Decoding Drug Performance: A Comprehensive Review Of In Vitro–In Vivo Correlation (IVIVC) In Drug Development

Since a long time, the basic problem faced by the pharmaceutical industry has been prediction of the behaviour of the drug within the body by using simple in vitro tests. This has become possible only through the studies of In Vitro-In Vivo Correlation, also called as IVIVC, which serves as an important scientific bridge. This method involves linking a mathematical correlation between the in vitro observation and the actual happening within the body when the medicine is introduced in patient's body. According to FDA, the IVIVC model can be described as a predictive mathematical model relating an in vitro characteristic of the dosage form to a relevant in vivo performance⁽²⁾. Generally speaking, in vitro observation in such models includes the rate or extent of dissolution, whereas in vivo observation means the plasma drug concentration or cumulative drug absorption into blood stream respectively ⁽²⁾. As compared to FDA, USP defines the concept as establishing rational relationships between the biological characteristic generated by a pharmaceutical dosage form and physicochemical characteristics of drug ⁽³⁾.

IVIVC has a number of important features which make it interesting from both scientific and practical points of views. First, it does not mean any direct cause-effect relationship between in vitro observations and results obtained in the living body. Instead, IVIVC model establishes a predictive correlation which, when properly determined and proven scientifically, becomes a tool for predicting the effect of drug ⁽⁴⁾. IVIVC was introduced several decades ago. It was the early works by Edwards in 1951 and later Nelson in 1957 that were among the first to discover the links between dissolution rate of aspirin and theophylline and the corresponding concentrations of these drugs in patients' blood stream ⁽²⁾. Later, with increasing complexity of drug molecules and further sophistication of pharmaceutical industry, IVIVC became an important scientific and practical tool of development of poorly water-soluble drugs

Methods:

This review AIIMS to connect the old and new ideas about IVIVC . Trying to feel the gap between what we already know and what regulators are asking for. To do this, we looked carefully at the literature using important keywords like "deconvolution modelling"," level A correlation "and "bio pharmaceutical modelling" in pubeMed ,Google scholar, and science direct. We actually looked at articles and review papers from 1980 to 2025 to make sure we had both the old math ideas and the new trend like PBPK  and machin learning .we pick papers that sold good results from regulation or looked at the practical problems that come with complex dosage form the information we gathered was organised in a way that made sense starting with simple categories and moving on latest modelling technique.

Principle:

The idea of IVIVC is that a relationship can be observed between the in vitro rate of drug release and the in vivo rate of drug absorption ⁽⁵⁾ ‚ assuming the rate of dissolution of the drug from the dosage form is the rate-limiting step‚ particularly in the case of modified-release dosage forms .

IVIVC mainly correlates:

• In vitro parameter: Drug dissolution/release profile
• In vivo parameter: Fraction of drug absorbed or plasma drug concentration

Mathematical techniques such as deconvolution and convolution are commonly used to establish this relationship ^(6,7) . Deconvolution will help reach the in vivo absorption profile from plasma concentration-time curves‚ while convolution will help predict the plasma concentration from in vitro studies?If an appropriate IVIVC model could be developed then in vivo behavior could be predicted from in vitro tests‚ making it a powerful design tool in drug development?                                                        

Development, Evalution and  Recent Advances In IVIVC:

In vitro-in vivo correlation (IVIVC) is a predictive mathematical model establishing the association between an in vitro property of a dosage form (e?g? dissolution) and the in vivo response (e?g? plasma drug concentration or fraction of drug absorbed) in a patient? It is a commonly used pharmaceutical development tool as it allows for predicting in vivo behavior of a drug from in vitro methods and reduces the need for multiple time-consuming and expensive bioavailability and bioequivalence experiments ^(8,10)? According to the regulatory agencies‚ including the FDA‚ IVIVC is a quantitative relationship‚ which enables the prediction of biological performance from physicochemical characteristics and is useful for formulation evaluation‚ quality control‚ and regulatory flexibility ⁽⁸⁾?

IVIVC correlation can also be categorized as Level A‚ Level B‚ Level C‚ or Level D correlation? Level A is a point-to-point correlation between the in vitro dissolution and in vivo absorption profiles and is thought to provide the highest quality of information‚ and therefore‚ is more acceptable to the regulatory agency and more likely to grant a biowaiver? Level B correlation uses statistical moment analysis? Level C correlation establishes the relationship between a single dissolution parameter (like the amount released) and a pharmacokinetic parameter? Multiple Level C correlation improves predictability because more than one time point is considered? Level D correlation produces rank-order or qualitative correlation⁽¹⁰⁾?

IVIVC establishment requires formulation screening with different dissolution profiles‚ performing dissolution studies under physiological conditions‚ performing in vivo pharmacokinetics studies‚ and employing pharmacokinetic deconvolution methods to relate the dissolution and absorption profiles? An IVIVC is based on the physicochemical properties of the drug such as solubility and permeability using the Biopharmaceutics Classification System (BCS) ⁽¹⁰⁾? Dissolution is regarded as a drug product-dependent property‚ while pharmacokinetics (absorption‚ distribution‚ metabolism and excretion) are drug-dependent properties? The plasma concentration-time profile is therefore the result of dissolution and pharmacokinetics ⁽⁹⁾?

The most common method for establishing IVIVC uses prediction evaluation? This typically involves internal and external model validation? Internal model validation is performed by comparing predicted values of pharmacokinetic parameters (e?g? Cmax‚ AUC) to their observed values in the same dataset used to develop the model? External model validation is performed using independent datasets? Guidelines tell that the average prediction error (%PE) should not be more than 10% and individual prediction error should not exceed 15% (10)? %PE is the most commonly used performance measure?

In the pharmaceutical industry‚ IVIVC has several applications‚ including formulation development‚ dissolution specification setting‚ scale-up and post-approval changes‚ biowaivers‚ and as a quality control test that measures drug release and performance consistency within a drug product across multiple production lots? For example‚ IVIVC has been established for controlled-release formulations to determine dissolution specifications using deconvolution methods ⁽¹²⁾?

Difficulties in achieving an IVIVC may occur due to the dissolution not being the rate-limiting step in drug absorption or due to the in vitro conditions not simulating the gastrointestinal conditions? The industry also reported other difficulties pertaining to nonlinear pharmacokinetics‚ variability‚ and insufficient and/or poor quality of in vivo data as possible contributors to the relatively low success rates in IVIVC development ^(8,10)

More mechanistic approaches to IVIVC have also been developed including finite absorption time (FAT) and finite dissolution time (FDT) which take into account the time-limited nature of drug absorption to provide a more physiologically relevant description of drug absorption (8)? Several expansions of the IVIVC concept have also been made including Levy plots and time-scaled IVIVC to more easily present the relationship between in vitro and in vivo?One advance with great promise for helping the development of IVIVC is the use of physiologically based pharmacokinetic (PBPK) modeling approaches that predict absorption and other aspects of disposition by considering physiological processes‚ the drug‚ and the formulation ^(8,11)?

IVIVC Modelling:

While testing drugs, scientists tend to analyse both laboratory results and outcomes after drugs are ingested in human organisms. This relationship is made possible via a technique known as IVIVC. With help of mathematical formulas, the process of prediction of absorption rates becomes precise. Adjustments are done considering the hidden patterns detected by these methods. The government documents are backed by reliable estimations based on predictions. What appears to be research in laboratories will later be reflected in patient health.⁽¹⁵⁾

Generally speaking, IVIVC methods can be divided into two major categories: some depend on the development of models (compartments model being a part), other do not apply any models, and use the techniques such as deconvolution. Only a few of IVIVC rely on numeric analysis⁽²⁾

1. Deconvolution-Based Techniques: Going backwards usually leads you right here – deconvolution represents one of the most efficient methods of performing IVIVC, especially for Level A. Just imagine peeling off layers of the timeline – starting with blood plasma levels, predicting the rate at which drugs were absorbed, going backwards with help of certain data. Among the most common techniques of this kind include following approaches:

Wagner Nelson Approach :One of the possible ways of studying kinetics of drugs is called the Wagner–Nelson approach – this refers only to cases when only a single compartment was used for calculations.

Wagner-Nelson Equation is used for one-compartment method

F_a(t) =Ct+kt0tCtdt ke0∞Ctdt

F_a?(t) = fraction of drug absorbed

C_t? = plasma concentration at time t

 K_e? = elimination rate constant

The Loo–Riegelman method

This method is used when the drug follows multi-compartment kinetics:

F_a(t) = C_t+k_100tCtdt+k_120tXpdtk100∞Ctdt

F_a?(t) = fraction of drug absorbed

C_t? = plasma concentration at time

K₁₀,k₁₂= Rate constants

X_p= amount of drug in peripheral compartment

The way how medicine moves inside our  body gets turned into a sequence which shows  how much of the drug is absorbed in body and when it is absorbed. That flow line fits neatly against results from test tube studies, one moment at a time ^(2,13).

2. Convolution Techniques: Backward steps define deconvolution, while convolution moves ahead. Most often, this method checks whether our models hold up under testing After getting the forecasts, they get measured next to real plasma results, often through %PE scores ⁽¹⁴⁾. To confirm efficacy of drug  inside and outside the body, this technique helps us to  test if the IVIVC comes up to expectations when challenged.⁽¹⁶⁾

The convolution equation is used to predict the plasma concentration of drug which is given as,         

Ct=0t Iτ.Rt-τdτ

C(t)= plasma concentration

I(t)=input function (release/absorption)

R(t)= unit impulse response

3. Compartmental (Model-Dependent) Approaches: Inside the body, medicine travels through separate zones linked together - each step guided by the specific predictable patterns. These sections act like chambers, showing  how fast or slow the drug  spreads into the body . The  one-compartment models act like the whole human body is a single pool where medicines are mixed up evenly.Split into stages, multi-compartment models handle movement and removal separately. these models depend mostly on certain drug behaviour numbers - such as how fast it gets absorbed (Ka) and cleared out (Ke), along with where it spreads in the body - to match lab data to what actually  happens inside the body ^(13).

4. Statistical (Model-Free) Methods: If the fixed model of drug behaviour appears too rigid for you, these methods could be considered instead. They do not track the entire course but establish associations among general patterns based on statistics. In this case, the association reveals itself as the dependence of dissolution rate on retention rate on average. Timing comes into play when one parameter stands for the rate of drug dissolution, while another one represents the rate of retention in the body. Just one parameter makes an association in this particular case – the quantity of dissolved medicine directly correlates with its maximum plasma concentration. It may happen that the same parameter can be associated with mean retention time in the body. Here, the sole point does the trick instead of any curves. There is no need for any patterns; it is enough for a single match to exist. One step links to another with help of the timing method ⁽¹⁵⁾.

5. Modelling systems that are time-invariant and exhibit linear properties: Here, complex models used in IVIVC often borrow ideas from systems engineering, such as LTI (linear time-invariant) model, for example , consider that: human body is represented by dynamic system. In this case, the amount of drug released can be thought of as input, while blood concentration level can be regarded as the reaction to the input. The idea is to associate output with input using certain patterns ⁽¹³⁾. Thus, dissolution process along with absorption kinetics and disposition process are combined in one system.

6. Mechanistic and Physiologically Based Modelling: Starting to shift, modern modelling techniques now takes us into the  PBPK methods.When these models replicates the  actual body conditions, their assumptions  about behaviour of drug or medicine inside the body come out much more closer to that in  reality. Yet here's the catch - running these models well means gathering huge amount of information and with along with it needing strong computing models ⁽¹³⁾.

7. Regression and Curve Fitting Methods: At last, people often turn to former analysis when they want to match the  lab numbers with the actual ones based on how the body react. It connects the dots between test outcomes and physical changes seen in individuals. Starting strong, Level A links usually lean on linear regression to match how much medicine dissolves with how much the body takes in. When things curve instead of follow a slant, models shift toward nonlinear methods. High R² numbers signal trust - those near one mean the forecast holds firm ^[14]

IVIVC Level Classification:

The in vitro–in vivo correlation (IVIVC)  is majorly classified into four distinct  levels which helps to  provides a regulatory and scientific framework for analyzing how well in vitro dissolution data predict in vivo medication performance.  These levels  aren’t arbitrary; they come straight from FDA and EMEA guidances, and they tell us exactly how much information we’re capturing and how well we can predict what happens in the body from a simple dissolution test. Traditionally , IVIVC classification system consist of level A , level B,. level C and Multiple Level C and all these four levels varies with respect to structural complexity, depth of information, prognostic power etc^.[17][18]

These levels are arranged in a hierarchy based on their regulatory value, with Level A standing out as the most informative and the top choice for regulatory purposes like bioequivalence waivers. Detailed comparison is as follows

Table no. 1 : Level Classification Of IVIVC