Population Pharmacokinetics and Pharmacodynamics of Immunoglobulins: A Systematic Review Authors: S L van der Zeeuw, S J van Tilburg, B C Jacobs, B C P Koch, V A S H Dalm, M B S Crombag, T Preijers Published: 2026 ABSTRACT AbstractBackground and ObjectiveIntravenous and/or subcutaneous immunoglobulins (IVIg/SCIg) are used as immunoglobulin replacement therapy (IgRT) in primary and secondary immunodeficiencies (PID/SID) and as immunomodulatory therapy in immune-mediated diseases. Despite empiric immunoglobulin G (IgG) target concentrations in PID, dosing remains variable and suboptimal, particularly in immune-mediated diseases lacking clear IgG target concentrations. Population pharmacokinetic (popPK) and pharmacodynamic (PD) modelling may clarify interindividual variability and support individualised dosing. Recent developments extend beyond PID, emphasising the importance of integrated PK-PD approaches to optimise IgG therapy. This review aimed to provide a comprehensive overview of popPK(-PD) models describing polyclonal IgG therapy, and to emphasise the added value of PK-PD in understanding drug behaviour.MethodsA systematic literature search was conducted in Embase, MEDLINE, and Web of Science from inception to August 2025. Studies describing popPK(-PD) models for exogenous polyclonal IVIg and/or SCIg administration in humans were included.ResultsIn total, 14 studies were included. While most popPK models focus on PID (N = 10), an increasing number address immune-mediated diseases. Immunoglobulin G pharmacokinetics are typically described using two-compartment models with first-order kinetics, using body weight (BW) as a common covariate for clearance and volume of distribution. Endogenous IgG is often incorporated as a fixed PK parameter; however, this may not adequately capture interindividual variability and potential effects of Ig treatment on IgG concentrations. Two popPK-PD models link IgG concentrations to clinical outcomes in immune-mediated diseases. In contrast, such models are currently lacking for PID. Simulations showed consistent predicted IgG concentrations following SCIg in PID, while IVIg models in PID showed more variability.ConclusionsPopPK models have advanced understanding of IgG PK and its variability between patients and treatment regimens. These models provide a powerful framework to support individualised dosing strategies, thereby potentially reducing reliance on empirical, trial-and-error dosing approaches. However, clear target concentrations are lacking and must be established for individual patients. Models may be further optimised by incorporating essential Neonatal Fc receptor (FcRn) mechanisms using physiology-based pharmacokinetic (PBPK) or semi-mechanistic models. In conclusion, current models are useful to predict IgG concentrations across diseases during Ig treatment; however, a clearer relationship with PD is required. Hence, these findings can guide clinical trial design for tailored dosing regimens.Graphical abstract Supplementary InformationThe online version contains supplementary material available at 10.1007/s40262-026-01641-5. Key Points While current population PK models for immunoglobulin treatment can predict IgG concentration-time profiles with reasonable accuracy, defining clear target concentrations remains challenging due to the heterogeneity of clinical indications and patient populations.Body weight was often used as a covariate, while the specific disease type was only considered in a minority of models.Population PD models are largely lacking, underscoring the need to link IgG levels to the patient outcomes to clarify exposure-response relationships and enable individualised dosing. Introduction Polyclonal immunoglobulins (Ig), either by intravenous (IVIg) or subcutaneous (SCIg) administration, contain mostly (> 98%) immunoglobulin G (IgG) [1]. Facilitated SCIg (fSCIg) uses hyaluronidase to increase the dispersion and absorption of the infused IgG, thereby facilitating infusion of high volumes (up to 600 mL) subcutaneously [2]. Intravenous Ig, SCIg, and fSCIg are human-derived polyclonal plasma products, consisting of pooled plasma from thousands of selected, healthy donors, that provide a broad representation of antibody specificity in the general population [3]. The therapeutic applications are diverse, falling into two main categories. First, it serves as long-term replacement therapy for disorders such as primary immunodeficiencies (PID), which require chronic or lifelong administration [4]. The estimated prevalence of PID ranges from 1:1,000 to 1:5000 individuals [5]. Secondly, high-dose IVIg is employed for its potent anti-inflammatory and immunomodulatory effects in acute immune-mediated conditions, such as Guillain-Barré syndrome (GBS), Kawasaki disease, and immune thrombocytopenia (ITP) [1]. In addition, chronic immune-mediated diseases, such as chronic inflammatory demyelinating polyneuropathy (CIDP) and multifocal motor neuropathy (MMN), require maintenance treatment with high-dose IVIg or SCIg for long-term immunomodulatory effects [1]. The estimated prevalence of immune-mediated conditions ranges from 0.6:100,000 to 9.5:100,000 [6–9]. The mechanisms of action of IVIg and SCIg therapy in immune-mediated diseases are largely attributed to the pleiotropic immunomodulatory properties of polyclonal IgG, which interact with various components of the immune system [10]. One of the mechanisms is the blocking of activating Fc-gamma receptors (FcγR) on macrophages and other immune cells, preventing these cells from destroying antibody-coated (opsonised) targets, such as platelets in ITP [10, 11]. Furthermore, IVIg modulates the complement system by binding to activated components like C3b, thereby interfering with the formation of the membrane attack complex (MAC) and preventing subsequent cell lysis [12, 13]. The polyclonal nature of the therapy also facilitates anti-idiotype interactions, allowing antibodies in the preparation to recognise and neutralise the pathogenic autoantibodies driving immune-mediated diseases [11, 14, 15]. Saturation of the neonatal Fc-receptor (FcRn) with supraphysiological concentrations of IgG after high-dose IVIg administration may lead to reduced recycling of IgG and increased lysosomal degradation of pathogenic (auto)antibodies [16, 17]. In addition, therapeutic immunoglobulins may directly modulate immune cells, influencing the function of B cells, T cells, and dendritic cells by, for example, suppressing antibody production and promoting the expansion of regulatory T cells (Tregs) [18]. Ultimately, the pleiotropic mechanisms of action of IVIg and SCIg regulate the immune response by modulating pro- and anti-inflammatory signals to help restore a balanced immune response [10, 18]. The total IgG concentration is routinely monitored in plasma for primary and secondary immunodeficiencies [19, 20]. Despite being dosed on body weight (BW), the IgG concentrations after Ig treatment can be highly variable [19, 20]. In PID, the empiric therapeutic target concentration ranges from 6.5 to 10.0 g/L [21]. However, despite established IgG target concentrations, dosing based solely on IgG concentrations remains suboptimal, suggesting that other factors, in addition to pharmacokinetic (PK) targets, determine treatment success. Consequently, dosing often remains based on trial-and-error. In immune-mediated diseases, no clear target IgG concentration is defined [22]. However, multiple studies show that there may be a relationship between the PK of IgG and clinical outcome, suggesting that individualised dosing is necessary [23–25]. The need for individualised dosing is further emphasised by the observed variability in IgG PK in GBS and CIDP in previous studies [24, 26, 27]. PopPK analyses allow for obtaining a model that describes the concentration in the body of a specific drug over time [28]. They are a powerful tool to describe and understand differences in IgG concentrations between patients, patient groups and treatment regimens, ultimately offering possibilities to optimise and tailor dosing regimens. Pharmacodynamic (PD) models enable description of the relationship between the concentration of a specific drug and its subsequent physiological effect [29]. Therefore, by integrating PK and PD models, a comprehensive insight into a drug’s behaviour within the body can be obtained. This approach not only supports optimisation of dosing strategies but also provides deeper insight into interindividual variability (IIV) in drug concentrations and response. In recent years, several new population pharmacokinetic (popPK) models for Ig therapy have been published, not only for PID, but also for a broader range of immune-mediated diseases. Moreover, the integration of PD components into PK models for IgG has gained increasing attention. However, in both PID and immune-mediated diseases, IgG dosing remains suboptimal. This review aimed to provide a complete and comprehensive overview of polyclonal IgG therapy, highlighting recent modelling developments expanding the scope beyond PID and emphasising the added value of PK-PD in understanding drug behaviour. Methods Search Strategy This systematic review was performed following the 2020 PRISMA guidelines [30]. A literature search was performed in Embase, MEDLINE and Web of Science on the 6th of August 2025. The search strategies consisted of search terms including “Immunoglobulin”, “Pharmacokinetics” and “Pharmacodynamics”. Full search strategies can be found in Supplementary Table 1. Inclusion and Exclusion Criteria This review included papers that report on an original or updated popPK and/or population pharmacodynamic (popPD) model of polyclonal immunoglobulins (IVIg and/or SCIg) in humans. Publications were excluded if they discussed no exogenous IgG administration, non-IVIg or SCIg, previously published popPK and/or popPD models without new data or models other than popPK and/or popPD models (e.g., physiology-based pharmacokinetic [PBPK] models). Only articles published in English were considered for inclusion. Study Selection First, titles and abstracts were screened by two reviewers (ST, TP). Second, a full-text review of the remaining publications was performed by two reviewers (SZ, ST). Conflicts were discussed among all reviewers before deciding on the final selection. Data Quality The quality of the included records was assessed by two reviewers (SZ, ST) using the clinical PK statement checklist [31]. An overview of the reported information in different sections of the articles was made using the following checklist. Data Collection Data collection from the included studies was performed by two reviewers (SZ, ST). The following data items were extracted from the included publications: patient characteristics, treatment characteristics (including dosages and dosing intervals), details of the sampling strategy, full details of the PK model (population parameters, IIV, and residual variability) and, if present, full details of the PD model. If data from the studies were unavailable, it was reported as “not reported”. For comparison, all PK parameters mentioned throughout this review were scaled to 70 kg BW as the reference, unless otherwise stated. If lean body mass (LBM) was incorporated in a model, relevant PK parameters were scaled to LBM corresponding to 70 kg BW (male, 170 cm) using the Boer formula [32]. In addition, all IIV values were converted to represent coefficient of variation percentages (CV%) using the following equation: e2-1×100%. All PK parameter estimates were converted to represent days by multiplying the estimates by 24 when originally reported in hours. Pharmacokinetic estimates originally reported in mL were converted to L by dividing the values by 1000. PK Simulations To provide an overview of the comparability of the IgG popPK models discussed in this review, simulations were performed using published estimates of the typical PK parameters. In this simulation, all virtual patients (male, 70 kg and 170 cm) received the same standard dose dependent on the clinical indication; PID IV: 400 mg/kg loading dose, followed by 400 mg/kg maintenance dose once every four weeks, PID SC: 100 mg/kg loading and maintenance dose weekly, CIDP and MMN IV: 2000 mg/kg over five days loading dose, followed by 1000 mg/kg maintenance dose once every three weeks, CIDP and MMN SC: 2000 mg/kg over five days loading dose, followed by 1000 mg/kg maintenance once every three weeks, GBS: IV 2000 mg/kg for five days loading dose. Lean body mass was calculated using the Boer formula and the same patient characteristics [32]. Endogenous IgG concentrations (i.e., endogenously produced IgG concentrations) were assumed to be 10 g/L for CIDP, MMN, and GBS, and 4.0 g/L for PID. The simulations were performed in R (version 4.3.2) utilising the ‘deSolve’ and ‘MASS’ packages [33, 34]. Results In total, 684 publications from three databases (n = 223 MEDLINE, n = 354 Embase, n = 107 Web of Science Core Collection) were identified in the initial search. After removing duplicates, 485 publications were screened. After applying all inclusion and exclusion criteria, 14 publications were selected for this review. Data Quality Information regarding sampling strategies was presented in approximately 40% of the included studies [27, 35–39]. For the other studies, sampling strategies were retrieved from the original studies, which collected the PK data. In 30% of the included studies, exclusion of patients and/or samples including quantification was presented [40–43]. Two of the included studies did not describe which software was used for the modelling [40, 44]. Co-administration of IVIg/SCIg with other drugs was described in one study [27]. Data on the background of the study and the discussion and conclusions were included in almost all of the studies. Pharmacokinetic Modelling of IVIg and SCIg Primary and Secondary Immunodeficiencies Primary immunodeficiency is an overarching term for congenital disorders characterised by impaired immune system function [45]. Predominant antibody deficiency (PAD) is the most common type of PID, which requires immunoglobulin replacement therapy (IgRT) due to altered antibody production [46]. Immunoglobulin replacement therapy doses typically range from 400 to 800 mg/kg once every 4 weeks [1]. In contrast, SCIg is generally administered weekly, with administered cumulative doses equivalent to IVIg [47]. Both PID and SID are characterised by reduced IgG concentrations necessitating replacement therapy. Since PIDs typically stem from genetic disorders, replacement therapy is often required lifelong. Endogenous IgG production varies greatly depending on the underlying PID. For example, patients with X-linked agammaglobulinemia (XLA) lack mature B-cells and have extremely low or absent endogenous immunoglobulins [48]. In comparison, endogenous IgG concentrations in patients with common variable immunodeficiency (CVID) are generally lower than 4.5 g/L, as opposed to a reference of more than 6 g/L [21, 49]. Whereas PID arises from congenital defects of the immune system, secondary immunodeficiencies (SID) may be acquired by diverse, non-hereditary factors, such as haematological malignancies and use of immunosuppressants [50, 51]. Secondary immunodeficiency is common in patients with multiple myeloma. A large retrospective cohort study found that 20% of patients with multiple myeloma developed SID [52]. General Information on the Included Models Of the 14 studies included in this systematic review, ten studies constructed a popPK model for patients receiving IgRT (Table 1). Seven out of ten studies constructed a model on patients with PID only [36, 37, 40–42, 44, 53], one out of ten on PID and SID combined [54], and one out of ten studies on very low birth-weight neonates [35] and one out of ten on numerous indications, including PID [38]. In four out of ten studies, only IVIg data were incorporated (Table 2) [35, 36, 38, 54]. In five out of ten studies, both IVIg and SCIg data were included [37, 40, 41, 44, 53]. One out of ten studies included data on IVIg, SCIg and fSCIg [42]. Intravenous immunoglobulin was administered once every three or four weeks and SCIg weekly or once every 2 weeks. All studies included data from at least one study performed in children. In three out of ten studies, data were incorporated from studies performed in Asia [36, 38, 41]. Eight out of ten studies included data from studies performed in the USA [35, 37, 40–42, 44, 53, 54], six out of ten from studies performed in Europe [40–42, 44, 53, 54] and five out of ten studies were performed in Canada [37, 41, 42, 44, 53]. Table 1Demographic characteristics of the included modelsAuthor, yearCountryIncluded clinical trialsSample sizePopulationGender M/F, N (%)Age, year median (min, max)Weight, kg median (min, max)IndicationIgG baseline, g/L, median (min, max)Primary immunodeficiency and secondary immunodeficiencyLandersdorfer et al. (2013) [40]USAEUNCT00168025NCT00322556NCT00419341NCT00542997N += 151NCT00168025: 46/34 (57.5%/42.5%)NCT00322556: 26/29 (52.7%/47.3%)NCT00419341: 22/27 (55.1%/44.9%)NCT00542997: 35/16 (68.6%/31.4%)NCT00168025: 25.0 (3, 69)NCT00322556: 23.0 (4, 81)NCT00419341: 32.0 (5, 72)NCT00542997: 18.0 (3, 60)NCT00168025: 66.5 (14.0, 130.0)NCT00322556: 62.0 (18.0, 135.0)NCT00419341: 66.0 (21.0, 104.00)NCT00542997: 53.5 (13.0, 96.0)PIDNRDumas et al. (2019) [53]USACAEUNCT01218438NCT01412385N = 10257/45 (55.9%/44.1%)30.0 (2.0, 83)63.7 (13.2, 161.8)PID9.9 (3.4, 19.9)Tortorici et al. (2019) [54]USAEUNCT00168025NCT00322556NIS-Nr: 182PID: N = 90SID: N = 97Total = 187PID: 49/41 (54.4%/45.6%)SID: 58/39 (59.8%/40.2%)PID mean (SD): 29.8 (20.3)SID mean (SD): 69.5 (10.4)PID mean (SD): 62.6 (26.4)SID mean (SD): 76.8 (16.1)PID, SIDPID: 9.17 (3.93, 27.2, N = 77)SID: 6.15 (2.05, 17.1, N = 74)Luo et al. (2020) [41]USACAEUAsiaNCT00419341NCT00168025NCT00322556NCT00542997NCT02711228,2016–001631–12NCT01199705NCT01458171N = 202117/85 (57.9%/42.1%)21 (3, 81)58.7 (13, 135)PIDNRTegenge et al. (2020) [35]USAhttps://doi.org/10.1016/s0022-3476(88)80070-2N = 2012/8 (60%/40%)1–6 daysGA: 26–35 weeksBW at birth: 0.78–1.38 kgVery low BW neonatesNRZhang et al. (2020) [44]USACAEUNCT00419341NCT00168025NCT00322556NCT00542997NCT02711228N = 173NCT00419341: 22/27 (44.9%/55.1%)NCT00168025: 46/34 (57.5%/42.5%)NCT00322556: 26/39 (47.3%/52.7%)NCT00542997: 35/16 (68.6%/31.4%)NCT02711228: 8/9 (47.1%/52.9%)NCT00419341: 32.0 (5.0, 72.0)NCT00168025: 25.0 (3.0, 69.0)NCT00322556: 23.0 (4.0, 81.0)NCT00542997:18.0 (3.0, 60.0)NCT02711228: 19.0 (14.0, 66.0)NCT00419341: 66.0 (21.0, 104.0)NCT00168025: 66.5 (14.0, 130.0)NCT00322556: 62.0 (18.0, 135.0)NCT00542997:54.0 (13.0, 96.0)NCT02711228: 72.5 (50.6, 96.0)PIDNRLee et al. (2021) [36]AsiaProspective observational studyN = 109/1 (90%/10%)9.5 (3, 64)26.7 (9.3, 75)PID0.7 (0.3-5.09)aLi et al. (2022) [42]USACAEUNCT00814320NCT01412385NCT01218438NCT00161993NCT00157079NCT00546871NCT00782106NCT03277313N = 340179/161 (53.2%/46.8%)31.5 (2.0-83.0)66.0 (11.9–162)PIDNRNavarro-Mora et al. (2022) [37]USACANCT00389324NCT01465958NCT02604810N = 9540/55 (42.1%/57.9%)NCT00389324: 42.5 (15.8) meanNCT01465958: 10.8 (3.7) meanNCT02604810: 36.8 (21.36) mean65.7 (16.7, 153.0)PIDNRLee et al. (2024) [38]AsiaProspective observational studyN = 7943/36 (54%/46%)15.0 (0.08–70.0)43.0 (3.61–98.5)PID, Kawasaki disease, GBS, thrombocytopenia, myasthenic crisis, lupus9.08 (0.61–33.03)Immune-mediated diseasesTortorici et al. (2021) [43]USACAEUAsiaAUSNCT01545076https://doi.org/10.1111/jns5.12017N = 235149/86 (63.4%/36.3%)58 (22, 83)82 (42.3, 133)CIDP12.5 (5.6-33.0)aFokkink et al. (2022) [27]EUhttps://doi.org/10.1056/NEJM199204233261705https://doi.org/10.1016/s0140-6736(03)15324-xN = 354Model building 101/76 (57%/43%)Validation 112/65 (63%/37%)Model building: mean (range): 53 (5–89)Validation: mean (range): 55 (18–86)Model building: mean (range): 72 (18–122)Validation: mean (range): 79 (52.6–118)GBSNRLi et al. (2024) (Front Neurol.) [39]USACAEUNCT00666263N = 4432/12 (72.7%/27.3%)52.0 (31.0, 72.0)83.5 (56.3, 107.0)MMN20.2 (11.6, 37.0)aLi et al. (2024) (Ann Clin Transl Neurol.) [23]USACAEUNCT00666263N = 4432/12 (72.7%/27.3%)52.0 (31.0, 72.0)83.5 (56.3, 107.0)MMN20.2 (11.6, 37.0)aAUS Australia, CA Canada, CIDP chronic inflammatory demyelinating polyneuropathy, EU Europe, GBS Guillain-Barré syndrome, IgG immunoglobulin G, MMN multifocal motor neuropathy, NR not reported, PID primary immunodeficiency, SID secondary immunodeficiencyaTreatment-naïve/endogenous IgG concentrations Demographic characteristics of the included models USA EU NCT00168025 NCT00322556 NCT00419341 NCT00542997 NCT00168025: 46/34 (57.5%/42.5%) NCT00322556: 26/29 (52.7%/47.3%) NCT00419341: 22/27 (55.1%/44.9%) NCT00542997: 35/16 (68.6%/31.4%) NCT00168025: 25.0 (3, 69) NCT00322556: 23.0 (4, 81) NCT00419341: 32.0 (5, 72) NCT00542997: 18.0 (3, 60) NCT00168025: 66.5 (14.0, 130.0) NCT00322556: 62.0 (18.0, 135.0) NCT00419341: 66.0 (21.0, 104.00) NCT00542997: 53.5 (13.0, 96.0) USA CA EU NCT01218438 NCT01412385 USA EU NCT00168025 NCT00322556 NIS-Nr: 182 PID: N = 90 SID: N = 97 Total = 187 PID: 49/41 (54.4%/45.6%) SID: 58/39 (59.8%/40.2%) PID mean (SD): 29.8 (20.3) SID mean (SD): 69.5 (10.4) PID mean (SD): 62.6 (26.4) SID mean (SD): 76.8 (16.1) PID: 9.17 (3.93, 27.2, N = 77) SID: 6.15 (2.05, 17.1, N = 74) USA CA EU Asia NCT00419341 NCT00168025 NCT00322556 NCT00542997 NCT02711228,2016–001631–12 NCT01199705 NCT01458171 1–6 days GA: 26–35 weeks USA CA EU NCT00419341 NCT00168025 NCT00322556 NCT00542997 NCT02711228 NCT00419341: 22/27 (44.9%/55.1%) NCT00168025: 46/34 (57.5%/42.5%) NCT00322556: 26/39 (47.3%/52.7%) NCT00542997: 35/16 (68.6%/31.4%) NCT02711228: 8/9 (47.1%/52.9%) NCT00419341: 32.0 (5.0, 72.0) NCT00168025: 25.0 (3.0, 69.0) NCT00322556: 23.0 (4.0, 81.0) NCT00542997:18.0 (3.0, 60.0) NCT02711228: 19.0 (14.0, 66.0) NCT00419341: 66.0 (21.0, 104.0) NCT00168025: 66.5 (14.0, 130.0) NCT00322556: 62.0 (18.0, 135.0) NCT00542997:54.0 (13.0, 96.0) NCT02711228: 72.5 (50.6, 96.0) USA CA EU NCT00814320 NCT01412385 NCT01218438 NCT00161993 NCT00157079 NCT00546871 NCT00782106 NCT03277313 USA CA NCT00389324 NCT01465958 NCT02604810 NCT00389324: 42.5 (15.8) mean NCT01465958: 10.8 (3.7) mean NCT02604810: 36.8 (21.36) mean USA CA EU Asia AUS NCT01545076 https://doi.org/10.1111/jns5.12017 https://doi.org/10.1056/NEJM199204233261705 https://doi.org/10.1016/s0140-6736(03)15324-x Model building 101/76 (57%/43%) Validation 112/65 (63%/37%) Model building: mean (range): 53 (5–89) Validation: mean (range): 55 (18–86) Model building: mean (range): 72 (18–122) Validation: mean (range): 79 (52.6–118) USA CA EU USA CA EU AUS Australia, CA Canada, CIDP chronic inflammatory demyelinating polyneuropathy, EU Europe, GBS Guillain-Barré syndrome, IgG immunoglobulin G, MMN multifocal motor neuropathy, NR not reported, PID primary immunodeficiency, SID secondary immunodeficiency aTreatment-naïve/endogenous IgG concentrations Table 2Dosing and sampling information of the included modelsAuthor, yearFormulationDoseNumber of samplesTime period samplesMeasurement techniquePrimary immunodeficiency and secondary immunodeficiencyLandersdorfer et al. (2013) [40]IVIg, SCIgNCT00168025: IVIg 200–888 mg/kg once every 3 to 4 weeksNCT00322556: IVIg median 460.1 mg/kg (range 13.3, 875.0) once every 3 or 4 weeksNCT00419341: SCIg 54–406 mg/kg QWNCT00542997: SCIg 72–262 mg/kg QWN = 3837235–1446 daysImmunoturbidimetric analysisDumas et al. (2019) [53]IVIg, SCIgIVIg: 300–1000 mg/kg Q4WSCIg: QW equivalent dosesModel building: N = 1657 (IVIg: 601, SCIg: 1056)Validation: N = 406 (IVIg: 160, SCIg: 246)Total: N = 2063 (IVIg: 761, SCIg: 1302)42–476 daysELISATortorici et al. (2019) [54]IVIgPID: 13.3–959.0 mg/kg once Q4WSID: 90.9–678.0 mg/kg once Q4WN = 257421–364 daysNRLuo et al. (2020) [41]IVIg, SCIgIVIg: 13.3-913 mg/kg once every 3 or 4 weeksSCIg: 26.7–379 mg/kg QW or once Q2WN = 45020-364 daysNCT00419341: Immunoturbidimetric analysisTegenge et al. (2020) [35]IVIg500 or 750 mg/kgExtensive (8 samples per patient) and sparse (2 or 3 samples per patient)0–28 daysRadial immunodiffusion using a commercial kitZhang et al. (2020) [44]IVIg, SCIgNCT00168025: IVIg 200–888 mg/kg once every 3 or 4 weeksNCT00419341: SCIg mean 179.6–224.3 mg/kg QWNCT00322556: IVIg 200–888 mg/kg once every 3 or 4 weeksNCT00542997: SCIg 117.0–120.7 mg/kg per weekNCT02711228: SCIg QW or once Q2WN = 407821–364 daysNRLee et al. (2021) [36]IVIg360–600 mg/kg once every three or four weeksN = 300–28 daysELISALi et al. (2022) [42]IVIg, SCIg, fSCIgNCT00814320: IVIg > 300 mg/kg/3–4 weeks IVIg: Q3W/Q4W, SCIg: QW/Q4W (fSCIg)NCT01412385: IVIg 300–1000 mg/kg once every 3 or 4 weeks, SCIg QW equivalent dose IVIgNCT01218438: IVIg 300–1000 mg.kg Q4W, SCIg 20% 145% of IVIg dose QWNCT00161993: NRNCT00157079: NRNCT00546871: IVIg 300–1000 mg/kg Q4W and SCIg 130%–137% of IVIg dose QW NCT00782106: NRNCT03277313: IVIg 300–1000 mg/kg Q4W and fSCIg every 3–4 weeksN = 5094NRNCT01412385: ELISANCT01218438: ELISANavarro-Mora et al. (2022) [37]IVIg, SCIgIVIg: 495 mg/kg (median, range 278–902) once every 3 or 4 weeksSCIg: 184.8 mg/kg (median, 72.0–303.5) once every 3 or 4 weeksN = 190619–340 daysNRLee et al. (2024) [38]IVIgPID: 360–600 mg/kg once every 3 or 4 weeksKawasaki: 2000 mg/kg in a single infusionGBS: 2000 mg/kg over 2–5 daysThrombocytopenia paediatric: 800–1000 mg/kg as a single infusionThrombocytopenia adult: 2000 mg/kg over 2–5 daysMyasthenic crisis: 2000 mg/kg over 5 daysLupus: 2000 mg/kg over 5 daysN = 2928–47 daysELISAImmune-mediated diseasesTortorici et al. (2021) [43]IVIg, SCIgIVIG: induction 2000 mg/kg over 2–5 days, followed by maintenance 1000 mg/kg Q3WSCIg: 200 or 400 mg/kgN = 15580–266 daysNRFokkink et al. (2022) [27]IVIg400 mg/kg/day for 5 daysModel-building: N = 589Final validation: N = 6890–182 daysImmunoturbidimetric analysisLi et al. (2024) (Front Neurol.) [39]IVIg400–2000 mg/kg once every 2–4 weeksN = 3090–420 daysNRLi et al. (2024) (Ann Clin Transl Neurol.) [23]IVIg400–2000 mg/kg once every 2–4 weeksN = 3090–420 daysNRELISA Enzyme-Linked Immunosorbent Assay, (f)SCIg (facilitated) SC immunoglobulin G, GBS Guillain-Barré syndrome, IVIg intravenous immunoglobulin G, NR not reported, PID primary immunodeficiency, Q2W every 2 weeks, Q3W every 3 weeks, Q4W every 4 weeks, SC subcutaneous, SID secondary immunodeficiency Dosing and sampling information of the included models NCT00168025: IVIg 200–888 mg/kg once every 3 to 4 weeks NCT00322556: IVIg median 460.1 mg/kg (range 13.3, 875.0) once every 3 or 4 weeks NCT00419341: SCIg 54–406 mg/kg QW NCT00542997: SCIg 72–262 mg/kg QW IVIg: 300–1000 mg/kg Q4W SCIg: QW equivalent doses Model building: N = 1657 (IVIg: 601, SCIg: 1056) Validation: N = 406 (IVIg: 160, SCIg: 246) Total: N = 2063 (IVIg: 761, SCIg: 1302) PID: 13.3–959.0 mg/kg once Q4W SID: 90.9–678.0 mg/kg once Q4W IVIg: 13.3-913 mg/kg once every 3 or 4 weeks SCIg: 26.7–379 mg/kg QW or once Q2W NCT00168025: IVIg 200–888 mg/kg once every 3 or 4 weeks NCT00419341: SCIg mean 179.6–224.3 mg/kg QW NCT00322556: IVIg 200–888 mg/kg once every 3 or 4 weeks NCT00542997: SCIg 117.0–120.7 mg/kg per week NCT02711228: SCIg QW or once Q2W NCT00814320: IVIg > 300 mg/kg/3–4 weeks IVIg: Q3W/Q4W, SCIg: QW/Q4W (fSCIg) NCT01412385: IVIg 300–1000 mg/kg once every 3 or 4 weeks, SCIg QW equivalent dose IVIg NCT01218438: IVIg 300–1000 mg.kg Q4W, SCIg 20% 145% of IVIg dose QW NCT00161993: NR NCT00157079: NR NCT00546871: IVIg 300–1000 mg/kg Q4W and SCIg 130%–137% of IVIg dose QW NCT00782106: NR NCT03277313: IVIg 300–1000 mg/kg Q4W and fSCIg every 3–4 weeks NCT01412385: ELISA NCT01218438: ELISA IVIg: 495 mg/kg (median, range 278–902) once every 3 or 4 weeks SCIg: 184.8 mg/kg (median, 72.0–303.5) once every 3 or 4 weeks PID: 360–600 mg/kg once every 3 or 4 weeks Kawasaki: 2000 mg/kg in a single infusion GBS: 2000 mg/kg over 2–5 days Thrombocytopenia paediatric: 800–1000 mg/kg as a single infusion Thrombocytopenia adult: 2000 mg/kg over 2–5 days Myasthenic crisis: 2000 mg/kg over 5 days Lupus: 2000 mg/kg over 5 days IVIG: induction 2000 mg/kg over 2–5 days, followed by maintenance 1000 mg/kg Q3W SCIg: 200 or 400 mg/kg Model-building: N = 589 Final validation: N = 689 ELISA Enzyme-Linked Immunosorbent Assay, (f)SCIg (facilitated) SC immunoglobulin G, GBS Guillain-Barré syndrome, IVIg intravenous immunoglobulin G, NR not reported, PID primary immunodeficiency, Q2W every 2 weeks, Q3W every 3 weeks, Q4W every 4 weeks, SC subcutaneous, SID secondary immunodeficiency Absorption and Bioavailability All models using SCIg data included a first-order absorption rate constant (Ka), with no significant covariate relationships (Tables 3 and 4) [37, 40–42, 44, 53]. Reported bioavailability estimates for SCIg ranged from 66.0 to 73.9%. Of the six popPK studies that included data on SCIg administration, only Li et al. (2022) included product type (SCIg products with or without hyaluronidase) as a covariate on bioavailability, as it is the only study to have included fSCIg data in their popPK analysis (14.5% of included patients) [42]. The relative bioavailability of SCIg was 70.5%, and 79.4% for fSCIg, described as the difference from the bioavailability of IVIg (assumed to be 1), suggesting that fSCIg has a higher bioavailability than unfacilitated SCIg. Although the exact value (CV%) was not reported, IIV was included on bioavailability in one model only [53]. Table 3Model structure, estimation method, covariates and model evaluationAuthor, yearSoftwareModelEstimation methodCovariate selection methodCovariates testedCovariates in final modelModel evaluationPrimary immunodeficiency and secondary immunodeficiencyLandersdorfer et al. (2013) [40]NRTwo-compartmentNRNRBWBW on CL and VcGoodness-of-fit plots, bootstrap (N = NR), VPC, shrinkageDumas et al. (2019) [53]NONMEMOne-compartmentFOCE+IForward inclusion (p < 0.01) and backward elimination (p < 0.005)Age, BW, geographic region, race, sexBW on CLOFV, AIC, goodness-of-fit parameters and goodness-of-fit plotsTortorici et al. (2019) [54]NONMEMTwo-compartmentFOCE+IStepwise backward eliminationBW, age, gender, disease type (PID, SID)BW on CL and Vc, Disease type on VcOFV improvement by 7.78 points, clinical relevance, goodness-of-fit plots, plausibility of parameter estimates, bootstrap (N = 1000)Luo et al. (2020) [41]NONMEMTwo-compartmentFOCE+IForward inclusion (p = NR) and backward elimination (p = NR)BW, ethnicity (Japanese/non-Japanese), age, sexBW on CL and VcGoodness-of-fit plots, PCVPC, bootstrap resampling (N = 500)Tegenge et al. (2020) [35]NONMEMTwo-compartmentFOCE+INRBW-Goodness-of-fit plots, VPC, difference in OFVZhang et al. (2020) [44]NRTwo-compartmentNRNRBWBW on CL and VcGoodness-of-fit plots, scatterplot conditional weighted residuals vs predicted, bootstrap (N = 500), VPCLee et al. (2021) [36]MonolixOne-compartmentSAEMForward inclusion (p < 0.05) and backward elimination (p < 0.001)Age, BW, ethnicity, gender, presence of bronchiectasis, genotypeBW on CL and VcGoodness-of-fit plots, PC VPC, bootstrap (N = 1000)Li et al. (2022) [42]NONMEMTwo-compartmentFOCE+IForward inclusion (< 0.01) and backward elimination (< 0.001)BW, BMI, LBM, sex, age, IgG product, hyaluronidase product yes/noProduct on F1, LBM on CL, Q, Vc, VpBootstrap (N = 1000), confidence intervals model parameters (95% CI), VPCNavarro-Mora et al. (2022) [37]NONMEMTwo-compartmentNRBackward elimination (retaining COV is OFV increase > 10.84 points, or if there is a strong physiological rationale for retainment if OFV increase is not met)Age, BW, sex, SCIg formulationBW on CL, Vc, and VpGoodness-of-fit plots, visual predictive check, bootstrap (N = 1000)Lee et al. (2024) [38]MonolixTwo-compartmentSAEMStepwise forward inclusion (< 0.05) and backwards elimination (< 0.001)Age, BW, baseline IgG level, ethnicity, sex, genotype, disease type, comorbidityBW on CL, Vc and Vp, disease type on Vp (PID as reference)Goodness-of-fit plots, acceptable relative SE (RSE) of <30% for fixed-effect estimates and <50% for random-effect estimates, pcVPC, bootstrap (N = 1000)Immune-mediated diseasesTortorici et al. (2021) [43]NONMEMTwo-compartmentNRWAMBW, age, sex, baseline IgG, IgG treatment-naïve/pretreated, Japanese/non-Japanese, US/non-US regionsBW on CL, Q, Vc, and VpGoodness-of-fit plots, pcVPCFokkink et al. (2022) [27]NONMEMTwo-compartmentFOCE+IForward inclusion (< 0.05) and backward elimination (< 0.01)GBS-DS MRC-SS, MP, dose, preceding diarrhea, sex, age, mechanical ventilation, VNTRMP on CL, GBS-DS on CLOFV, realistic parameter estimates, shrinkage (<20%), conditional number (<1000), goodness-of-fit plots, external evaluation, VPC, NPDELi et al. (2024) (Front Neurol.) [39]NONMEMOne-compartmentNRNRAge, sex, BW, BMI, LBM, creatinine clearanceLBM on VcGOF, VPCLi et al. (2024) (Ann Clin Transl Neurol.) [23]NONMEMOne-compartmentNRNRAge, sex, BW, BMI, LBM, creatinine clearanceLBM on VcGOF, VPCAIC Akaike Information Criterion, BMI body mass index, BW body weight, CL clearance, FOCE+I first-order conditional estimation with interaction, GBS-DS Guillain-Barré Syndrome disability score, LBM lean body mass, MP methylprednisolone, MRC-SS Medical council sum score, NR not reported, OFV objective function value, (pc)VPC (prediction corrected) visual predictive check, PID primary immunodeficiency, Q intercompartmental clearance, SAEM stochastic approximation expectation-maximisation, SID secondary immunodeficiency, Vc volume of distribution of the central compartment, VNTR variable number of tandem repeats, Vp volume of distribution of the peripheral compartment, WAM Wald’s Approximation Method. Model structure, estimation method, covariates and model evaluation AIC Akaike Information Criterion, BMI body mass index, BW body weight, CL clearance, FOCE+I first-order conditional estimation with interaction, GBS-DS Guillain-Barré Syndrome disability score, LBM lean body mass, MP methylprednisolone, MRC-SS Medical council sum score, NR not reported, OFV objective function value, (pc)VPC (prediction corrected) visual predictive check, PID primary immunodeficiency, Q intercompartmental clearance, SAEM stochastic approximation expectation-maximisation, SID secondary immunodeficiency, Vc volume of distribution of the central compartment, VNTR variable number of tandem repeats, Vp volume of distribution of the peripheral compartment, WAM Wald’s Approximation Method. Table 4Pharmacokinetic parameter estimates and random effects modelAuthor, yearFixed-effectsRandom-effectsInter-individual variability (CV%)Residual unexplained variabilityCL (L/day)V (L)F1Ka (L/day)RelationshipCLVF1KaAdd (g/L)PropPrimary immunodeficiency and secondary immunodeficiencyLandersdorfer et al. (2013) [40]CL = 0.142Q = 0.252Vc = 3.94Vp = 4.1866.0%0.439NRNRNRNRNRNRNRDumas et al. (2019) [53]CL = 0.09216b (BW/70 kg)^0.576Vc = 4.0173.9%0.096ExponentialCL = 27.6Vc = 39.2–––5.3%Tortorici et al. (2019) [54]CL = 0.152b (BW/72)^0.796Q = 0.825Vc PID = 3.08b (BW/72)^1.1Vc SID = 8.75b (BW/72)^1.1Vp = 1.8––ExponentialCL = 53.1Vc = 75.4––0.9307.1%Luo et al. (2020) [41]CL = 0.139b (BW/58.7)^0.881Q = 0.300Vc = 4.01b (BW/58.7)^0.501Vp = 3.5166.8%0.506 g/dayExponentialCL = 36.02Q = 54.25Vc = 92.08Vp = 325.424.0130.57UnclearbUnclearbTegenge et al. (2020) [35]CL = 0.0027Q = 0.045Vc = 0.008Vp = 0.055––ExponentialCL = 52Vc = 22Vp = 29–––9%Zhang et al. (2020) [44]CL = 0.138b (BW/66)^0.768Q = 0.260Vc = 3.95b (BW/66)^0.448Vp = 4.4467.6%0.444ExponentialCL = 36.02Q = 61.52Vc = 92.58Vp = 252.7732.3654.85σ2 = 0.00632σ2 = 0.0101Lee et al. (2021) [36]CL = 0.0624b (BW/27 kg)^0.88Vc = 2.77b (weight/27 kg)^0.66––ExponentialCL = 13.97Vc = 8.00–––10.23%Li et al. (2022) [42]CL = 0.183b (LBM/47)^0.75Q = 0.353b (LBM/47)^0.75Vc = 3.01b (LBM/47)^1Vp = 1.40b (LBM/47)^1F1 without hyaluronidase = 70.5%F1 with hyaluronidase = 79.4%0.395 day−1ExponentialCL = 45.9Vc = 20.4––σ2 = 0.962cσ2 = 0.108cNavarro-Mora et al. (2022) [37]CL = 0.150b (BW/65.7)^0.744Q = 0.474Vc = 3.06b (BW/65.7)^0.686Vp = 1.93b (BW/65.7)^1.0470.5%0.246ExponentialCL = 28.9Vc = 19.3–82.3σ2 = 0.668σ2 = 0.00281Lee et al. (2024) [38]CL = 0.036b (BW/43)^0.75Q = 0.317Vc = 1.20b (BW/43)^1Vp disease type 2a = 0.52b (BW/43)^1b exp(1.67)Vp disease type 3a = 0.52b (BW/43)^1b exp(1.65)Vp disease type 4a = 0.52b (BW/43)^1b exp(1.46)Vp disease type 5a = 0.52b (BW/43)^1b exp(1.93)Vp disease type 6a = 0.52b (BW/43)^1 exp(1.97)Vp disease type 7a = 0.52b (BW/43)^1b exp(2.16)––ExponentialCL = 28.35Q = 67.27Vc = 45.41Vp = 58.19–––10.73%Immune-mediated diseasesTortorici et al. (2021) [43]CL = 0.435b (BW/82)^0.615Q = 0.50b (BW/82)^0.615Vc = 4.69b (BW/82)^0.773Vp = 1.87b (BW/82)^0.7730.8240.439 day−1ExponentialCL = 28.0Vc = 23.0–––12.1%Fokkink et al. (2022) [27]CL = 0.28b (BW/70)^0.75b 1.31 (if MP)b 1.22 (if GBS-DS = 3 or 4)b 1.83 (if GBS-DS = 5)Q = 0.25Vc = 2.87Vp = 2.65––Exponential49.9––––σ = 0.12Li et al. (2024) (Front Neurol.) [39]Kel = 0.05784Vc = 6.59b (LBM/56.54)^2.23––ExponentialNR18.0––σ2 = 0.00905Li et al. (2024) (Ann Clin Transl Neurol.) [23]Kel = 0.05832Vc = 6.48b (LBM/56.54)^2.17––ExponentialNR18.5––σ2 = 0.00905Add additive, BW body weight, CL clearance, F1 bioavailability, GBS-DS Guillain-Barré Syndrome disability score, Ka absorption rate constant, LBM lean body mass, MP methylprednisolone, NR not reported, PID primary immunodeficiency, Prop proportional, Q intercompartmental clearance, SID secondary immunodeficiency, V volume of distribution, Vc volume of distribution of the central compartment, Vp volume of distribution of the peripheral compartmentaDisease type 2 Kawasaki disease, Disease type 3 Guillan-Barré syndrome, Disease type 4 Thrombocytopenia, Disease type 5 Myasthenic crisis, Disease type 6 Systemic lupus erythematosus, Disease type 7 others, which includes neutropenic sepsis, pemphigus foliaceus, viral encephalitic/encephalitis and multiple myelomabMultiple proportional errors are reported but unclear how they are incorporatedcNot clear whether expressed as variance or standard deviation Pharmacokinetic parameter estimates and random effects model CL = 0.142 Q = 0.252 Vc = 3.94 Vp = 4.18 CL = 0.152b (BW/72)^0.796 Q = 0.825 Vc PID = 3.08b (BW/72)^1.1 Vc SID = 8.75b (BW/72)^1.1 Vp = 1.8 CL = 0.139b (BW/58.7)^0.881 Q = 0.300 Vc = 4.01b (BW/58.7)^0.501 Vp = 3.51 CL = 36.02 Q = 54.25 Vc = 92.08 Vp = 325.4 CL = 0.0027 Q = 0.045 Vc = 0.008 Vp = 0.055 Vc = 22 Vp = 29 CL = 0.138b (BW/66)^0.768 Q = 0.260 Vc = 3.95b (BW/66)^0.448 Vp = 4.44 CL = 36.02 Q = 61.52 Vc = 92.58 Vp = 252.77 CL = 0.183b (LBM/47)^0.75 Q = 0.353b (LBM/47)^0.75 Vc = 3.01b (LBM/47)^1 Vp = 1.40b (LBM/47)^1 F1 without hyaluronidase = 70.5% F1 with hyaluronidase = 79.4% CL = 0.150b (BW/65.7)^0.744 Q = 0.474 Vc = 3.06b (BW/65.7)^0.686 Vp = 1.93b (BW/65.7)^1.04 CL = 0.036b (BW/43)^0.75 Q = 0.317 Vc = 1.20b (BW/43)^1 Vp disease type 2a = 0.52b (BW/43)^1b exp(1.67) Vp disease type 3a = 0.52b (BW/43)^1b exp(1.65) Vp disease type 4a = 0.52b (BW/43)^1b exp(1.46) Vp disease type 5a = 0.52b (BW/43)^1b exp(1.93) Vp disease type 6a = 0.52b (BW/43)^1 exp(1.97) Vp disease type 7a = 0.52b (BW/43)^1b exp(2.16) CL = 28.35 Q = 67.27 Vc = 45.41 Vp = 58.19 CL = 0.435b (BW/82)^0.615 Q = 0.50b (BW/82)^0.615 Vc = 4.69b (BW/82)^0.773 Vp = 1.87b (BW/82)^0.773 CL = 0.28b (BW/70)^0.75b 1.31 (if MP)b 1.22 (if GBS-DS = 3 or 4)b 1.83 (if GBS-DS = 5) Q = 0.25 Vc = 2.87 Vp = 2.65 Add additive, BW body weight, CL clearance, F1 bioavailability, GBS-DS Guillain-Barré Syndrome disability score, Ka absorption rate constant, LBM lean body mass, MP methylprednisolone, NR not reported, PID primary immunodeficiency, Prop proportional, Q intercompartmental clearance, SID secondary immunodeficiency, V volume of distribution, Vc volume of distribution of the central compartment, Vp volume of distribution of the peripheral compartment aDisease type 2 Kawasaki disease, Disease type 3 Guillan-Barré syndrome, Disease type 4 Thrombocytopenia, Disease type 5 Myasthenic crisis, Disease type 6 Systemic lupus erythematosus, Disease type 7 others, which includes neutropenic sepsis, pemphigus foliaceus, viral encephalitic/encephalitis and multiple myeloma bMultiple proportional errors are reported but unclear how they are incorporated cNot clear whether expressed as variance or standard deviation Distribution In eight out of ten studies, IgG pharmacokinetics were described using a two-compartment model (Table 4) [35, 37, 38, 40–42, 44, 54]. Volume of distribution of the central compartment (Vc) ranged from 1.95 to 8.48 L per 70 kg and volume of distribution of the peripheral compartment (Vp) ranged from 1.63 to 6.07 L per 70 kg. Two studies reported a one-compartment model [36, 53]. Volume of distribution at steady state (Vss = Vc + Vp) ranged from 2.80 to 13.46 L [36–38, 40–42, 44, 53, 54]. In the study by Dumas et al. (2019), a one-compartment model was constructed using data from both IVIg and SCIg [53]. The volume of distribution of the central compartment (Vc) was estimated to be 4.01 L, with IIV (CV%) estimated to be 39.2%. The authors suggested that possible explanations for using a one-compartment model may include the fact that samples were collected only during steady state and certain dose adjustments unique to the country where the clinical trials were performed. In the study by Lee et al., a one-compartment model was found to best describe the data, based on the Bayesian information criterion (BIC) [36]. The Vc was estimated to be 2.77 L with IIV (CV%) of 8.0%. Body weight was included as a covariate on Vc with an estimated exponent of 0.66. The small sample size and the limited sampling spread may explain why a two-compartment model did not provide a better fit than a one-compartment model. In seven out of ten models, BW was included as a covariate on Vc (Tables 3 and 4) [36–38, 41, 42, 44, 54]. The allometric scaling components for Vc were fixed to 1 in two models [38, 42]. In the remaining five models, allometric scaling exponents ranged from 0.576 to 0.881 [36, 37, 41, 44, 54]. In three out of ten models, BW was included as a covariate on Vp [37, 38, 42]. While Li et al. and Lee et al. fixed the allometric scaling component to 1 [38, 42], Navarro-Mora et al. estimated it to be 1.04 [37]. One study incorporated LBM as a covariate using allometric scaling, and the exponents on Vc and Vp were fixed to 1 [42]. Elimination A large variability in clearance (CL) was observed between popPK models developed for PID. Estimated parameter values for CL ranged from 0.05 to 0.21 L/day per 70 kg (Table 4). In seven out of ten studies, BW was incorporated as a covariate on CL [36–38, 41, 44, 53, 54]. In six models, the allometric scaling component was estimated, and ranged from 0.576 to 0.881 [36, 37, 41, 44, 53, 54]. In the remaining model, the allometric scaling component was fixed to 0.75 [38]. Notably, Li et al. derived the LBM and incorporated it as a covariate on CL instead of BW, using a fixed allometric scaling component of 0.75 [42]. For the two-compartment models, parameter estimates for intercompartmental clearance (Q) ranged from 0.25 to 0.825 L (Table 4) [37, 40–42, 54, 55]. Only the model by Li et al. included LBM as a covariate on Q, using a fixed allometric scaling exponent of 0.75 [42]. Special Patient Populations Although all studies included both children and adults, age was not identified as a covariate affecting the PK of IgG (Table 3). Tegenge et al. developed a popPK model exclusively based on data from very low birth-weight neonates receiving IVIg [35]. The final model included an estimated allometric scaling component of 3.5 × 10−7. Interindividual variability (%CV) was included on CL (52%), Vc (22%) and Vp (29%), and in the extensive sampling population CL appeared to be higher in neonates (0.0027 L/day per kg) compared to older children and adults (0.0014 L/day per kg) when scaled to a reference of 70 kg, prompting the authors to suggest that increased dosing may be required to ensure effective treatment in this special patient population [35]. Covariates Three studies reported baseline IgG concentrations of the included patient populations [38, 53, 54]. In these cases, the baseline IgG concentrations consisted of both endogenous IgG and exogenous IgG, since all patients were required to have been treated with IgRT for at least 3 months prior to inclusion. Dumas et al. and Lee et al. reported median baseline IgG values of 9.9 g/L (range 3.4–19.9) and 9.08 g/L (range 0.61–33.03), respectively [38, 53]. In both studies, patients were already on stable IVIg treatment, explaining the high IgG baseline values compared to reported IgG concentrations in IgG treatment-naïve patients [56]. Reported baseline IgG concentrations in the study by Tortorici et al. were measured 28±2 days after IgRT dose [54]. Therefore, these concentrations represent both endogenous and exogenous IgG. The median [min-max] baseline IgG value in SID patients was 6.15 [2.05–17.1] and the median IgG value was 9.17 [3.93–27.2] in PID patients. There was high variability in baseline IgG concentrations in the study by Lee et al. (0.61–33.03 g/L), which can be explained by the wide range of included indications [38]. Baseline IgG concentration was investigated as a covariate on Vc, Vp, and CL; however, this relationship was not significant. Endogenous (treatment-naïve) IgG concentrations reported by Lee et al. ranged from 0.3 to 5.1 g/L, and mainly XLA patients were included [36]. Endogenous baseline IgG concentrations were incorporated as a popPK parameter in the included models using various methodological approaches. In several studies, the endogenous IgG concentration was fixed to 4 g/L [40, 41, 44, 54]. Navarro-Mora et al. evaluated a fixed value of 4 g/L and a lower value of 1.5 g/L. However, because changes in PK parameters were minimal, endogenous IgG was fixed at 4 g/L in the final model [37]. Lee et al. subtracted the measured endogenous IgG concentrations prior to treatment from the total IgG concentration [36]. If pre-treatment data were unavailable, an endogenous IgG concentration of 0.7 g/L was used. Tegenge et al. set the endogenous IgG concentration to 5 g/L based on observed pre-infusion IgG concentrations in neonates [35]. Li et al. estimated a typical value for endogenous IgG concentration [42]. Two studies did not report how they incorporated endogenous IgG concentrations [38, 53]. Of the ten studies, only Tortorici et al. used data from both PID and SID patients in the development of the model [54]. Disease type was incorporated as a covariate on Vc. In PID patients, the estimated Vc (3.08 L) was lower than the estimated Vc in SID patients (8.75 L). The authors concluded that this difference had no impact on the overall IgG exposure (the area under the concentration–time curve [AUC]0–28 days) and that there was no evidence to treat PID or SID patients differently. Given its role in regulating the recycling and systemic persistence of IgG, FcRn represents a biologically plausible covariate in popPK models of IgG-based therapeutics. Of the included studies, two assessed variable number of tandem repeats (VNTR) FcRn polymorphisms in a predominantly Asian population [36, 38]. The first study by Lee et al. investigated the impact of different VNTR FcRn polymorphisms on the PK of IgG [36]. The heterozygous FcRn polymorphism (VNTR2/3) was observed in four patients and the homozygous polymorphism (VNTR3/3) in six patients (Fig. 1). Ultimately, FcRn polymorphisms were not identified as a significant covariate. Notably, the small sample size may have influenced these findings. The second study investigated the effect of FcRn polymorphisms in a larger population of 79 patients [38]. The predominant polymorphism in this study population was VNTR3/3 (84.8%), followed by VNTR2/3 (13.9%) and VNTR3/4 (1.2%). Again, genetic polymorphisms of FcRn were not found to be a significant covariate of IgG PK. Since the majority of the included patients in the studies by Lee et al. (2024) were not diagnosed with PID, this may have influenced the findings, assuming different polymorphisms impact IgG PK in replacement therapy more than in other applications of IgG [38]. PRISMA-flowchart of included studies. IgG immunoglobulin, IVIg intravenous immunoglobulins, popPD population pharmacodynamics, popPK population pharmacokinetics, SGIg subcutaneous immunoglobulins Statistical Model Two main statistical models were incorporated into popPK models. First, inter-individual variability (IIV) quantified the variation in a PK parameter between subjects [57]. Second, residual unexplained variability (RUV) represented the remaining unexplained variability in a PK parameter after accounting for other sources of variance [57]. The underlying statistical models for each model are reported in Table 4. In nine out of ten studies, IIV was described using an exponential model [35–38, 41, 42, 44, 53, 54], while the remaining study reported IIV but did not specify the underlying relationship [40]. Residual unexplained variability was modelled using a proportional error in four studies [35, 36, 38, 53], whereas five studies used a combined additive and proportional error model [37, 41, 42, 44, 54]. The remaining study did not report the error structure [40]. Immune-Mediated Diseases In contrast to replacement therapy used in patients with PID or SID, immune-mediated neuropathies and other autoimmune diseases are typically treated with high doses of IVIg aimed at achieving an immunomodulatory effect [1, 58]. For these indications, IVIg is commonly administered as an induction therapy at a dose of 2000 mg/kg over a period of 2 to 5 days [1]. For chronic immune-mediated diseases, additional maintenance therapy is often required, involving a highly variable range of doses for IVIg or SCIg [1, 47]. Notably, patients with PID or SID often have limited endogenous IgG production, whereas individuals with immune-mediated diseases generally have normal or even elevated endogenous IgG concentrations. General information of the included models In total, four of the studies in Table 1 constructed a popPK model for patients with immune-mediated disease [23, 27, 39, 43]. Only Tortorici et al. used both IVIg and SCIg data in their model [43]. The remaining three studies used only IVIg data in their models [23, 27, 39]. Of the four models, one was constructed for GBS [27], one for CIDP [43] and two for MMN [23, 39]. The two studies by Li et al. in patients with MMN utilised the same patient cohort [23, 39]. One study focused on the PK [39] and the other study incorporated a PD component [23]; the latter will be described in more detail later in this review. All four studies included data from studies performed in Europe. Three out of four studies included data from studies performed in the USA [23, 39, 43], three out of four data from studies performed in Canada [23, 39, 43] and one out of four included data from studies performed in Asia [43]. Absorption and Bioavailability Table 4 summarises the popPK parameters for the included studies. Tortorici et al. described the absorption of SCIg as a first-order process, with a Ka of 0.439 day−1 fixed from a previous study because of limited data in the absorption phase [40, 43]. No significant covariates were described on Ka. Relative bioavailability of SCIg compared with IVIg was estimated at 83%, which is markedly higher than that reported in studies of PID and SID patients for unfacilitated SCIg (range 66–74%). Distribution In line with findings in PID, most studies described the distribution using a two-compartment model (n = 2) (Table 3) [27, 43]. Li et al. applied a one-compartment model, likely due to the availability of only trough concentrations, limiting the ability to characterise the distribution phase [39]. In the latter study, the estimated volume of distribution was 6.59 L for the reference patient with a LBM of 56.54 kg. The estimated exponent for allometric scaling based on LBM was 2.23. Tortorici et al. reported the Vc estimated as 4.15 L per 70 kg and Vp as 1.65 L per 70 kg, with an estimated exponent for allometric scaling of 0.773. Fokkink et al. reported a Vc of 2.65 L per 70 kg and Vp of 2.87 L per 70 kg with a fixed allometric scaling exponent of 1 [27]. Lee et al. described a covariate effect of disease type on Vp with exponents ranging from 1.46 to 2.61 based on the disease, resulting in typical parameter estimates for Vp of 3.6 to 11.5 L per 70 kg [38]. Elimination First-order elimination was applied in all models. Estimated clearance ranged from 0.28 to 0.41 L/day per 70 kg (Table 4), which is higher than values typically reported for PID patients (0.02 to 0.21 L/day per 70 kg, Table 4). Two studies incorporated allometric scaling; Tortorici et al. (2021) estimated an exponent of 0.615 for clearance, while Fokkink et al. fixed the exponent at 0.75 [27, 43]. Li et al. did not report allometric scaling on elimination [39]. Additionally, Fokkink et al. found that concomitant treatment with methylprednisolone and greater disease severity, as defined by the GBS disability score, significantly increased IVIg elimination (Table 4) [27]. Of note, Fokkink et al. reported no significant association between VNTR genotype and PK parameters [27]. Covariates Various approaches were used to account for endogenous IgG concentrations. Tortorici et al. modelled endogenous IgG based on observed treatment-naïve IgG concentrations, incorporating the residual error [43]. Fokkink et al. estimated a typical baseline value with associated IIV, and included disease type as a covariate, since CIDP patients were already on maintenance therapy and therefore had higher baseline concentrations compared to treatment-naïve GBS patients [27]. Li et al. estimated a typical endogenous IgG value (CBASE) with associated IIV [39]. Statistical Model The underlying statistical models for each model are reported in Table 4. Two of three studies used an exponential model to describe IIV [27, 43], while the third reported IIV but did not specify the underlying relationship [39]. Similarly, residual variability was modelled using a proportional error in two studies [27, 43], whereas the third did not specify the error structure [39]. Comparability The simulated concentration–time profiles for IVIg in patients with PID (Fig. 2A) demonstrated notable variability. Since the model presented by Tegenge et al. is specifically constructed for very low birth-weight neonates, this model was excluded from the simulations. Simulations based on the model by Lee et al. yielded higher IgG concentrations compared to those generated using the models of Lee et al. and Tortorici et al. This difference can be explained by the low Vc and Vd, which automatically led to increased IgG concentrations. Furthermore, Lee et al. included data from patients with multiple different conditions, including PID. As expected, estimated IgG concentrations in patients with GBS during induction treatment and CIDP/MMN during maintenance therapy were generally higher than those observed in PID, reflecting differences in dosing regimens and assumed endogenous IgG concentrations. In contrast to the variability observed with IVIg, only minor differences were noted in the simulated profiles for SCIg in PID patients (Fig. 2B). Moreover, IgG concentrations following SCIg administration appeared to be more stable over time compared to those following IVIg treatment. Simulations using the available published popPK models describing IgG concentrations after administration of IVIg and SCIg. Endogenous IgG concentrations were assumed to be 10 g/L for CIDP, MMN and GBS, and 4.0 g/L for PID. A Population predictions for dosing of IVIg. Of note, both models from Li et al. overlap as they report the same popPK model with Li et al. (Ann Cln Transl Neurol.) included additional data [23, 39]. Standard doses per indication: PID: IV 400 mg/kg loading dose, followed by 400 mg/kg maintenance dose once Q4W, CIDP and MMN: IV 2000 mg/kg over five days loading dose, followed by 1000 mg/kg maintenance dose once Q3W GBS: IV 2000 mg/kg for five days, B population predictions for dosing of SCIg. Standard doses per indication: PID: SC 100 mg/kg loading and maintenance dose QW, CIDP and MMN: SC 2000 mg/kg over five days loading dose, followed by 1000 mg/kg maintenance once Q3W. CIDP chronic demyelinating polyneuropathy, GBS Guillain-Barré syndrome, IV intravenous, IVIg intravenous immunoglobulins, MMN multifocal motor neuropathy, PID primary immunodeficiency, popPK population pharmacokinetics, QW weekly, Q3W every 3 weeks, Q4W every 4 weeks, SC subcutaneous, SGIg subcutaneous immunoglobulins, Pharmacodynamic Analyses In total, two studies from Table 1 comprised analyses in which a pharmacodynamic (PD) analysis was conducted, next to constructing a popPK model [23, 43]. Pharmacodynamics of IVIg in MMN In the study from Li et al., 44 adult patients with MMN receiving IVIg were included (see Tables 1 and 2) [23]. To describe the efficacy of therapy with IVIg, grip strength measurements were obtained using a dynamometer. With these assessments, exploratory exposure-response modelling was performed to evaluate the correlation between changes in grip strength and serum trough IgG concentrations. In an indirect response PK-PD model, the change in grip strength (∆GS) over time (∆t) was described using the following equation: in which MNT was the production rate of grip strength, Imax the maximum inhibitory effect, DRV, the IgG concentration in excess of baseline IgG concentration, C50 the concentration at which 50% of the maximum inhibitory effect was achieved, and DTR, which was the deterioration rate of grip strength (GS). The latent model derived parameter for GS and the logit transformation of Imax are further described in Supplemental Equations S1 and S2. In the final model, an estimate of 10.6 kg for latent model-derived parameter of GS in the absence of treatment (GBASE) was obtained with an IIV (%CV) of 86.8. The deterioration rate was estimated at 0.023 (1/hour) for patients without receiving IVIg treatment and for C50 a value of 9.41 mg/mL was obtained. For LIMAX (logit transformed Imax value), a value of 0.433 was estimated, which translated into a value of 0.61 mg/mL for Imax. The residual unexplained variability (RUV) of this PD model was typically associated with factors such as a proportional error term with an estimated value of 0.0602, with a low shrinkage value of 7.0%. Furthermore, estimation of the IIV (%CV) for the RUV term was also considered and a value of 24.5 was obtained with a low shrinkage of 0.8%. In the covariate analysis, no relationships explaining the obtained IIV for GBASE or the RUV term were obtained. Using the established population PK-PD model, Monte Carlo simulations were performed for different IVIg dosing regimens with administration intervals of 2, 3, and 4 weeks. On the basis of the predicted grip strength, the percentage of patients could be calculated with a meaningful clinical improvement, as a change in grip strength of 4 kg was considered the minimal effect change for meaningful clinical improvement. Using the calculated PD effects, the most optimal dosing regimen could be selected for each dosing interval, showing the usefulness of the PK-PD model in individualising the dosing regimens for IVIg. On the basis of the Monte Carlo simulations for different dosing regimens, it was demonstrated that in ≥ 70% of patients a clinically meaningful improvement in GS was obtained for an IVIg dose of ≥ 1600 mg/kg/month. Moreover, for high doses (> 1000 mg/kg) dose-splitting did not impact GS. Pharmacodynamics of IVIg in CIDP In a study by Tortorici et al., a population PD analysis was conducted to investigate the relationship between IgG exposure and the disease severity of CIDP [43]. The analysis aimed to predict how changes in the Inflammatory Neuropathy Cause and Treatment (INCAT) score are related to IgG concentrations. In total, data from 171 patients with CIDP receiving SCIg or placebo were analysed. For the individual PK parameters of all patients, a previously published population PK study that described Ig concentrations after SCIg administration was used. To establish the exposure-response model, data were obtained during the baseline visit and throughout the treatment with SCIg for all patients who had received at least one dose. The analysis used the change in IgG concentration from a pre-treatment baseline (ΔIgG) after SCIg administration as the key exposure metric. This was defined as the total serum IgG at the time of an INCAT assessment minus the IgG concentration at relapse after withdrawal from IVIg treatment, i.e., the endogenous concentration of IgG. This specific metric was selected because lower ΔIgG concentrations were found to be a clearer predictor of disease worsening than lower total IgG concentration scores in CIDP patients [59]. An increase in the INCAT score represents a worsening of the disease, while a decrease signifies improvement and no change is regarded as stable disease. The relationship was described using a latent variable exposure-response model, where the drug effect on the INCAT score (Et) was best characterised by a maximum drug Emax model as a function of ΔIgG (Supplemental Equation S3). For predicting the INCAT score, an ordered categorical probit model was applied using the following general form: in which probit is the inverse of the cumulative normal distribution function (Φ-1), m represented the observed INCAT score, Pr(INCAT≤m) representing the probability of the observed INCAT score being lower or equal to m, fb(m) the baseline (intercept) component, fnd the nondrug (or placebo) component, and fd(E(t)) the drug model component. In the final model, the nondrug component was not considered. Although the INCAT score (m) had a maximum of 10, the scores for the individual patient’s arms and legs, with a maximum of 5, were modelled. A baseline component fb(m) was modelled for all score values m ≤ 5 (Supplemental Equation S4). For the baseline parameter, covariate relationships were estimated. For age, a linear relationship demonstrated a decrease in INCAT baseline score of -0.0862 per year per deviation from 57 years, and being Japanese increased the baseline by 4.06. Although covariate relationships were tested for Emax as well, no relationships were retained for Emax in the final model. Using the final exposure-response model, a clear relationship between IgG concentrations and clinical outcomes was described. Higher ΔIgG concentrations were associated with a greater probability of the disease being stable or improving and, conversely, a lower probability of worsening. The model estimated that the ΔIgG concentrations corresponding to 20%, 50%, and 80% of the maximum predicted increase in probability of a stable or improved INCAT were 0.8, 2.8, and 8.1 g/L, respectively. At these ΔIgG concentrations, the predicted probability of having a stable or improved INCAT score was 70%, 78%, and 87%, respectively. Furthermore, the analysis demonstrated that the predicted probability of a stable or improved INCAT score on the basis of the IgG trough concentrations for weekly subcutaneous dosing was 81% for a dose of 200 mg/kg and 86% for a dose of 400 mg/kg. Therefore, the authors concluded that the dosing interval could be handled flexibly if the total weekly dose remained adequate allowing to individualise the dose of SCIg. Discussion In this systematic review, 14 immunoglobulin popPK models were included for various diseases, among which two were PK-PD models [23, 43]. Understanding the PK and PD of human immunoglobulin therapy for PID/SID and immune-mediated diseases is essential, given the dose-response relationships implicated across multiple indications. However, dosing often remains empirically set based on decade-old studies, frequently relying on trial-and-error approaches in current clinical practice. Moreover, clear target concentrations have not been established for all IVIg and SCIg indications. This review provides a detailed overview of current popPK and PD models for IVIg and (f)SCIg and may help to explore the variability in IgG concentrations between patients and various treatment regimens. A key distinction between PID and SID, and immune-mediated diseases lies in the IgG dosing regimen and the underlying pathophysiology of the different conditions. In PID, IgG doses are substantially lower than those administered in immune-mediated diseases such as CIDP [1], resulting in correspondingly lower plasma IgG concentrations (Fig. 2). In addition, a form of PID (antibody deficiency disorders [PAD]) causes defects in the adaptive immune system, which leads to low antibody levels [45]. Endogenous IgG concentrations are generally higher in immune-mediated diseases since they usually stem from causes that do not affect the antibody production [60, 61]. Therefore, differences in endogenous IgG production could further explain the differences in IgG concentrations. Differences in dosing regimens contribute to generally lower estimated CL values in PID patients compared with those receiving high-dose immunoglobulin therapy for indications that require immunomodulatory effects. A possible explanation for this increased CL is saturation of FcRn after high-dose immunoglobulin treatment, leading to reduced recycling of IgG. Additionally, the study by Lee et al., which included multiple indications, reported that the estimated Vp for IVIg in immune-mediated diseases was 5–10 times higher than in replacement therapy [38]. This difference is likely attributable to increased vascular permeability and consumption at sites of inflammation. In the context of PID, IgG levels were monitored to assess therapeutic efficacy, with the aim of restoring plasma IgG concentrations to be comparable to those observed in healthy individuals. In contrast, higher IgG doses are required in immune-mediated diseases to achieve immunomodulatory effects. At these elevated concentrations, IgG can inhibit components of the complement system and modulate immune responses [12]. Therefore, high-dose IVIg or SCIg therapy is warranted in such conditions to ensure sufficient immunological impact, which is not required in PID [62]. Across the included popPK models, the reported bioavailability values of SCIg are comparable [37, 40–42, 44, 53]. Reported values for bioavailability range from 66.0 to 73.9%, consistent with the available literature. According to a study by Berger et al, the bioavailability of four different SCIg products ranged from 65.0 to 69.0% [63]. Among the included popPK models, only Li et al. have incorporated data from fSCIg treatment [42]. The bioavailability appears to be higher for fSCIg compared to SCIg. Li et al. reported an estimated bioavailability of 79.4%, which corresponds to 92.2% when expressed as a ratio relative to IVIg [42]. These findings are consistent with other reports in the literature, where the bioavailability of fSCIg based on the AUC has been reported to be approximately 93% [64]. Given its hydrophilic properties, IgG has a relatively small volume of distribution [65]. This is reflected in the reported population parameter estimates across the included popPK models. Most models estimated the Vc between 2.0 and 8.5 L per 70 kg, with the exception of the model by Tegenge et al., which was specifically targeted toward very low birth-weight neonates [35]. In this special population, the estimated Vc and Vp were 8.7 and 60.0 mL, respectively. This is markedly smaller than the volumes reported in adult populations and aligns with the reported circulating blood volume for preterm neonates [66]. Most popPK models included combined data of both adults and children and accounted for physiological differences using allometric scaling for weight. However, age was not included as a significant covariate in any of the included models. Interestingly, in patients with SID, the estimated Vc was 8.5 L per 70 kg, which is considerably higher than the values typically reported in patients with PID, ranging from 2.0 to 5.0 L per 70 kg [54]. A similar estimate for Vc in SID patients has been reported in the literature [67]. A study conducted in 30 bone marrow transplant patients treated with IVIg at 250 or 500 mg/kg/week reported a volume of distribution of 9.4 L per 70 kg [67]. The hypothesis regarding the higher Vc in SID patients is that it may be attributed to a gradual, ongoing decrease in endogenous IgG due to progressive immunodeficiency [54, 68]. This contributes to a decrease in plasma concentration and thus an increase in Vc. It is important to note that this is a relative increase in Vc. Although the exact cause of this difference in Vc between PID and SID patients remains unclear, the finding underscores the importance of considering disease type as a potential covariate in future popPK modelling efforts, particularly when integrating data from heterogeneous patient populations. The long half-life of IgG is largely attributed to FcRn, and (genetic) variation in its expression may contribute to variable PK of therapeutic immunoglobulin. Variable number of tandem repeats within the FCGRT gene (encoding FcRn) have been associated with differences in FcRn expression, with monocytes from individuals homozygous for three repeats (VNTR3/VNTR3) showing higher transcriptional activity compared to those with the VNTR2/VNTR3 genotype [69]. However, of the three popPK studies described in this review that tested VNTRs as a covariate, none reported a significant impact on any of the PK parameters [27, 36, 38]. In line with this, studies using non-compartmental PK analysis in patients with MMN and GBS reported no significant association between FCGRT polymorphisms and PK after IVIg treatment [25, 70]. In contrast, studies in patients with CVID and myasthenia gravis have reported associations between these genetic variants and altered PK or clinical outcomes [71–73]. In addition, there is evidence that FCGRT polymorphisms affect the PK of monoclonal antibodies [74]. Especially in patients receiving high-dose IVIg, FcRn is likely to be saturated and differences in FcRn expression may shift the threshold at which no free receptor is available. This subsequently results in reduced recycling of IgG and enhanced lysosomal degradation. Studies reporting the absence of a significant effect of FcRn typically have a relatively small sample size. The inconsistent findings regarding the influence of genetic variation in FcRn on immunoglobulin pharmacokinetics highlight the need for further investigation in larger studies. Among the included popPK models, estimated CL values for adults ranged from 0.05 to 0.4 L/day. Most CL estimates are in line with the estimated clearance of endogenous IgG reported in literature (0.21 L/day) [75]. However, there are some exceptions. There is considerable variability in estimated CL values. Most models have incorporated BW (or LBM) as a covariate on CL. Therefore, covariate relationships probably do not explain the observed variability. The lowest value of 0.05 L/day was reported by Lee et al., who incorporated multiple immune-mediated diseases into their model using PID as a reference [38]. An explanation for this low CL value is unclear. The highest CL value (0.4 L/day) was reported by Tortorici et al. [43]. This was discussed as being in line with previously reported CL values in the literature; however, no specific reference was provided. Allometric scaling is a widely used approach for extrapolating PK parameters based on BW. Standard exponents of 0.75 for CL and 1.0 for V are commonly applied. However, these fixed exponents may not be suitable for all patient populations. In particular, they do not account for the maturation of metabolic enzymes, which limits their applicability in neonates [76]. In the study by Tegenge et al., incorporating an estimated allometric exponent of 3.5 × 10−7 did not improve the performance of the popPK model compared to models using fixed exponents [35]. This lack of improvement may be attributed to the narrow BW range within the study population, which constrained the ability to reliably estimate the allometric exponent and resulted in a very low value. Body weight was commonly incorporated as a covariate in IgG popPK models. Most models use BW; however, a selection of models has incorporated LBM as a covariate [23, 39, 42]. Lean body mass does not take fat mass into account, making it a potentially more appropriate weight parameter to use for dosing immunoglobulins in obese patients [32, 65]. Given the hydrophilic properties of IgG, it is unlikely that it distributes significantly into adipose tissue [65]. Consequently, standard dosing based on BW in obese patients may lead to elevated plasma IgG concentrations. To mitigate this risk, individualised dosing strategies using alternative weight metrics, such as LBM, ideal body mass (IBM) or adjusted body mass (ABM), may provide a more appropriate dosing approach. Although alternative weight-based dosing approaches have been described in the literature, they are primarily focused on cost reduction [77–79]. Using LBM, IBM or ABM requires less IgG, thereby lowering treatment costs [77]. Limited evidence suggests dosing based on IBM does not increase 30-day hospital readmission rates or length of stay [79]. In Canada, ABM-based dosing has already been adopted as standard of care [80]. Further research comparing standard BW-based dosing with alternative dosing strategies, particularly those focused on clinically relevant PD outcomes, may provide more robust evidence regarding the feasibility, efficacy and safety of dosing based on alternative weight metrics. Ideally, such research would be supported by popPK-PD models to provide additional insight into the sources of variability in IgG concentrations and associated clinical outcomes. In the context of immunodeficiencies and IgRT, endogenous IgG can significantly affect the IgG concentrations achieved with treatment. Although based on a simulation study, patients with a lower endogenous IgG concentration (1.5 g/L) seem to reach lower trough IgG concentrations after treatment with IVIg or (f)SCIg than patients with higher endogenous IgG concentrations (4.0 g/L) [81]. Especially for treatment-naïve patients, it is suggested that an individualised dosing regimen that takes endogenous IgG into account may be required for this group of patients to ensure trough level target attainment [81]. Most popPK models treat endogenous IgG as a fixed parameter, typically based on data from treatment-naïve patients with immunodeficiencies. A commonly used value is 4.0 g/L, as similar values have been reported in treatment-naïve CVID patients [56, 82]. However, this approach does not adequately reflect the substantial variability in endogenous IgG concentrations among PID patients. Endogenous concentrations may vary depending on the specific type of immunodeficiency. For instance, patients with X-linked agammaglobulinemia (XLA) have no endogenous IgG production, resulting in lower levels compared to patients with CVID [83]. One popPK model did estimate a parameter describing the endogenous IgG. However, the data consisted mostly of patients who received IgG treatment prior to the study, which may compromise the precision of the estimated value of 6.15 g/L. Incorporating additional data from treatment-naïve patients could improve this estimate. Furthermore, accounting for the specific type of immunodeficiency may enhance the predictive performance. In immune-mediated diseases, the endogenous IgG concentration seemed less influential. Unlike in immunodeficiencies, there are no target IgG concentrations established, as efficacy is mostly determined by assessing clinical outcomes such as grip strength [22]. This is also reflected in the outcomes modelled by popPD. Inflammatory neuropathy cause and treatment scores in MMN did not seem to be mediated by endogenous IgG [43]. However, increases in delta IgG were associated with an increased probability of stable or increased INCAT scores. This suggests that the administered IgG dose, rather than achieving a specific target IgG concentration, is one of the factors driving clinical response in this indication. In this review, the focus is on polyclonal IgG, but compared to that, a relatively large number of studies have been published on the PK of therapeutic monoclonal antibodies (mAbs) [75, 84]. Therapeutic mAbs share PK characteristics similar to those of polyclonal products. Both types of antibodies are eliminated mainly by intracellular proteolysis after nonspecific pinocytosis and are often described using two-compartment models with relatively small volumes of distribution [75]. The median estimate for Vc is 3.1 with a range of 2.4–5.5 L in mAbs compared with 1.95–8.48 L per 70 kg for IVIg [75]. For Vp, the median range is 2.7–2.8 L compared with 1.63–6.07 L per 70 kg for IVIg [75, 84]. However, in contrast to polyclonal IgG, target-mediated drug disposition plays a significant role in the elimination of mAbs, where the drug binds with high affinity to a target, and this drug-target complex is subsequently internalised, leading to an increased turnover [75, 84]. In addition, anti-drug antibodies (ADAs) can significantly increase the elimination of mAbs [75, 84]. These processes are often reflected in population PK models for mAbs, where non-linear kinetics or a combination of linear and non-linear elimination pathways are incorporated. In contrast, all included popPK models for IVIg describe elimination using a first-order linear approach. The main reason is that IVIg is less immunogenic and lacks a clear antigenic target. A limitation of the popPK models for high‑dose IVIg identified in this review is that they generally do not explicitly describe FcRn turnover or saturation. Nevertheless, several studies on monoclonal antibodies and in animal models demonstrate that FcRn‑mediated nonlinear kinetics can be incorporated into population PK models using relatively simple structural models [16, 85, 86]. In addition, (minimal) physiologically based pharmacokinetic ((m)PBPK) models, which were beyond the scope of this systematic review, may offer mechanistic insights into immunoglobulin PK. Their primary advantage is the ability to incorporate detailed physiological processes for IgG disposition [87]. However, a major drawback of PBPK models is their limited ability to characterise interindividual variability compared with popPK approaches. For future research, extending population PK models for IVIg or SCIg with explicit FcRn kinetics, potentially informed by minimal PBPK structures or simplified FcRn‑turnover mechanisms, may provide a better description of IgG nonlinearity at high doses. Available research in other fields, such as oncology, may offer additional insights that could prove useful for application in immunoglobulins [88]. The availability of PD analyses for immunodeficiencies is notably limited, particularly for PID, where a literature search revealed no popPD studies. This knowledge gap may hinder the optimisation of dosing regimens and the development of a deeper understanding of the exposure-response relationship in this patient population. In contrast, two PD analyses have been conducted for CIDP. For instance, the PATH study provided a pharmacometric model that linked IgG exposure to clinical efficacy, as measured by the INCAT score [43]. This study demonstrated that higher IgG exposure is associated with better clinical outcomes, as defined by the INCAT score. This relationship between IgG PK and PD has been described previously, focusing on grip strength as a PD outcome [26]. For PID patients, establishing a dose-response relationship might be more complex, as time-to-event analysis based on the occurrence of a (serious) infection has to be performed [89, 90]. However, specific ADA titres or immune cell counts could be used as measurements of the physiological response to treatment with IVIg or SCIg as well [91, 92]. Moreover, the number of hospital days or antibiotic use might also be a crude measure. Nevertheless, PD models remain essential for establishing an optimal dosing regimen to maintain disease stability. Therefore, the overall scarcity of PD analyses across both PID and immune-mediated diseases underscores a significant unmet need for further research to better characterise the PD of immunoglobulin therapies and to develop more individualised treatment approaches that also take the PD into account. In particular, the absence of PD models for PID is a research gap that should be explored in future studies. Conclusion Over the past decades, increasing efforts have been made to better understand the PK of human polyclonal immunoglobulin therapy through popPK(-PD) modelling. To reflect this progress, this review included 14 published popPK studies on IVIg, SCIg, and fSCIg. Most of these studies described the disposition of IgG using a two-compartment model with first-order kinetics. Body weight, through allometric scaling, was often significantly associated with PK parameters. Besides body weight, only a few disease-specific covariates were reported to significantly explain variability between patients. Intravenous Ig is used for an increasing number of indications, and it is apparent that PK may differ across conditions, highlighting the need to account for the disease in modelling efforts. While current models can predict IgG concentration-time profiles with reasonable accuracy, defining clear target concentrations remains challenging due to the heterogeneity of clinical indications and patient populations. The PopPK-PD models offer a powerful framework to evaluate treatment effects in relation to IgG exposure after therapy. Expanding the development of popPK-PD and PBPK models, along with a better understanding of the factors that determine the efficacy of IVIg or SCIg, is required to move towards model-informed precision dosing of immunoglobulin therapy. Nevertheless, the available models could already be used to inform clinical trial design to investigate the feasibility of novel, personalised dosing strategies. Supplementary Information Below is the link to the electronic supplementary material. Supplementary file (DOCX 102 KB) Acknowledgements The authors wish to thank dr W.M. Bramer from the Erasmus MC Medical Library for developing and updating the search strategies. Funding No funding was obtained for this study. Declarations Conflict of interest SZ, MC, and TP have no competing interests to declare. ST received funding for research from Octapharma (all paid to institution). VASHD received funding for research from Takeda, CSL Behring, Pharming and Moderna (all paid to institution), and honoraria for lectures from AstraZeneca, Pfizer, Pharming, Takeda, CSL Behring and Johnson & Johnsons (all paid to institution). BCJ received funding for research from Sanquin, Grifols, Octapharma and CSL-Behring (all paid to institution). BK is an Editorial Board member of Clinical Pharmacokinetics and was not involved in the selection of peer reviewers for the manuscript nor any of the subsequent editorial decisions. Ethics approval Not applicable. Consent to participate Not applicable. Consent for publication Not applicable. Availability of data and material Not applicable. Code availability Not applicable. Author contributions Conceptualisation: Shannon van der Zeeuw, Sander van Tilburg, Tim Preijers; Methodology: Shannon van der Zeeuw, Sander van Tilburg, Tim Preijers; Formal analysis and investigation: Shannon van der Zeeuw, Sander van Tilburg; Writing – original draft preparation: Shannon van der Zeeuw, Sander van Tilburg, Tim Preijers; Writing – review and editing: Shannon van der Zeeuw, Sander van Tilburg, Rose Crombag, Birgit Koch, Virgil Dalm, Tim Preijers, Bart Jacobs; Funding acquisition: none; Resources: Shannon van der Zeeuw, Sander van Tilburg; Supervision: Rose Crombag, Birgit Koch, Virgil Dalm, Tim Preijers, Bart Jacobs. References Associated Data Supplementary Materials Supplementary file (DOCX 102 KB)