,145,146,163,172,177].

Drugs with a Consistent Pharmacokinetics Change Direction
For the vast majority of drugs (114), data gathered in this review are consistent among studies. Although not all studies presented a full set of PK parameters, the evidence exists to support the notion that in pregnancy, drug exposure levels per given dose are decreased for most medications. In addition, lower plasma protein binding (higher free drug level) is a consistent finding. This tandem trending of higher Cl rate, higher V d , and higher free fraction is observed for most drugs except for those metabolized by CYP1A2 and CYP2C19, which show a trend toward decreased metabolism during pregnancy.

Drugs with Variable Pharmacokinetic Change Directions
Studies of seven drugs were found to yield conflicting PK results among studies in pregnancy. Three of these drugs are part of the antimalarial drug group (pyrimethamine [199,200], sulfadoxine [199,200], and DHA [192][193][194]197,198]), two are antithrombotic drugs (unfractionated heparin [113,114] and low-molecular-weight heparin [46,114-117]), one is an antibiotic (ampicillin [67,68]), and the last is an anticancer chemotherapeutic drug (doxorubicin [205,216]). The average quality score of the consistent antibiotic and antithrombotic studies tended to be higher than the quality score of the inconsistent studies from the same group (14.4 versus 11.5, p < 0.05, and 16.4 versus 15.5, p = 0.119, for the antibiotic and antithrombotic drugs, respectively). Nevertheless, the average quality score of the consistent studies was not higher than that of the inconsistent studies for both the antimalarial drugs (18.2 versus 19.1, p = 0.62) and the anticancer chemotherapeutics (11.5 versus 14.5: averages). Thus, variability of quality scores cannot account for the inconsistent PK directions that were demonstrated.
Ampicillin [67,68]. The pregnancy PK of ampicillin had been reported in two studies [67,68]. Both studies presented PK parameters during delivery and demonstrated conflicting results regarding the half-life of elimination. While the elimination half-life presented in one study [67] was longer among pregnant women compared to the control group (58.3 min versus 44.8 min, respectively), the other study [68] demonstrated a difference in the opposite direction (52.4 min versus 69.9 min, respectively). We believe that one of the potential sources for these conflicting results is the choice of control group: while the control group in the former [67] comprised healthy nonpregnant individuals, the post-pregnant women (who may be still under some influence of pregnancy-associated physiological changes) served as their own control in the latter study [68].
Pyrimethamine and sulfadoxine [199,200]. The pregnancy PK of this antimalarial drug combination had been studied in Papua New Guinea [199] and in four African countries (Mozambique, Sudan, Zambia, and Mali) [200]. These two publications present conflicting results. Concerning pyrimethamine, the Papua New Guinea time-concentration plots showed average pregnancy levels to be lower at most time points than the nonpregnant comparison, while data from the African countries indicated the opposite (measurements in pregnancy were higher). This same phenomenon was also evident in some, but not all, data reported on sulfadoxine.
Appraising the methodologies used by these two research groups, we have identified a potential source for this conflict regarding the raw data. In both studies, pregnancy was associated with significant anemia, and both papers (Table 1) reported an average reduction of~20% in hemoglobin values during pregnancy. However, while the Papua New Guinea study used plasma for drug assays, the African study used whole blood from dried blood spots, with no correction for hematocrit values. This limitation of the dried blood spot method may have caused an overestimation of drug levels per blood spot area in pregnant women in the African study, as a result of a relative abundance of plasma per blood spot due to severe anemia [246]. Although there are likely to be other factors contributing to the discrepancies between the two studies, we speculate that the difference in the sample matrix is the major cause, and that pyrimethamine and sulfadoxine apparent clearance is higher during pregnancy. This also highlights the importance of methodological standardization in PK studies, including sample analysis procedures.
Dihydroartemisinin [192][193][194]197,198]. Five studies met the inclusion criteria that investigated the effect of pregnancy on the PK of DHA, the active metabolite of artesunate, for severe malaria (Table 16). Inconsistencies in PK parameter changes exist in the AUC and clearance of DHA; a statistically significant reduction in AUC (decreased exposure) and an increase in oral clearance in pregnancy were observed in one study [197], while the change directions were opposite in the other [198]. However, this can be explained by increased disease severity at PK sampling in the latter [198], as systemic exposure of DHA is higher in infected patients with a severe course of malaria than in those with a mild course [198,247]. The increased DHA exposure in acute malaria during pregnancy after oral artesunate is probably a result of increased bioavailability due to decreased presystemic elimination through glucuronidation in the intestine [198]. Hepatic metabolism of DHA occurs through enzymes such as CYP2B6, UGT1A9, and UGT2B7, but data on these isoenzymes in pregnant women with acute infection are still limited.

Low-molecular-weight heparin [46,114-117] and heparin [113,114].
Six studies investigated the PK of heparin and low-molecular-weight heparin by using factor anti-Xa activity as a surrogate marker of enoxaparin (n = 2), dalteparin (n = 3), and unfractionated heparin (n = 2) in pregnant women (Table 12). The statistically significant discrepancies in the pharmacokinetic parameters can be mainly attributed to the different study designs, dosing regimens, and indications for heparin in the study population (therapeutic versus prophylactic administration). However, the most important parameter in these studies is the C max (2-4 h after administration) of the factor anti-Xa activity because it determines whether the woman is properly controlled for thromboembolic events. Studies with a dose increase design had an increase in the C max of anti-Xa activity [114,116]. The remaining studies revealed lower C max values during pregnancy, even with higher doses [46,113,115,117]. Those studies [46,114,117] showed higher clearance during pregnancy, which was statistically significant in two of them [46,117]. The recommended therapeutic range of 0.6-1.0 IU/ml [248] was achieved in only half of the population in one of the two studies [117]. It should be noted that the Barbour et al.
[116] study compared women in the third trimester to women in early pregnancy (as the control group). Peak levels of anti-Xa activity (equivalent to C max ) were 0.63 IU/ml in early pregnancy versus 0.69 IU/ml in the third trimester. These control values were somewhat higher than the C max values reported for the other nonpregnant populations in the other studies [46,114,115].

Study Limitations
Most studies that demonstrated significant PK changes had relatively small sample sizes. The mixture of small sample sizes with different pharmacological/research methodologies poses substantial challenges to comparing and summarizing their study results. Another limitation stems from the fact that, for many drugs, pregnancy-related PK changes were considered to be significant on the basis of a single study, often of low quality, with small numbers of women and a small subset of PK parameters. Although we show single studies with statistically significant results in the "consistent" category for simplicity of presentation, single studies do not inform on the consistency of the changes. Further replication studies are required. The quality assessment of the studies included in this review was performed using the ClinPK checklist for assessing methodological quality in clinical PK studies. This checklist provides meticulous guidelines for quality assessment, but having been only recently published, it will need refinement and external validation.
We are acutely aware of the fact that by excluding studies lacking a comparison group of nonpregnant women we may miss a significant amount of PK data. However, in the context of our research question, we find it imperative to not only document certain kinetic patterns but also provide quantitative or semiquantitative estimates of the extent and directionality of those pregnancy-associated PK changes. Comparing cohort data for pregnant women to normal population averages would expose our study to a multitude of biases, mainly due to the fact that the most dominant contributors to the "normal population" PK parameter values, in textbooks and seminal papers, are healthy men (Lexicomp and Micromedex databases, for example, report "adult" data with no gender, yet the citation lists are rich with male volunteer publications). Moreover, in the majority of studies included in this systematic review, pregnant women served as their own controls (in the prepregnancy or postpartum state), which isolates the pregnancy as the most dominant factor in the assessment.
Lastly, trimester-specific PK changes were difficult to summarize. While most of the studies provided third trimester results, others reported separate results from the second and third trimesters, and few reported separate results from all trimesters. Physiological changes in pregnancy take place progressively during gestation (reviewed by Costantine [8] and Loebstein et al. [9]). As such, we hypothesized that this would lead to trimester-specific differences in drug disposition. Unfortunately, however, many studies in this review did not report trimesterspecific changes, which could possibly have contributed to the conflicting PK results in some studies described above.

Conclusions
Our systematic analyses confirmed that many drugs are subject to pregnancy-associated PK changes, which may alter plasma/serum drug concentration profiles. However, we have also found a paucity of clinically useful data on whether dose adjustment is necessary for these PK changes. Where such PK studies were done, generally only a few PK parameters were estimated, sample sizes were small, and maternal and/or fetal outcomes were not examined. Further studies that address these limitations are needed to optimize drug therapy for pregnant women.

Author Contributions
Conceptualization: GK GP TL.