2010 process drift chodankar

8/17/2019 2010 Process Drift Chodankar

1/38

PQRI-FDA Workshop onProcess Drift

Nandkumar Chodankar (Ph D Tech)Group CEO Pharma Business

Excel Industries Ltd.

Mumbai, India

Process Drift in theManufacturing of APIs

PQRI FDA


2/38

Main Discussions

Background –

Validation & Deviation, Cause - Effect

Drift-

Definition,

Cause of drift,

Categories of drift,

Multiple Effects of Drift

Monitoring and Controlling Drift

Examples from the API Industry

Dec. 1st 2010 2/38Nandkumar Chodankar (Ph D Tech)


3/38

Background: Prerequisites for theConsistent Process Validation

Validation should be about overall systemicintegrity.

For consistent product quality and output,the critical process parameters should beidentified during process development.

Critical process parameters may be defined asprocess step(s), process condition(s), test(s)

requirement(s) or other relevant parameter(s) oritem(s) that must be controlled withinpredetermined criteria (range) to ensure that theproduct (API) meets the defined specs.



4/38

Prerequisites for theProcess Validation

Qualification of the equipment & calibrationof the monitoring instruments – prior tovalidation.Qualification of the critical utility services

like water, steam, power, air, nitrogen, etc.Equipment cleaning process (validation)and sanitization.

All the documents relating to the processdevelopment, justification, critical stepsand parameters and relevant SOPs.Trained Personnel (Unit Process &Operations).



5/38

Critical Process Parameters

DoE is important for developing robustprocess and to identify the dependent vs.independent factors/parameters (riskprobability).

If the process capability is properlydelineated, the process should consistentlystay within the defined limits of its criticalprocess parameters and product specs.



6/38

Validated RangeDefined Limits

The variables due to chance cause usuallyremain within a defined narrow validatedrange and are controllable.

As long as these parameters vary withinthe normal or expected manner stablepattern of chance of variations develops.(normal curve or bell-shaped jar curve)

For example Temp. 90 ±2⁰C

+

-90⁰C 92⁰C88⁰C84⁰C 96⁰C

Target

T i m e



7/38

Validated Range

Alert Range

Worst Case (?)

Parameter – Establishing theRange (Design Space?)

Target

+-

- +

90⁰C 92⁰C88⁰C

87⁰C

85⁰C

93⁰C

95⁰C

(90⁰C ±2)

(90⁰C ±3)

(90⁰C ±5)


Working Range


8/38

Validation rangeBatch

Number

Temp

C

Rate of

Add.

l/min

React.

Time-

min.

Molar

Prop.

1 mole:

Impurities % conve

rsion

%


9/38

Deviation - Chance Cause

The possibilities of variations in any process aredue to the chance cause .

1. Quality of input material (specification range)

2. The equipment & instruments used (qualified)

3. The method of processing (unit operations)

4. The manufacturing conditions (parameters)

5. The method of testing (validated)6. The personnel involved (trained)



10/38

Assignable Vs.Non-assignable

Assignable cause: The variation can betraced to a known cause, such as

“negligence” or not adhering to an

approved method, not following approvedparameter, etc.

Process is in “a state of control” when

assignable cause have been eliminated andonly chance cause of variation is present.



11/38

Non-assignable cause A Concern

Non-assignable cause is a variation thatcannot be traced to a known cause.

When non-assignable cause is present in

addition to the chance cause the variationwill be excessive, and the process will beout of control . We need to reduce the non-assignable cause to control the process.

Non-assignable cause is the result ofincomplete study of the processrequirements.



12/38

Drift – Definition The gradual departure from an intended course

due to external influences - over time. A Ship or plane may drift in its course due to

unexpected circumstances.

Trending is one of the methodologies to trackdrift

P a r a m e t e r

Working Upper Range

Time (Batch No. 1 Data)

Validated Upper Range

Working Lower Range

Validated Lower Range

Alert

Alert

Target

Target



13/38

Process Drift & its Impact

“Process Drift” is the change of “NormalProcess Behavior” over time. This isobserved in most of the manufacturingprocesses.

Process drift may have a negative impact on

Quality of the product, yield, cost,reputation, etc.

Appropriate strategies must beimplemented to detect /monitor andaddress the process drift.



14/38

Risk Analysis – What Can Go wrong?

After Validation of the critical processparameters, one needs to continuously reviewthe trend of the batches for any drift (process

parameters vis-à-vis quality parameters).The risk of any such drift on the quality

parameters of the drug product, like impurityprofile, efficacy and safety - needs to be

evaluated.Use of statistical methods, DoE, determining

probability factors may be the approaches.



15/38

Risk Analysis – What Can Go wrong?

Risk analysis by asking questions.

1. What can go wrong?

2. If something goes wrong is there a way toknow that something has gone wrong, andbring the system back in compliance?

3. If there are no means to monitor or identify

this, what will be the effect on the quality ofthe product and the patient?



16/38

Strategy to Monitor & Control

The methodology to detect the drift and itseffect on the quality parameters is mostimportant in the lifecycle of the product .

Use of PAT is good to control drift and eliminate

manual adjustments or process tuning?

This will depend on the type of the drift that isexpected (or experienced during trend analysis).



17/38

What could be the causeof Process Drift?

Incomplete process development. Aging of the facility and the equipment.

Process pipeline scaling / contamination.

Out of calibration of the monitoringinstruments (Measuring- accuracy /sensitivity)

Fluctuations in the process parameters.

Changes in the raw material quality.

Indirect impact of the utilities & other factors.



18/38

Process Parameter Drift

Working Upper Range


Working Lower Range


Alert

Alert

Time

T e m p e

r a t u r e

R a t e o f A d d i t i o n

R e f l u x R a t e

Target

Target

Target



19/38

Multiple Parameters vs. Time

TargetWorking Upper Range


Working Lower Range


Alert

Alert

Multivariate Drift Effect

Time

P a r a

m e t e r s



20/38

Can we categorize Drifts?

1. Changing conditions can cause cyclical drift

Lack of preventive maintenance can affect theequipment and instruments used.

Recycling of catalyst / solvents.

Seasonal changes.2. Changes in the equipment (equivalent, new)

or raw material specifications (new suppliers).

3. Irregular drifts due to changes in thepersonnel, SOPs, equipment wear & tear, etc.

4. Linear Drift is seen for catalyst/solvent recycle


li i l if


21/38

Cyclic, Linear, Irregular Drift Pattern

Time

Target


TargetLinear Drift Pattern

Cyclic Drift Pattern

Target

Target

Irregular Drift Pattern

Comparison of Batches Trending


22/38

Comparison of Batches - Trending

Batch no. 1 no. 2 no. 3 no. 4 no. 5 no. 6 no. 7 no. 8

T e m p e r a t u r e

R a t e o f A d

d i t i o n

R e f l u x R a

t e

Target

Target

Target

no. 9



23/38

Combined /MultipleImpact of Drift

Many process variables mentioned maychange simultaneously and make itdifficult to evaluate their combined impact

on the quality. We need to generate dataof such drifts and their effect on thequality parameters (Risk Evaluation).

It is important to appropriately detect andclassify the type of drift so that correctivemeasures can be considered to counter /eliminate its effect.



24/38

Multivariate Effect Because of the multivariate process behavior

“Univariate based statistical process control (SPC)technology” is less useful to appropriately detectand classify the drift.

Some of the drift impact can be minimized by

1. Adequate, meaningful and appropriate trending

2. Shorter cycles of preventive maintenance.

3. Detecting & correcting drifts quickly and

minimizing the risk of defects in the quality.4. During process development, besides critical

parameters, reviewing other variables which may

have indirect combined effect on the quality.Dec. 1st 2010 24/38Nandkumar Chodankar (Ph D Tech)

C bi d Eff t Q lit


25/38

Combined Effect on Quality Aspect- Impurity Profile

0.2%

0.10%

0.05%0.025%

0.15%

Batch no. 1 no. 2 no. 3 no. 4 no. 5 no. 6 no. 7 no. 8 no. 9

Impurity -A Impurity -B Impurity -C Impurity –D- Unidentified


Batch no. 1 no. 2 no. 3 no. 4 no. 5 no. 6 no. 7 no. 8

T e m p e r a t u r

e

R a t e o f A d d i

t i o n

R e f l u x R a t e

Target

no. 9

D l t f iti l


26/38

Identify the following:1.Control Objective(s)

2.Input Variables: Classify as

a) Manipulated or b) Disturbance variablesInput may change continuously or at discreteintervals of time.

3.Output Variables: Classify asa) Measured or b) unmeasured variablesMeasurement may be continuous or at discrete intervalsof time

Development of criticalstrategy for process modeling


D l t f iti l


27/38

4. Constraints: Classify asa) Hard or b) Soft

5. Operating Characterization: Classify as

a) Continuous; b) Batch; or c) Semi-continuous6. Safety, Environmental and Economic

considerations

7. Control Structure: The controllers can be Feed-back or

Feed-forward in nature.



Development of critical


28/38


Process

Manipulated Inputs

Disturbance Inputs

Measured Outputs

Unmeasured Outputs

b) Control Representation

Process

Manipulated Inputs Measured Outputs

Unmeasured OutputsDisturbance Inputs

Controller

a) Input/Output Representation


M it i &


29/38

Monitoring &Control Strategy

PID Controller

PLC: Programmable Logic Controller, amicroprocessor-based electronic device for

implementing control algorithms.Time Domain

Ziegler-Nichols Tuning


Control Categories


30/38

Control CategoriesOperating Modes

Continuous

Batch

Semi-continuous

Operating Conditions

Start-up

Study state operation

Controlled shut down & Emergency shut down

Monitoring Maintenance

Optimization



31/38

Control Design

Operator controlled

Automatic Control

Alarms

Interlocks

Data logging

Automatic Tuning

SequencingProgrammable Logic Control

Field Monitoring



32/38

Study State Operation

System should be reliable with minimalsupervision

System should make “in-specification”

product routinelyMonitored by operator with a minimum

number of variables

Control loop interaction is automatic



33/38

PAT: On-line Control of Drift A few of the parameters can be monitored and

controlled using PAT - online instrumentation(Cause effect- control). In API Manufacturingplants many of the utilities, flow rates, etc., arecontrolled using SCADA /PLC which can minimize

drifts in the set process parameters. Crystal growth, shape of crystals, polymorphs can

be monitored and controlled using moderninstruments like NIR, Raman Spectra, X-rayFluorometers, etc.

However, a time lag between testing andmonitoring signal generation can result in a drift.


Ph t Chl i ti Pl t


34/38

Photo Chlorination Plant



35/38

Example 1- Impurity Profile DriftProduct 1-Two solvents system

Unidentified impurity (A)

I m p u r i t y –

( A )

%

Target 0.005

Working Upper Range

Batch Numbers

(Accepted Upper Range)

Detectable Lower Range

NMT 0.1%

0.02% Alert0.04%

0.01%

0.010.02 0.04

0.060.08



36/38

Example 1- Impurity Profile Drift

Product 1; Unidentified impurity (A) - NMT 0.1%

Two solvents system Critical Parameters:

Carbon treatment (before and after treatment results)

Distillation temperature & Rate of distillation andvolume distilled

Final volume and content of each / ratio

RPM & Temperature and rate of cooling

Final temperature range before filtration Filtration lots, washing, etc.

Impurity was within the acceptable limit but wasdrifting towards 0.1%


S


37/38

SummaryMost important is to have a robust process which is

possible by perfecting DoE Using Quality by Design,

Risk analysis,

PAT.

Batch-to-batch meaningful “Trend Analysis” to identify &monitor drift can help in taking corrective measures tominimize the drift effect on the quality. This can lead tocontinuous “State of Control” in the product lifecycle

Drift may be monitored and controlled by using moderninstruments like NIR, Raman Spectra, X-ray Fluorometer, withrapid testing approach.

However, combined effect, a time lag between signal/ testingand controls can still result in drift.



38/38

Thank [email protected]

[email protected]
mailto:[email protected]:[email protected]:[email protected]:[email protected]

2010 process drift chodankar

Documents