QbD_LSSYB_DOE_PHARMA DEVELOPMENT OPTIMISATION_MCS_STUDY MATERIAL_156933 PDF

Pursullence GBS LLP उत्तिष्ठ भारतः presents BREATH APEX Insight Fellowship Program New York , USA CSSC, NEW YORK , USA ACCREDITED MULTIPLE CAPSTONE PROJECTS BASED INDUSTRIAL GLOBAL CERTIFICATION PROGRAM 1 Business Problem Scenario EmCare Pharmaceuticals is a Mumbai based leading pharmaceutical company specializing in the development and manufacturing of solid oral dosage forms founded by Dr. Amit Khare in 1976. EmCare have a strong commitment to delivering high-quality products to patients and ensuring regulatory compliance. However, in recent months, the company has been facing challenges related to the content uniformity of their solid dosage forms, which has affected product quality and safety, resulting in customer dissatisfaction. Content uniformity is a critical quality attribute (CQA) for solid oral dosage forms, ensuring that each unit contains the specified amount of active pharmaceutical ingredient (API) within acceptable limits. However, recent internal analysis have revealed that the content uniformity of EmCare Pharmaceuticals' solid dosage forms has been consistently above the desired target of RSD ≤ 2% for the past few months. Business Problem Scenario In the midst of the challenges faced by EmCare Pharmaceuticals regarding weak content uniformity in their solid dosage forms, a significant incident occurred when a VP, Operations, Mr. Stuart Paul, GMK Healthcares, UK, a platinum client contacted the company head to express their concerns about the product quality and safety issues. The client, a healthcare provider, had been administering EmCare Pharmaceuticals' solid oral dosage form to their patients for an extended period. “During routine quality checks and patient feedback analysis, our team detected inconsistencies in the content uniformity of the medication. This is happening over last few months quite consistently. It is a serious concerns that affect the efficacy and safety of the product, prompting us to reach out to you. You have to address the issue promptly and immediately”,Mr. Stuart to Dr. Lalit Soni, Head, Operations, EmCare Pharmaceuticals. Business Problem Scenario Client also emphasized the potential risks associated with such inconsistencies, including the possibility of patients receiving inadequate doses or experiencing unintended overdosing. The client also expressed concerns about the impact on patient health, the credibility of EmCare Pharmaceuticals' products. “Improve fast or else, we will think otherwise,” Mr. Stuart to Dr. Lalit. The incident served as a wake-up call for EmCare Pharmaceuticals, reinforcing their commitment to delivering high-quality medications & meeting customer expectations. It prompted the company to prioritize the implementation of a rigorous LEAN SIX SIGMA DOE driven Quality by Design (QbD). The complaint from the client had a significant impact on EmCare Pharmaceuticals' management, highlighting the urgency to address the weak content uniformity issue to ensure consistent product quality & patient safety. EmCare Pharmaceuticals is looking out for professionals with expertise of LEAN SIX SIGMA DOE driven Quality by Design (QbD). ” HOW TO ASSESS WHAT CUSTOMER WANT AND DELIVER THE SAME EVERY “single” TIME....? VOC ANALYSIS..!! VOC Is Voice Of Customers...Customers Always Speaks Out What they expect, Their Heart Or Emotions...So VOC Indirectly Is Voice Of Emotions / Expectations... For Pharma Industry Customer Can Include Patients, Regulatory Bodies, Healthcare / Pharma companies, Even Some Times Governments. They Represent People And Thus Concern Of People Health & Safety, Product Efficacy or Quality. VOC (Voice of Expectations or Emotions) In Life Sciences / Healthcare / Pharma Industry is Patient Health, Safety And Product Efficacy & Quality… All these are called Quality Target Product Profile(QTPP)… Quality Target Product Profile (QTPP): The Quality Target Product Profile (QTPP) is a therapeutic characteristics / Pharmacological action that a drug product should possess to meet its intended purpose i.e. curing certain medical condition. It outlines the desired therapeutic attributes that the formulation should aim to achieve. To Fulfill VOC, One Need To Work At Business End – I.E. On Product.. Let’s Call It Voice Of Product...In Other Words Companies Need To Keep Answering Following Question Constantly: To Honor Or Match Customer Expectations /Emotions- QTPP, What Do We Need To Improve In Product ? Critical Quality Attributes (CQAs) Critical Quality Attributes (CQAs) Critical Quality Attributes (CQAs) are specific physical, chemical, biological, or microbiological characteristics that are critical to ensure the QTPP i.e. quality and therapeutic performance of a drug product. They are the “measurable” characteristics of a product- drug / device that need to be controlled and monitored to ensure Patient Health, Safety, Product Efficacy & Quality… Examples Of Critical Quality Attributes (CQAs) 1.) Assay: The assay is a CQA that measures the concentration of the active pharmaceutical ingredient (API) in the tablet. It ensures that the tablet contains the specified amount of the API, which is crucial for its therapeutic efficacy. 2.) Dissolution Rate: The dissolution rate is a CQA that determines how quickly the tablet disintegrates and releases the API into the surrounding medium. It affects the bioavailability and onset of action of the drug. 3.) Content Uniformity: Content uniformity is a CQA that ensures the uniform distribution of the API within the tablet. It ensures that each tablet within a batch contains a consistent amount of the API, enabling accurate dosing and consistent therapeutic effect. 4.) Hardness, 5.) Disintegration Time, 6.) Stability However, Final Product-drug / device Is Output Of Something –I.E. Effective Input Material & Efficient Processing...It Is Called Voice Of Process... Hence To Improve Performance Of Final Product, Companies Always Need To Answer Following Question: To Improve Performance Of Final Product, What Should We Control / Optimize At Process Level i.e. material attributes and process parameters... CRITICAL MATERIAL ATTRIBUTES (CMAs) & CRITICAL PROCESS PARAMETERS (CPPs) Examples Of CMAs 1. Particle Size Distribution: It refers to the range of particle sizes present in a material or formulation. It can influence the product's performance, such as dissolution rate, bioavailability, flowability, and stability. 2. Moisture Content: Moisture Content refers to the amount of water or moisture present in a material or product. Moisture content is a critical material attribute as it can affect the stability, physical properties, and chemical reactivity of the material. Examples Of CPPs 1. Mixing Time: It refers to the duration or period for which the components of a formulation or mixture are blended together. Mixing time affects the uniformity and homogeneity of the mixture. 2. Compression Force: It refers to the force applied during tablet compression to form a compacted tablet. It influences the tablet's hardness, thickness, and disintegration characteristics, which in turn affect dissolution and drug release. Voice Of Customer (QTPPs) Depends On Voice Of Product (CQAs) That Further Depends On Voice Of Process (CMAs & CPPs) However Pharma Industry Is Highly Regulated Industry... Pharma Industry Need To Follow Quality Regulations Set By International Regulatory Bodies Through INTERNATIONAL CONFERENCE ON HARMONIZATION (ICH) OF TECHNICAL REQUIREMENTS FOR REGISTRATION OF PHARMACEUTICALS FOR HUMAN USE Thus Let’s Understand QBD In Eyes Of ICH..!! What is Quality by Design (QbD)? International Conference on Harmonization (ICH) Guidelines : Working with regulators in the European Union (the European Medicines Agency) and Japan, the US FDA has introduced Quality by Design objectives through the International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use. ICH Introduced Quality Guidelines: ICH Q8 (Pharmaceutical Development), ICH Q9 (Quality Risk Management), and ICH Q10 (Pharmaceutical Quality System). What is Quality by Design (QbD)? QBD is a systematic approach to Product development that begins with predefined objectives (QTPPs & CQAs) and emphasizes product and process understanding (CMAs & CPPs) and process control, based on sound science (DOE) and quality risk management. (ICH Q8 R2) Pharmaceutical Development - ICH Q8(R2) Product Profile ▸ Quality Target Product Profile (QTPP) CQA’s ▸ Determine “potential” critical quality attributes (CQAs) Risk Assessments ▸ Link raw material attributes and Material Attributes & Process Parameters to CQAs and perform risk assessment – i.e. find Critical - MAs / PPs ▸ Develop a design space – With Scientific & statistical foundation of DOE Design Space ▸ Design and implement a control strategy – find optimum levels of CMAs & Control Strategy CPPs to improve performance of CQAs at final product – drug / device level Continual Improvement ▸ Manage product lifecycle, including continual improvement Advantages of QbD to Pharma / Life Sciences Industry ▸ QbD applied to Drug substance development, Drug Product (ICH Q8 R2), Analytical method development. ▸ Like QbD is required for NDA submissions, since Jan’ 2013, FDA strongly recommends QbD in ANDA submissions & for Biopharmaceuticals products too. ▸ QbD Improve yields, lower cost, less investigations, reduced testing, etc. ▸ QbD Timely launch of products. ▸ QbD High Return on investment / cost savings. ▸ QbD Lead to Systematic development thus reducing Re-working & Rejections of the batches. ▸ QbD Significant reduction in regulatory oversight post approval. ▸ QbD Driver to Business Benefits. Since QbD is powered by Design of Experiments (DOE) , Statistical Process Control (SPC) & Quality Risk Management (QRM) with LEAN SIX SIGMA –DMADV pathway, We are here to learn LEAN SIX SIGMA YELLOW BELT – DOE For QbD Based Pharma Development Optimisation LET’S BEGIN WITH SIX SIGMA YELLOW BELT…. 21 DO YOU KNOW ? 22 INDEED.COM SALARY SURVEY Six Sigma certified individuals usually fall into the $100,000 + pay brackets, and are amongst the highest-paid professionals globally. The average salary for “ Six Sigma Black Belt" ranges from approximately $70,727 per year for Business Consultant to $129,996 per year for Director of Quality. Salary information comes from 93,588 data points collected directly from employees, users, and past and present job advertisements on Indeed.com in the past 36 months. https://www.indeed.com/salaries/six-sigma-black-belt-Salaries 23 These analytically driven problem solving expertise are high IN DEMAND all over the world…! Many Universities are encouraging their graduates and PG students learn these expertise and get in to the market with problem solving skill as an inherent part of their profile..… 30 31 32 33 34 35 36 37 38 39 40 Introduction to SIX SIGMA What is Six Sigma? Sigma is a Greek letter that represent the Standard deviation…i.e. amount of variation (Expected performance – Actual performance) in a process over the period of time. Six Sigma is a statistics based problem solving technique Invented By Bill Smith in Motorola in late 1980s. What is Six Sigma? Six Sigma Prioritize Significant Xs That Impact Y, Based On Statistical Validation. Six sigma is a such a high performance bar….99 out of 100 is not sufficient….Even 99.99 out of 100 is not sufficient.. It expects ONLY 3.4 defects / defective parts per million opportunities / attempts….i.e. performance need to be right for 99.99967 out of 100 times…i.e. 9,99,997 times of 10 lakhs attempts…. That High Standard!!! WHY IS SIX SIGMA THAT STRINGENT…. How can you say 99.99% is not enough….????? If this is the question…Let’s understand difference between 99% vs. 99.99% vs. Six Sigma…. Ask it to Cardiac Surgeon ….what if a cardiac surgeon is right for 99.99% in his life time…..it means 100 patients have lost their lives due to doctor’s error…will we accept it…. What if in country like India, railways are accurate for 99.99%..per year…there are at least 100 accidents per year…..and on an avg. 10,000 people are losing their lives… will we accept it….? Big no….!! And that’s why a benchmark as high as six sigma is needed. Its not a destination. It is a journey of continual improvement…. Remember customers don’t accept even 1 defective product…… right ? 44 Six Sigma Highlights… Hierarchy Of Lean Six Sigma ** MASTER BLACK BELT IS LIKE PhD IN FIELD OF BUSINESS EXCELLENCE ** BLACK BELT IS LIKE POSTGRADUATE IN FIELD OF BUSINESS EXCELLENCE ** EXPERIENCED GREEN BELT IS LIKE GRADUATE IN FIELD OF BUSINESS EXCELLENCE ** EXPERIENCED YELLOW BELT IS LIKE DIPLOMA IN FIELD OF BUSINESS EXCELLENCE ** FOR SAKE OF UNDERSTANDING APPROXIMATE RELEVANCE OF THE LEVELS IN BUSINESS EXCELLENCE WITH THAT OF ACADEMIC HIERARCHIES 45 46 47 Introduction to ADVANCED QbD LEAN SIX SIGMA – DMADV-C Advanced QbD-LEAN SIX SIGMA (DMADV-C) MEASURE DESIGN CONTROL (What are (How to (What &How (What Are (What Is An Prove/ How To Statistically Deep Is A Input ‘Optimum’ Validate An Control CQAs Significant Problem – Parameters Settings for Optimum By Ensuring Inputs [CMAs Output / [CMAs & [CMAs & Solution ?) CQAs Under & CPPs] CQAs ?) CPPs] That CPPs] To Statistical Improve Improve Control DEFINE ANALYZE VALIDATE Output CQAs ? CQAs?) Limits STATISTICAL PREDICTIVE QUALITY RISK DESIGN OF MONTE CARLO STATISTICAL CAPABILITY ANALYTICAL MANAGEMENT EXPERIMENTS SIMULATIONS PROCESS ONTROL ANALYSIS MODELINGS 49 LSS Step QBD LEAN Six Sigma Yellow Belt - End to End Problem Solving Steps Tool Phases # 1 Identify The Current Business Issue (Qualitative problem) VOC Analysis-complaint / Feed Back Log 2 Collect & Study Historical Data (Quantitative problem ) Historical Data Collection DEFINE 3 Calculate Baseline / Pre- Solutioning Pharma Process Capability (Statistical Problem) Capability Index - Parts per Million ( PPM) & Sigma Score (Zlt) 4 Compare Baseline Capability With Industry Benchmark & Set Improvement Goal Improvement Goal Setting 5 Identify List Of 'Controllable', Material Attributes & Process Parameters. CI Matrix & Fishbone Diagram 6 Prioritize 2 Key 'Impactable & Critical' Material Attributes & Process Parameters. And Their Current Settings. CI Matrix & SOP MEASU RE 7 Select 2 Levels For Prioritized Material Attributes & Process Parameters. : (Low (-) & High(+) Brainstorming 8 Design Combinations Of 2^2 (2 Factors & Their 2 Levels ) & Execute The Experiment With 3 Replications. Experimental Design & Data collection ANALYS Multivariate Modeling, Coefficient Of Determination- Data 9 Identify Statistically Significant Process Parameter And It’s Levels E Analysis, P Value , Main Effect & Interaction Effect plot, cube plot DESIGN 10 Design Optimum Combination Of Statistically Significant Process Input Parameters And Their Levels. Process Optimisation-DOE: Prescriptive and Predictive analytics 11 Based on Optimum Values of CMAs/ CPPs, deploy scenario analysis. Scenario Analysis. 12 Deploy Monte Carlo Simulations for different scenarios of CMAs / CPPs. Probability Distribution And Monte Carlo Simulations VALIDA TE 13 Validate CQA Optimisation Based On: Study sensitivity of CQA for diff. Scenarios and study related Risk. Sensitivity Analysis,, Statistical Risk Assessment 14 Validate CQA Optimisation Based On: Statistical Process Control SPC- X bar & R charts 15 Control Phase Data Collection Data Collection 16 Statistical Process Control (SPC) Control Chart CONTR OL Capability Index - Parts per Million ( PPM) & Sigma Score (Zlt), Bar 17 Control Phase Capability Analysis & Improvement Sustenance Proofing Chart 18 Project Sign Off & Closure Lessons Learnt & Documenting DEFINE PHASE “Define What is the Problem ?” 51 DEFINE PHASE STEP 1.1 : Identify The Current Business Issue. “Study Current Client Complaints and understand what client expect.” 52 LSS Step QBD LEAN Six Sigma Yellow Belt - End to End Problem Solving Steps Tool Phases # 1 Identify The Current Business Issue (Qualitative problem) VOC Analysis-complaint / Feed Back Log DEFINE MEASURE ANALYSE DESIGN VALIDATE CONTROL Identify the current business issue / Problem “ During routine quality checks and patient feedback analysis, our team detected inconsistencies in the content uniformity of the medication. This is happening over last few months quite consistently. It is a serious concerns that affect the efficacy and safety of the product, prompting us to reach out to you. You have to address the issue promptly and immediately…. - Mr. Stuart Paul, Vice President, GMK HealthCares Identify the current business issue / Problem In current project, Performance of ‘content uniformity’ (CU) for solid dosage form-tablets is not appropriate for last few months / batches. This affect patient health and satisfaction (QTPP) for client of Emcare Pharma. Content uniformity % RSD ( Relative Std. Deviation) is NOT COMPLYING TO THE STANDARD OF < Or = 2%. Content uniformity (CU) need to be optimised i.e. % RSD need to be controlled. Identify the current business issue / Problem Under QbD regime, CU is a Critical Quality Attribute (CQA). To optimize the performance of the CU, we need to find out Critical Material and Processing parameters (CMAs & CPPs) that affect the performance of CQA and their optimum setting/levels. If those CMAs & CPPs are not set optimally / correctly, This can RISK the performance of CU. (QbD Element # 1: Quality Risk Management). Hence, we need to find out scientifically correct “settings / levels” for CMAs & CPPs to improve / optimize performance of CU with Design of Experiments (DOE) (QbD Element # 2) , and sustain the optimum performance of CQA with statistical process control (SPC) QbD Element # 3 to avoid any Non compliances (NCs) in future. All this will be executed under QbD LEAN SIX SIGMA DMADV framework….!! LSS Step QBD LEAN Six Sigma Yellow Belt - End to End Problem Solving Steps Tool Phases # 1 Identify The Current Business Issue (Qualitative problem) VOC Analysis-complaint / Feed Back Log 2 Collect & Study Historical Data (Quantitative problem ) Historical Data Collection DEFINE MEASU RE ANALYS E DESIGN VALIDA TE CONTR OL DEFINE PHASE STEP 2.1 : Collect & Study historical Data, calculate Baseline / Pre-Solutioning process capability & compare it with industry benchmark. “Baseline data analysis to assess where are we standing currently.” 58 59 BATCHES WITH TOTAL NO. OF % CONTENT CU (% RSD) >2% "C" BATCHES BATCHES UNIFORMITY-NC ("NC" BATCHES) 100 45 55 45.00 Baseline / Historical % CONTENT UNIFORMITY-NC for last 3 months’ is 45 %. % CONTENT UNIFORMITY-NC of 45% is less or very less or high or very high..?? How to find it out ? 60 DEFINE PHASE STEP 1.3 : “ Calculate Historical / Pre- Solutioning Process Capability...” Use Capability Indices to Capture Process Capability 61 LSS Step QBD LEAN Six Sigma Yellow Belt - End to End Problem Solving Steps Tool Phases # 1 Identify The Current Business Issue (Qualitative problem) VOC Analysis-complaint / Feed Back Log 2 Collect & Study Historical Data (Quantitative problem ) Historical Data Collection DEFINE 3 Calculate Baseline / Pre- Solutioning Pharma Process Capability (Statistical Problem) Capability Index - Parts per Million ( PPM) & Sigma Score (Zlt) MEASU RE ANALYS E DESIGN VALIDA TE CONTR OL Answer : Process Capability Analysis 63 Process Capability is how ‘capable’ a process performance is..! In other words, how efficiently process is meeting client expectations CONSISTENTLY or in simple words, where are you standing against client set expectations…..!! Process capability is calculated in various capability indices like DPMO, PPM, Z lt, Z st, Cp, Cpk, Pp, Ppk…. Here, we will use a.) DPMO ( Defect Per Million Opportunities ) / PPM Parts per million & b.) Zlt (Sigma Long Term). 64 Calculation of DPMO or also called PPM* = DPO * 10,00,000 DPO or PPU = No. Of Defective (i.e. Non meeting targets) / Total No. Of Responses Industrial Benchmark For DPMO: DPMO should be < 5,000 ; Avg. Industry DPMO = 25000 – 35000 65 Baseline Capability In DPMO = DPO * 10,00,000 DPO or DPU = No. Of NCs / Total No. Responses DP0 or DPU = 45 / 100 = 0.45 Hence, Baseline Capability In DPMO = 0.45 * 10,00,000 = 4,50,000 Sigma Long Term (Z Lt) = NORMSINV(1-DPO) =NORMSINV(1-0.45) = 0.1257 66 LSS Step QBD LEAN Six Sigma Yellow Belt - End to End Problem Solving Steps Tool Phases # 1 Identify The Current Business Issue (Qualitative problem) VOC Analysis-complaint / Feed Back Log 2 Collect & Study Historical Data (Quantitative problem ) Historical Data Collection DEFINE 3 Calculate Baseline / Pre- Solutioning Pharma Process Capability (Statistical Problem) Capability Index - Parts per Million ( PPM) & Sigma Score (Zlt) 4 Compare Baseline Capability With Industry Benchmark & Set Improvement Goal Improvement Goal Setting MEASU RE ANALYS E DESIGN VALIDA TE CONTR OL DEFINE PHASE STEP 1.4 : Improvement Goal Setting... “Set a Goal...” 68 BASELINE PERFORMANCE ANALYSIS & IMPROVEMENT GOAL SETTING A. Baseline % NC : 45 % B. Baseline Capability In DPMO :- DPO * 10,00,000 = 0.45*10,00,000 = 4,50,000 C. Baseline Sigma Lt = NORMSINV(1-DPO) = NORMSINV(1-0.45) = 0.1257 D. Industry Benchmark For SIGMA Score : > 4 E. Avg. Industry SIGMA SCORE : 2 To 3 F. LSSYB Project Improvement Goal Setting : “ To Improve Current Sigma Score From 0.1257 To 1 By Specific timeline” ( Ex: + 3 months from current month.) 69 BASELINE PERFORMANCE ANALYSIS & IMPROVEMENT GOAL SETTING BASELINE / HISTORICAL PHASE DPMO /PPM & SIGMA SCORES # OF BATCHES WITH CU (% (DEFECTIVE) PARTS PER (DEFECTIVE) PARTS PER UNIT TOTAL # OF BATCHES RSD) > 2% MILLION SIGMA SCORES (DPU) (NC BATCHES) (DPMO OR PPM) a b b/a (b/a)*1000000 NORMSINV(1-(b/a)) 100 45 0.45 4,50,000 0.1257 1.) To Reduce Current DPMO From -------- to ------- By---- 2024. (In Next 3 Months ) GOAL STATEMENT 2.) To Increase Current SIGMA SCORE From -------- to ------- By---- 2024. (In Next 3 Months ) 70 MEASURE PHASE “Identify Critical Material Attributes & Critical Process Parameters Impacting Problem , Their Experimental Levels And Execute Experiments..” 71 LSS Step QBD LEAN Six Sigma Yellow Belt - End to End Problem Solving Steps Tool Phases # 1 Identify The Current Business Issue (Qualitative problem) VOC Analysis-complaint / Feed Back Log 2 Collect & Study Historical Data (Quantitative problem ) Historical Data Collection DEFINE 3 Calculate Baseline / Pre- Solutioning Pharma Process Capability (Statistical Problem) Capability Index - Parts per Million ( PPM) & Sigma Score (Zlt) 4 Compare Baseline Capability With Industry Benchmark & Set Improvement Goal Improvement Goal Setting 5 Identify List Of 'Controllable', Material Attributes & Process Parameters. CI Matrix & Fishbone Diagram MEASU RE ANALYS E DESIGN VALIDA TE CONTR OL A. IDENTIFY LIST OF 'CONTROLLABLE' MATERIAL ATTRIBUTES & PROCESS PARAMETERS In measure phase, we identify key factors - material or process factors or parameters that can have significant impact upon Project Y – Critical Quality Attribute (CQA),in current case it is Content Uniformity ( CU). IN OTHER WORDS, WE WOULD LIKE TO KNOW FACTORS THAT- IF WE WILL NOT SET OPTIMALLY, THERE WILL BE SERIOUS RISK OF CQA CROSSING COMPLYING LIMIT I.E. GOING OUT OF CONTROL. Thus Measure phase focuses on Quality Risk Management, one of the key element of QbD. With this goal in a mind, Process optimization and QbD DOE specialist will request process head / manager to call upon an meeting that includes senior members from the process- supervisors, asst. manager, and finally few of the senior employees who are working in the process for few years, who know the drug development process and its key material and process parameters / factors in details. 73 A. IDENTIFY LIST OF 'CONTROLLABLE' MATERIAL ATTRIBUTES & PROCESS PARAMETERS You will brief them the current problem with CQA ( In current case- Content uniformity-project Y) that need to be optimised – increased or decreased. You will ask the participants to comes up with CONTROLLABLE process parameters -input material, method / processing and machine parameters. CONTROLABLE parameters are those parameters that 1.) has std. numerical value, they have currently set based on guideline instructed in the std. operating procedure (SOPs) document. 2.) We Can Change Their Settings easily For The Purpose Of Experiment with out much of financial burden or administrative burden -approvals. 74 LIST OF PROBABALE PROCESS VARIABLES / PARAMETERS THAT IMPACT CRITICAL QUALITY ATTRIBUTE (CQA) CONTROLLABLE IMPACTABLE CRITICAL LIST DOWN THOSE MAs/ PPs N/3 # THAT FULFILLS 2 CONDITIONS ( EVEN SMALLEST DEVIATION Prioritisation CATEGORY OF VARIABLE- FROM ITS OPTIMUM LEVEL CAN (IS GIVEN PARAMETER 1. THAT IS SET AT THE SPECIFIC MA OR PP SEVERLY AFFECT PROJECT Y?) NEED TO BE WORKED 'NUMERICAL' VALUE. UPON IMMEDIATELY / 2. CAN EASILY CHANGE THIER SETTINGS (** IF NOT SURE WHETHER HIGH URGENTLY - YES / NO ) FOR THE PURPOSE OF EXPERIMENT. OR LOW, CONSIDER IT LOW ) 1 Material Attributes Particle Size Distribution 2 Material Attributes Moisture Content 3 Material Attributes Purity 4 Material Attributes Bulk Density 5 Material Attributes Flowability 6 Process Parameters Mixing Time 7 Process Parameters Compression Force 8 Process Parameters Drying Temperature 9 Process Parameters Granulation Moisture Content 10 Process Parameters Coating Inlet Air Temperature 75 LSS Step QBD LEAN Six Sigma Yellow Belt - End to End Problem Solving Steps Tool Phases # 1 Identify The Current Business Issue (Qualitative problem) VOC Analysis-complaint / Feed Back Log 2 Collect & Study Historical Data (Quantitative problem ) Historical Data Collection DEFINE 3 Calculate Baseline / Pre- Solutioning Pharma Process Capability (Statistical Problem) Capability Index - Parts per Million ( PPM) & Sigma Score (Zlt) 4 Compare Baseline Capability With Industry Benchmark & Set Improvement Goal Improvement Goal Setting 5 Identify List Of 'Controllable', Material Attributes & Process Parameters. CI Matrix & Fishbone Diagram 6 Prioritize 2 Key 'Impactable & Critical' Material Attributes & Process Parameters. And Their Current Settings. CI Matrix & SOP MEASU RE ANALYS E DESIGN VALIDA TE CONTR OL B. PRIORITIZE KEY 'IMPACTABLE & THEN 2 CRITICAL' PROCESS PARAMETERS AND THEIR CURRENT SETTINGS Out of the List of ‘CONTROLABLE’ Mas / PPs , ask meeting participants to prioritize highly ‘IMPACTABLE’ parameters that affect / impact the performance of Project Y. In other words, Project Y is highly sensitive towards these impactable Parameters so much that even a small/ fractional change in settings of these parameters can significantly affect performance of Project Y. Participants can either choose these parameters with ‘consensus’ and if consensus among them is not possible towards selection of the parameters then ‘majority’ among them. Ex: if there are 6 participants in meeting /brainstorming session, out of 6 members, at least 4 should find given parameter highly impactable, if not, that parameter is filtered out as low impactable parameter. Finally, choose 2 highly CRITICAL process parameters from highly IMPACTABLE parameters. Critical parameters are those parameters that need to be worked upon immediately / urgently. 77 LIST OF PROBABALE PROCESS VARIABLES / PARAMETERS THAT IMPACT CRITICAL QUALITY ATTRIBUTE (CQA) CONTROLLABLE IMPACTABLE CRITICAL LIST DOWN THOSE MAs/ PPs N/3 # THAT FULFILLS 2 CONDITIONS ( EVEN SMALLEST DEVIATION Prioritisation CATEGORY OF VARIABLE- FROM ITS OPTIMUM LEVEL CAN (IS GIVEN PARAMETER 1. THAT IS SET AT THE SPECIFIC MA OR PP SEVERLY AFFECT PROJECT Y?) NEED TO BE WORKED 'NUMERICAL' VALUE. UPON IMMEDIATELY / 2. CAN EASILY CHANGE THIER SETTINGS (** IF NOT SURE WHETHER HIGH URGENTLY - YES / NO ) FOR THE PURPOSE OF EXPERIMENT. OR LOW, CONSIDER IT LOW ) 1 Material Attributes Particle Size Distribution LOW 2 Material Attributes Moisture Content HIGH YES 3 Material Attributes Purity LOW 4 Material Attributes Bulk Density HIGH 5 Material Attributes Flowability HIGH 6 Process Parameters Mixing Time HIGH 7 Process Parameters Compression Force HIGH 8 Process Parameters Drying Temperature HIGH 9 Process Parameters Granulation Moisture Content HIGH YES 10 Process Parameters Coating Inlet Air Temperature HIGH 78 LSS Step QBD LEAN Six Sigma Yellow Belt - End to End Problem Solving Steps Tool Phases # 1 Identify The Current Business Issue (Qualitative problem) VOC Analysis-complaint / Feed Back Log 2 Collect & Study Historical Data (Quantitative problem ) Historical Data Collection DEFINE 3 Calculate Baseline / Pre- Solutioning Pharma Process Capability (Statistical Problem) Capability Index - Parts per Million ( PPM) & Sigma Score (Zlt) 4 Compare Baseline Capability With Industry Benchmark & Set Improvement Goal Improvement Goal Setting 5 Identify List Of 'Controllable', Material Attributes & Process Parameters. CI Matrix & Fishbone Diagram 6 Prioritize 2 Key 'Impactable & Critical' Material Attributes & Process Parameters. And Their Current Settings. CI Matrix & SOP MEASU RE 7 Select 2 Levels For Prioritized Material Attributes & Process Parameters. : (Low (-) & High(+) Brainstorming ANALYS E DESIGN VALIDA TE CONTR OL C. SELECT 2 LEVELS FOR PRIORITIZED PROCESS PARAMETERS :- (LOW (-) & HIGH(+) Each controllable parameters have their own current numerical standard setting as per the guideline of Standard Operating Procedures (SOPs) / Pharmacopeia Guidelines. For experimental purpose, select 2 levels around current std. settings. Low (-) and High (-). Based on discussion and expert views of meeting participants, they need to select 2 levels for each parameter that will be considered for an experimental design. Ex: CRITICAL CURRENT LEVELS # PARAMETERS SETTINGS Low (-1) High(+1) 1 Bulk Density 0.8-1.0 g/mL 0.8 g/mL 1.0 g/mL 2 Mixing Time 10-20 minutes 10 minutes 20 minutes 80 LSS Step QBD LEAN Six Sigma Yellow Belt - End to End Problem Solving Steps Tool Phases # 1 Identify The Current Business Issue (Qualitative problem) VOC Analysis-complaint / Feed Back Log 2 Collect & Study Historical Data (Quantitative problem ) Historical Data Collection DEFINE 3 Calculate Baseline / Pre- Solutioning Pharma Process Capability (Statistical Problem) Capability Index - Parts per Million ( PPM) & Sigma Score (Zlt) 4 Compare Baseline Capability With Industry Benchmark & Set Improvement Goal Improvement Goal Setting 5 Identify List Of 'Controllable', Material Attributes & Process Parameters. CI Matrix & Fishbone Diagram 6 Prioritize 2 Key 'Impactable & Critical' Material Attributes & Process Parameters. And Their Current Settings. CI Matrix & SOP MEASU RE 7 Select 2 Levels For Prioritized Material Attributes & Process Parameters. : (Low (-) & High(+) Brainstorming 8 Design Combinations Of 2^2 (2 Factors & Their 2 Levels ) & Execute The Experiment With 3 Replications. Experimental Design & Data collection ANALYS E DESIGN VALIDA TE CONTR OL D. DESIGN COMBINATIONS OF 2^2 (2 FACTORS & THEIR 2 LEVELS ) & EXECUTE THE EXPERIMENT WITH 3 REPLICATIONS DESIGN EXPERIMENTAL COMBINATIONS: Based on the formula of Levels (L) Raised To Factors (F), find out how many no. of combinations, which is also called ‘Runs’ need to be generated to deploy experiment with scientific foundation. There is diff. between "Trial and Errors" that is based on intuitions and Scientific experimental design. In current simulated project, we have 2 levels and 2 factors, hence total no. of runs / combinations will be 4 i.e. 2 levels raised to 2 factors. In experimental science, there is a concept called ‘Replications’. To enhance the accuracy of experiments- we ‘re-do' entire set of experiment. This process of re-doing of entire set of experiment is called Replications. Experimental accuracy is directly proportional to no. of replications. However, while increasing no. of replications, its impact of experimental duration and cost of experimentation need to be taken in to consideration. In current simulated project, set of experiments is replicated for 3 times, i.e. total of 4 runs per experiment * 3 = 12 runs. Algorithm / formulae in Pursullence system will automatically generate 12 combinations of different levels for 2 Input / process factors. 82 D. DESIGN COMBINATIONS OF 2^2 (2 FACTORS & THEIR 2 LEVELS ) & EXECUTE THE EXPERIMENT WITH 3 REPLICATIONS EXPERIMENTAL DESIGN FACTOR 1 : Particle Size Distribution FACTOR 2 : Compression Force REPLICATIONS (μm) (kN) 10 15 20 15 I 10 25 20 25 20 25 10 25 II 20 15 10 15 10 15 20 15 III 10 25 20 25 83 ANALYSE PHASE “ Analyze Statistically Significant Process Parameters and Their Levels..” 84 LSS Step QBD LEAN Six Sigma Yellow Belt - End to End Problem Solving Steps Tool Phases # 1 Identify The Current Business Issue (Qualitative problem) VOC Analysis-complaint / Feed Back Log 2 Collect & Study Historical Data (Quantitative problem ) Historical Data Collection DEFINE 3 Calculate Baseline / Pre- Solutioning Pharma Process Capability (Statistical Problem) Capability Index - Parts per Million ( PPM) & Sigma Score (Zlt) 4 Compare Baseline Capability With Industry Benchmark & Set Improvement Goal Improvement Goal Setting 5 Identify List Of 'Controllable', Material Attributes & Process Parameters. CI Matrix & Fishbone Diagram 6 Prioritize 2 Key 'Impactable & Critical' Material Attributes & Process Parameters. And Their Current Settings. CI Matrix & SOP MEASU RE 7 Select 2 Levels For Prioritized Material Attributes & Process Parameters. : (Low (-) & High(+) Brainstorming 8 Design Combinations Of 2^2 (2 Factors & Their 2 Levels ) & Execute The Experiment With 3 Replications. Experimental Design & Data collection ANALYS Multivariate Modeling, Coefficient Of Determination- Data 9 Identify Statistically Significant Process Parameter And It’s Levels E Analysis, P Value , Main Effect & Interaction Effect plot, cube plot DESIGN VALIDA TE CONTR OL END TO END USER MANUAL FOR DEPLOYMENT OF PURSULLENCE’S DOE- 2^2 PROCESS OPTIMISATION SYSTEM 1 2 3 4 Once Done With Step 1 & 2, Algorythm In Select 2 ‘Factors’ And Pursullence Process Select Drop Down For Their 2 ‘Levels’ And Add It Optimisation -DOE Add Project Y Project Y should be Either In Top Left Corner ‘Factor Analytics System Will "Increased / Decreased" Table’ Automatically Create 12 Combinations In Design Table. 8 7 6 5 Statistical Algorithms Select 1 Of 10 Datasets Of Check R-Sqr. And Then R- Within System Will Paste 12 Data points In 12 Combinations Shared Sqr. Adj. Value, Both Populate Descriptive & Final Column Of Design In The Tab Of Should Be In Green, or Diagnostic Analytical Table. 'Experimental Outcome' atleast R-Sqr. Adj > 50% Inferences. From Project Data File. 86 END TO END USER MANUAL FOR DEPLOYMENT OF PURSULLENCE’S DOE- 2^2 PROCESS OPTIMISATION SYSTEM 9 10 11 12 If R-Sqr. Adj is NOT > 50% Factors (A Or B) / Check Main Effect Plot , You May Check Another If In Green, Check P Interaction (A*B) With 'P' Whether Statistically Dataset Of 12 Values For Both Factors Value Less Than 0.05 (In Significant Factor/s Have Combinations From Step And Interaction. Red), Is Statistically Direct Or Inverse Impact 3. Significant. On CQA. 15 14 13 Acceptable Means, If Check Cube /Square Check Interaction Plot Project Y Need To Be Plot And Assess Levels And See Whether Line Of Improved, Select Highest Of Both Factors, At Factor 'A' Deeply/ Lightly Of 4 Values Of Project Y. Which, Project Y Is Intersect Line Of Factor If Project Y Need To Be "Acceptable“ in current 'B' Or Both Lines Are Reduced, Select Lowest case LOWEST CU value Almost Parallel Or Exactly Of 4 Value Of Project Y. among all 4 values. Parallel. 87 SNAPSHOTS OF PURSULLENCE’S DOE- 2^2 PROCESS OPTIMISATION SYSTEM 88 SNAPSHOTS OF PURSULLENCE’S DOE- 2^2 PROCESS OPTIMISATION SYSTEM INFERENCES : Since R square Adjusted value is > 50%, we can say, that the power of a statistical model developed through DOE IS >50%, Power OF what ? Power of predicting CU-% RSD value with the experiment based on selected 2 factors : PSD & CF is moderately high. R SQUARE and R SQUARE ADJ. values should be higher…Higher the better. 89 SNAPSHOTS OF PURSULLENCE’S DOE- 2^2 PROCESS OPTIMISATION SYSTEM P value for Compression Force is 0.0251, it means, p value * 100 = 0.0251 * 100 = 2.51 % Risk to say Compression Force is Critical factor that can impact CQA-Content Uniformity. 100 - % Risk = % Guaranty / % Assurance. So for Compression Force, 100 - 2.51 = 97.49 % guaranty to say, Compression force is critical factor. This assurance / guaranty should be more than 95%. It means % risk should be less than 5%, i.e. p value should be less than 0.05. Pursullence DOE Analytics Software Tool Highlight such a factors in RED. 90 SNAPSHOTS OF PURSULLENCE’S DOE- 2^2 PROCESS OPTIMISATION SYSTEM INFERENCES : 1.) In DOE analysis, when PSD is set at level 1 (10 μm), CU % RSD is a 1.44, When PSD is set at level 2 (20 μm), CU % RSD is 1.19, it means CU % RSD is lowered at PSD Of 20 μm. CU favorability is inclined to Level 2 of PSD. 91 SNAPSHOTS OF PURSULLENCE’S DOE- 2^2 PROCESS OPTIMISATION SYSTEM INFERENCES : 2.) In DOE analysis, when Compression Force (CF) is set at level 1 (15 kN), CU -% RSD is a 1.56, When CF is set at level 2 (25 kN),, CU % RSD is a 1.07, it means CU % RSD is lowered at CF level 2, i.e. 25 kN. CU favorability is inclined to Level 2 of Compression Force. 92 SNAPSHOTS OF PURSULLENCE’S DOE- 2^2 PROCESS OPTIMISATION SYSTEM INFERENCES : 3.) Since 2 lines are not crossed in Interaction plot, There is no Interaction between PSD and Compression Force 93 SNAPSHOTS OF PURSULLENCE’S DOE- 2^2 PROCESS OPTIMISATION SYSTEM II III I IV II : FACTOR 1 – LEVEL 1; FACTOR 2- LEVEL 2 III : FACTOR 1 – LEVEL 2; FACTOR 2- LEVEL 2 I : FACTOR 1 – LEVEL 1; FACTOR 2- LEVEL 1 IV : FACTOR 1 – LEVEL 2; FACTOR 2- LEVEL 1 94 DESIGN PHASE “ Design Optimum Levels Of Process Parameters To Be Implemented...” 95 LSS Step QBD LEAN Six Sigma Yellow Belt - End to End Problem Solving Steps Tool Phases # 1 Identify The Current Business Issue (Qualitative problem) VOC Analysis-complaint / Feed Back Log 2 Collect & Study Historical Data (Quantitative problem ) Historical Data Collection DEFINE 3 Calculate Baseline / Pre- Solutioning Pharma Process Capability (Statistical Problem) Capability Index - Parts per Million ( PPM) & Sigma Score (Zlt) 4 Compare Baseline Capability With Industry Benchmark & Set Improvement Goal Improvement Goal Setting 5 Identify List Of 'Controllable', Material Attributes & Process Parameters. CI Matrix & Fishbone Diagram 6 Prioritize 2 Key 'Impactable & Critical' Material Attributes & Process Parameters. And Their Current Settings. CI Matrix & SOP MEASU RE 7 Select 2 Levels For Prioritized Material Attributes & Process Parameters. : (Low (-) & High(+) Brainstorming 8 Design Combinations Of 2^2 (2 Factors & Their 2 Levels ) & Execute The Experiment With 3 Replications. Experimental Design & Data collection ANALYS Multivariate Modeling, Coefficient Of Determination- Data 9 Identify Statistically Significant Process Parameter And It’s Levels E Analysis, P Value , Main Effect & Interaction Effect plot, cube plot DESIGN 10 Design Optimum Combination Of Statistically Significant Process Input Parameters And Their Levels. Process Optimisation-DOE: Prescriptive and Predictive analytics 11 Based on Optimum Values of CMAs/ CPPs, deploy scenario analysis. Scenario Analysis. 12 Deploy Monte Carlo Simulations for different scenarios of CMAs / CPPs. Probability Distribution And Monte Carlo Simulations VALIDA TE 13 Validate CQA Optimisation Based On: Study sensitivity of CQA for diff. Scenarios and study related Risk. Sensitivity Analysis,, Statistical Risk Assessment 14 Validate CQA Optimisation Based On: Statistical Process Control SPC- X bar & R charts 15 Control Phase Data Collection Data Collection 16 Statistical Process Control (SPC) Control Chart CONTR OL Capability Index - Parts per Million ( PPM) & Sigma Score (Zlt), Bar 17 Control Phase Capability Analysis & Improvement Sustenance Proofing Chart 18 Project Sign Off & Closure Lessons Learnt & Documenting END TO END USER MANUAL FOR DEPLOYMENT OF PURSULLENCE’S DOE- 2^2 PROCESS OPTIMISATION SYSTEM 16 17 18 Prescriptive Analytics Algorithm In Pursullence Process Optimisation - Based On Experimental Algorithm In Pursullence Process DOE Analytics System Will Allow Performances And Prescribed Levels Optimisation -DOE Analytics System You To Choose Optimum Levels For Of Factors, System's Predictive Will Automatically Build Regression Both Factors. Analytical Algorithm Will Predict Model At Back End. Optimum Performance Of Project Y. 19 Once Project Y Is Optimised Experimentally, Next It Can Be END Implemented Operationally And Validate Project Y Optimisation. 97 SNAPSHOTS OF PURSULLENCE’S DOE- 2^2 PROCESS OPTIMISATION SYSTEM Ensure PRESCRIBED VALUES OF CMA & CPP ( In above case- PSD & Compression Force are Feasible to set and Cost Effective. 98 LSS Phases Step # Six Sigma Yellow Belt - End to End Problem Solving Steps Tool 1 Identify The Current Business Issue (Qualitative problem) VOC Analysis-complaint / Feed Back Log 2 Collect & Study Historical Data (Quantitative problem ) Historical Data Collection DEFINE 3 Calculate Baseline / Pre- Solutioning Pharma Process Capability (Statistical Problem) Capability Index - Parts per Million ( PPM) & Sigma Score (Zlt) 4 Compare Baseline Capability With Industry Benchmark & Set Improvement Goal Improvement Goal Setting 5 Identify List Of 'Controllable', Material Attributes & Process Parameters. CI Matrix & Fishbone Diagram Prioritize 2 Key 'Impactable & Critical' Material Attributes & Process Parameters. And Their Current 6 CI Matrix & SOP MEASURE Settings. 7 Select 2 Levels For Prioritized Material Attributes & Process Parameters. : (Low (-) & High(+) Brainstorming 8 Design Combinations Of 2^2 (2 Factors & Their 2 Levels ) & Execute The Experiment With 3 Replications. Experimental Design & Data collection Multivariate Modeling, Coefficient Of Determination- Data ANALYSE 9 Identify Statistically Significant Critical Material Attributes & Process Parameters And It’s Levels Analysis, P Value , Main Effect & Interaction Effect plot, cube plot Design Optimum Combination Of Statistically Significant Critical Material Attributes & Process 10 Process Optimisation-DOE: Prescriptive and Predictive analytics DESIGN Parameters And Their Levels. 11 Develop A Solution Implementation Plan To Deploy Solutions. Solution Impleemntation plan VALIDATE CONTROL DEVELOP A SOLUTION IMPLEMENTATION PLAN TO DEPLOY SOLUTIONS. We have proven the significance of the factors and their levels statistically through experiments. Now we need to validate it operationally i.e. we will start implementing settings on daily / shift wise basis on floor. The difference between factorial settings of experimental phase (1st) and factorial settings of post experiment implementation phase (2nd) is- The purpose of experimental phase is to find optimum levels of factors and the purpose of ‘post experiment solution implementation’ is to validate result that were seen in experiment, on short term basis. Had it not been validated in this phase, we will scrap the results of experiments and go for next experiment. Once experimental results are VALIDATED IN VALIDATE PHASE, we can go to next phase (3rd) of upgrading SOPs with “new normal” factorial settings to implement settings for ongoing / long term basis. 100 DEVELOP A SOLUTION IMPLEMENTATION PLAN TO DEPLOY SOLUTIONS. In order to better understand the effects of NEW SETTINGS and to learn how to make the implementation more effective, each solution is implemented for short term basis. An Implementation plan is developed for solutions and solutions are implemented on pilot basis i.e. for short period of time. 101 DEVELOP A SOLUTION IMPLEMENTATION PLAN TO DEPLOY SOLUTIONS. Optimised Levels based on STATISTICALLY Who should Status of Implementation Sr. PRESCIRPTION ANALYTICS IN Timeline of Responsible to Status SIGNIFICANT ensure Solution (Implemented / Not No PURSULLENCE'S PO-DOE Implementation Implement Validation CMAs / CPPs Implementation Initiated ) SYSTEM 1 Asst. Manager should mail Asst. status of implementation Supervisor manager every week to Manager, cc to Sr. manager 2 102 VALIDATE PHASE “ Validate Project Y Improvement / Optimisation.” 103 LSS Step QBD LEAN Six Sigma Yellow Belt - End to End Problem Solving Steps Tool Phases # 1 Identify The Current Business Issue (Qualitative problem) VOC Analysis-complaint / Feed Back Log 2 Collect & Study Historical Data (Quantitative problem ) Historical Data Collection DEFINE 3 Calculate Baseline / Pre- Solutioning Pharma Process Capability (Statistical Problem) Capability Index - Parts per Million ( PPM) & Sigma Score (Zlt) 4 Compare Baseline Capability With Industry Benchmark & Set Improvement Goal Improvement Goal Setting 5 Identify List Of 'Controllable', Material Attributes & Process Parameters. CI Matrix & Fishbone Diagram 6 Prioritize 2 Key 'Impactable & Critical' Material Attributes & Process Parameters. And Their Current Settings. CI Matrix & SOP MEASU RE 7 Select 2 Levels For Prioritized Material Attributes & Process Parameters. : (Low (-) & High(+) Brainstorming 8 Design Combinations Of 2^2 (2 Factors & Their 2 Levels ) & Execute The Experiment With 3 Replications. Experimental Design & Data collection ANALYS Multivariate Modeling, Coefficient Of Determination- Data 9 Identify Statistically Significant Process Parameter And It’s Levels E Analysis, P Value , Main Effect & Interaction Effect plot, cube plot DESIGN 10 Design Optimum Combination Of Statistically Significant Process Input Parameters And Their Levels. Process Optimisation-DOE: Prescriptive and Predictive analytics 11 Based on Optimum Values of CMAs/ CPPs, deploy scenario analysis. Scenario Analysis. VALIDA TE CONTR OL BASED ON OPTIMUM VALUES OF CMAS/ CPPS, DEPLOY SCENARIO ANALYSIS STEP 1: In Design phase we found optimum / better values for selected 2 Input factors that calculate optimum value for CQA. In Validate phase, we need to validate values of input factors that we found in design phase, are really optimum? Let’s validate it… Pursullence DOE Predictive Analytics tool will generate range of + /- 30 % deviation around optimised values for both input factors. Ex: Say, PSD and Compression Force are our CMAs / CPPs and CU-% RSD is our CQA. In Design phase, we found 10.5 µm is an optimum value for PSD and 24.5kN is an optimum value for Compression Force. In first step of Validate phase, analytics tool will automatically calculate 30% (+/-) range of values around 10.5 µm i.e. 10.5*(1-0.30) = 7.35 µm (min.) & 10.5*(1+0.30) = 13.65 µm (max.) for PSD. And likewise analytics tool will also calculate +/- 30% range for second factor of compression force: min. of 17.15kN to max. of 31.85kN. STEP 2: Add these MIN & MAX values for both factors in the table given. 105 BASED ON OPTIMUM VALUES OF CMAS/ CPPS, DEPLOY SCENARIO ANALYSIS As highlighted in black, Copy Min and max values for both Input factors from table of scenario analysis and type it in the Table of Probability Distribution (Min-Max). 106 LSS Step QBD LEAN Six Sigma Yellow Belt - End to End Problem Solving Steps Tool Phases # 1 Identify The Current Business Issue (Qualitative problem) VOC Analysis-complaint / Feed Back Log 2 Collect & Study Historical Data (Quantitative problem ) Historical Data Collection DEFINE 3 Calculate Baseline / Pre- Solutioning Pharma Process Capability (Statistical Problem) Capability Index - Parts per Million ( PPM) & Sigma Score (Zlt) 4 Compare Baseline Capability With Industry Benchmark & Set Improvement Goal Improvement Goal Setting 5 Identify List Of 'Controllable', Material Attributes & Process Parameters. CI Matrix & Fishbone Diagram 6 Prioritize 2 Key 'Impactable & Critical' Material Attributes & Process Parameters. And Their Current Settings. CI Matrix & SOP MEASU RE 7 Select 2 Levels For Prioritized Material Attributes & Process Parameters. : (Low (-) & High(+) Brainstorming 8 Design Combinations Of 2^2 (2 Factors & Their 2 Levels ) & Execute The Experiment With 3 Replications. Experimental Design & Data collection ANALYS Multivariate Modeling, Coefficient Of Determination- Data 9 Identify Statistically Significant Process Parameter And It’s Levels E Analysis, P Value , Main Effect & Interaction Effect plot, cube plot DESIGN 10 Design Optimum Combination Of Statistically Significant Process Input Parameters And Their Levels. Process Optimisation-DOE: Prescriptive and Predictive analytics 11 Based on Optimum Values of CMAs/ CPPs, deploy scenario analysis. Scenario Analysis. 12 Deploy Monte Carlo Simulations for different scenarios of CMAs / CPPs. Probability Distribution And Monte Carlo Simulations VALIDA TE CONTR OL DEPLOY MONTE CARLO SIMULATIONS FOR DIFFERENT SCENARIOS OF CMAs / CPPs. STEP 3: Once Min-Max values are fed to analytics tool, in the table, Analytics tool will execute Monte Carlo simulations (MCS). Let’s understand How? First sub-step to deploy MCS is to find out INPUT factors. In our case, that we have already found 2 Input factors. Second sub-step to deploy MCS is to choose data distribution type for Input factors. Here, we will use data distribution type of “Uniform Distribution”. What is Uniform Distribution? 108 DEPLOY MONTE CARLO SIMULATIONS FOR DIFFERENT SCENARIOS OF CMAs / CPPs. STEP 3: To execute MCS, analytics tool will randomly choose any value between the range of +/- 30% for both Input factors. Each of the values will have uniform / equal chance to be chosen. Ex: In above case PSD has a range of 7.35 to 13.65. Likewise for Compression Force: 17.5 to 31.85. To execute Monte Carlo simulations, based on the algorithm within analytics tool, tool will randomly choose values between 7.35 to 13.65 for PSD with uniform or equal chance for each value to be chosen. Analytics tool will automatically select values between +/-30% range for both factors and calculate CQA. 109 DEPLOY MONTE CARLO SIMULATIONS FOR DIFFERENT SCENARIOS OF CMAs / CPPs. AT QbD LEAN SIX SIGMA Yellow Belt Level, once you find optimum levels for both Input factors, analytics tool will give you +/-30% deviation values/range for both factors, just select it and add in the table. In the next module of MC_Simulations, just clear the values of CQA from 2nd column and click on the BLUE Balloon. Algorithm in analytics tool will automatically deploy Uniform distribution for both Input factors based on the regression equation, calculate CQA. Not only that analytics tool will automatically iterate this process for 1000 times and based on statistical modeling, will generate 1000 values for CQA. 110 DEPLOY MONTE CARLO SIMULATIONS FOR DIFFERENT SCENARIOS OF CMAs / CPPs. WHY?? The sole purpose of Validate / Verify phase is to validate improvement in CQA. To validate / verify improvement in CQA, we need sufficient data. However, In real world, it is not feasible to collect data in ample amount / larger scale because it consumes lots of time and cost. Here we would need to scientifically simulate the Input values and generate sufficient data points. Statistical models simulate values for Input factors for 1000 times based on recent / historical values and calculate CQA for 1000 times as if we are setting those Input factors in real world and calculate CQA for 1000 days. Simulations model that we are using here is Monte Carlo simulations. Following is a brief about Monte Carlo Simulations, 111 DEPLOY MONTE CARLO SIMULATIONS FOR DIFFERENT SCENARIOS OF CMAs / CPPs. When we want to predict the performance for future, we either do it thru guesswork also called Intuitions, or based on recent / historical data analysis. Guesswork is always “risky”. Hence companies prefer recent / historical data analysis for predicting the future. However, the limitation of predicting based on historical data is that we make inferences for FUTURE based on recently “happened” data. Hence there always lies an “uncertainty or risk” in inferring based on PAST PERFORMANCES. Although, data based predictions is highly accurate in comparison with that of intuitions / guess work. Now, question is How to increase accuracy of predictions based on recent / historical past performances. As discussed, here, we need to coin the concept of “uncertainty”, i.e. how much uncertainty lies in predicting based on past data and can that risk be absorbable? Here, [100 % - % uncertainty] = probability of assurance or guaranty, also called ‘Confidence’. And to enhance that assurance and reduce the uncertainty, we need to have similar data…loads of data. However we can’t have “that much” data that reduce uncertainty in predictions significantly. Or capturing sufficient data consumes loads of time and cost. To rescue us, here comes a great friend - Monte Carlo Simulations (MCS). 112 DEPLOY MONTE CARLO SIMULATIONS FOR DIFFERENT SCENARIOS OF CMAs / CPPs. MCS is kind of a “Time-Machine” that travels in future for 1000 RUNS and come back to give us 1000 values of Input factors and thus calculate CQA. Now, it will be highly accurate to validate based on these 1000 datasets than just recent 12 experimental runs. Thus, MCS help us filter out the impurity of uncertainty up to higher extent. However, MCS comes up with certain statistical conditions of knowing data distribution. Data distribution in simple terms is nothing but the behavior or spread of the data. Anticipation of data behavior is critical in predictive analytics. MCS simulate Input values to calculate CQA based on FULFILLING the condition of data distribution. Here MCS uses data distribution type of Uniform distribution and generate 1000 values for BOTH INPUT FACTORS. 113 DEPLOY MONTE CARLO SIMULATIONS FOR DIFFERENT SCENARIOS OF CMAs / CPPs. Besides that, system will also find “mean” value (avg. value) and “confidence intervals” of 95%, 90%, 85% and 80% around mean value. Let’s understand confidence interval first….While predicting ANY THING for future, we cannot predict exactly. In other words, we cannot be sure for exact value for any predictions. Predictions are always made in the “range”! Let’s take a day to day example. After any examination, if somebody ask you that how much will you score? You can be sure for 1st class, or distinction or say between 60%-70%, but can you predict exact 66.39%...No one can….!! Once we say between 60%-70%, may be next question will come to your way, how much sure /confident you are of the score between 60%- 70%..? Are you….80% sure, or 90% or 95% or 99% sure…Of course, no one can be 100% sure, as we are predicting the future… Right?? So if we will answer that I am 90% confident that I will score between 60%-70%, that ‘range ’ is called ‘‘Confidence Interval’’ and 90%surety is called “Confidence Level”… 114 DEPLOY MONTE CARLO SIMULATIONS FOR DIFFERENT SCENARIOS OF CMAs / CPPs. However, when you answer I will get the score in between 60%-70%. It is by ‘hunch’ or ‘intuition’. What if we could calculate ‘Confidence Interval’ (CI) of certain ‘Confidence Level’ more accurately…how can we calculate it? ….any guesses…Of course, our best friend statistics will come here to rescue us.. Statistics can calculate confidence interval / range with certain confidence level. Pursullence’s system has an algorithm at back end that statistically calculate CI with Confidence Level of 90% and 95%..! By thumb rule, higher the confidence level, higher will be the confidence interval and vice versa! Thus, Pursullence system can predict confidence interval of Avg. NPV value at certain confidence level. Now Let’s get back to next STEP In Validate Phase, 115 As highlighted in red, first of all, clear previous data from 2nd column till last value in cell no. 1002. There can be 1000 values from cell no. 3 to 1002. Select Cell No. 3 , drag it till cell no. 1002 and click “Delete” button. Then, click on Blue ballon. Algorythm in Pursullence analytics tool will simulate 1000 values in next couple of minutes. Analytics System will also show how many iterations are left at top of the table. 116 LSS Step QBD LEAN Six Sigma Yellow Belt - End to End Problem Solving Steps Tool Phases # 1 Identify The Current Business Issue (Qualitative problem) VOC Analysis-complaint / Feed Back Log 2 Collect & Study Historical Data (Quantitative problem ) Historical Data Collection DEFINE 3 Calculate Baseline / Pre- Solutioning Pharma Process Capability (Statistical Problem) Capability Index - Parts per Million ( PPM) & Sigma Score (Zlt) 4 Compare Baseline Capability With Industry Benchmark & Set Improvement Goal Improvement Goal Setting 5 Identify List Of 'Controllable', Material Attributes & Process Parameters. CI Matrix & Fishbone Diagram 6 Prioritize 2 Key 'Impactable & Critical' Material Attributes & Process Parameters. And Their Current Settings. CI Matrix & SOP MEASU RE 7 Select 2 Levels For Prioritized Material Attributes & Process Parameters. : (Low (-) & High(+) Brainstorming 8 Design Combinations Of 2^2 (2 Factors & Their 2 Levels ) & Execute The Experiment With 3 Replications. Experimental Design & Data collection ANALYS Multivariate Modeling, Coefficient Of Determination- Data 9 Identify Statistically Significant Process Parameter And It’s Levels E Analysis, P Value , Main Effect & Interaction Effect plot, cube plot DESIGN 10 Design Optimum Combination Of Statistically Significant Process Input Parameters And Their Levels. Process Optimisation-DOE: Prescriptive and Predictive analytics 11 Based on Optimum Values of CMAs/ CPPs, deploy scenario analysis. Scenario Analysis. 12 Deploy Monte Carlo Simulations for different scenarios of CMAs / CPPs. Probability Distribution And Monte Carlo Simulations VALIDA TE 13 Validate CQA Optimisation Based On: Study sensitivity of CQA for diff. Scenarios and study related Risk. Sensitivity Analysis,, Statistical Risk Assessment CONTR OL VALIDATE CQA OPTIMISATION : STUDY SENSITIVITY OF CQA FOR DIFF. SCENARIOS AND STUDY RELATED RISK STEP 4: Once 1000 iterations are executed, we have 1000 values for CQA. In next Module, analytics tool will do the risk analysis. SENSITIVITY ANALYSIS: All that experimenter need to check is the probability of the risk (%). It means, a.) if we set values for both Input factors at Optimum value that we find in Design phase or b.) if we set them at any value between 30% +/- range, What will be CQA value and out of 1000 scenarios generated by MCS, how many time, will CQA goes beyond the Target value. In other words, we check HOW SENSITIVE IS CQA / OUTPUT TO VALUES OF INPUT FACTORS. This is called Sensitivity Analysis. In current case, target for CU-%RSD is 2%. So how many times, out of 1000 iterations, % RSD goes beyond target of 2%, is Risk. 118 VALIDATE CQA OPTIMISATION : STUDY SENSITIVITY OF CQA FOR DIFF. SCENARIOS AND STUDY RELATED RISK This risk value supposed to be as low as possible, and not more than 40%. If so, analytics tool will show it in red. If it is in red, it indicates, that the optimum value that you found in Design phase might looked optimum for 1 scenario, however for 1000 different scenarios, those SO CALLED OPTIMUM VALUES of input factors are not optimum. So go back to design phase prescriptive / prediction table and prescribe better values for Input factors. Get +/- 30% deviation values, deploy MCS for 1000 iterations and recheck % Risk. And keep repeating these steps, until you find optimum values of Input factors that give lower % risk. Once you find Values for Input factors that shows lower % risk, it VALIDATE /VERIFY that these values are Statistically Optimum values for both of the Input factors, thanks to Scenario analysis, Data Distribution-Uniform Distribution, Monte Carlo simulations & Sensitivity analysis based Risk analysis. 119 Pursullence Analytics Tool will provide detailed statistical analysis in the table as highlighted here including Mean, Confidence interval (95%, 90%) for Mean / average and other statistical terms. Check probability of risk i.e. % Risk, if % Risk is low, analytics tool will highlight it in Green , if high, in Red. If % Risk is low, This validate that Input factors – CMAs / CPPs are at “Statistically optimum” values. Thus, change SOP ( std. operating procedural) values for Input factors to above validated values for future settings. You can also check 50 values for X bar and Range chart in same module that will be explained on next few slides. 120 LSS Step QBD LEAN Six Sigma Yellow Belt - End to End Problem Solving Steps Tool Phases # 1 Identify The Current Business Issue (Qualitative problem) VOC Analysis-complaint / Feed Back Log 2 Collect & Study Historical Data (Quantitative problem ) Historical Data Collection DEFINE 3 Calculate Baseline / Pre- Solutioning Pharma Process Capability (Statistical Problem) Capability Index - Parts per Million ( PPM) & Sigma Score (Zlt) 4 Compare Baseline Capability With Industry Benchmark & Set Improvement Goal Improvement Goal Setting 5 Identify List Of 'Controllable', Material Attributes & Process Parameters. CI Matrix & Fishbone Diagram 6 Prioritize 2 Key 'Impactable & Critical' Material Attributes & Process Parameters. And Their Current Settings. CI Matrix & SOP MEASU RE 7 Select 2 Levels For Prioritized Material Attributes & Process Parameters. : (Low (-) & High(+) Brainstorming 8 Design Combinations Of 2^2 (2 Factors & Their 2 Levels ) & Execute The Experiment With 3 Replications. Experimental Design & Data collection ANALYS Multivariate Modeling, Coefficient Of Determination- Data 9 Identify Statistically Significant Process Parameter And It’s Levels E Analysis, P Value , Main Effect & Interaction Effect plot, cube plot DESIGN 10 Design Optimum Combination Of Statistically Significant Process Input Parameters And Their Levels. Process Optimisation-DOE: Prescriptive and Predictive analytics 11 Based on Optimum Values of CMAs/ CPPs, deploy scenario analysis. Scenario Analysis. 12 Deploy Monte Carlo Simulations for different scenarios of CMAs / CPPs. Probability Distribution And Monte Carlo Simulations VALIDA TE 13 Validate CQA Optimisation Based On: Study sensitivity of CQA for diff. Scenarios and study related Risk. Sensitivity Analysis,, Statistical Risk Assessment 14 Validate CQA Optimisation Based On: Statistical Process Control SPC- X bar & R charts CONTR OL VALIDATE CQA OPTIMISATION : STATISTICAL PROCESS CONTROL STEP 5: After sensitivity analysis and risk assessment, Analytics tool will also automatically create statistical control charts- X bar chart (for Mean) & R chart (for standard deviation). All data points or most of the data points on this chart/graph should be within upper and lower control limits. This indicate that the CQA performance over the period of 1000 iterations is steady-stable and under control with out much chances of unpredictable variation. This is another statistical way to validate / verify the improvement of CQA. Let’s understand statistical control chart first and then X bar and R Chart, 122 VALIDATE CQA OPTIMISATION : STATISTICAL PROCESS CONTROL Statistical Process Control (SPC) Statistical Process Control (SPC) : SPC control variation in performance of Project Y (Output) by constantly monitoring & maintaining it under statistically defined ‘control limits’.  SPC monitor Project Y over the period of time.  SPC is an Early Warning System if the Project Y is out-of-control.  SPC detect “Unusualness Or Abnormality” in process behavior over the period of time. 123 Statistical Process Control USES Control Chart….  Control Charts were developed by Walter Shewhart in the 1920’s. Control Charts plot data of ‘Y’ over time (data points connected by lines), in order to detect trends and/or unusual events.  Control Chart is like Line Charts, but with the addition of ‘control limit’ lines (upper control limit (UCL) and lower control limit (LCL)), and an average (or ‘center’) line. 124 SPC Graphical Output This point is ‘unusual’ because it falls outside the control limits and thus need to be investigated- Why is it Out of control limit…? Is it by chance or by special cause…. 125 VALIDATE CQA OPTIMISATION : STATISTICAL PROCESS CONTROL The X bar-R chart, also known as the X bar and R chart, is a statistical control chart commonly used in statistical process control (SPC) to monitor and control the process mean and variability. It consists of two separate charts: the X bar chart and the R chart. X bar Chart: The X bar chart is used to monitor the average CQA over time. It displays the average values of samples taken from the 1000 iterations. Analytical tool will automatically create 50 samples of 20 CQA values in each sample = total of 1000 CQA values. Each sample will have average / mean value, so total of 50 samples and thus 50 average values. Analytical tool will create graph / control chart from 50 average values. The X-bar chart helps identify any shifts or trends in the mean /average values of CQAs that may indicate a deviation from the desired target value. 126 VALIDATE CQA OPTIMISATION : STATISTICAL PROCESS CONTROL To interpret the X-bar R chart: 1) The centerline represents the overall average of the process. 2) Control limits are typically set at three standard deviations above and below the average. 3) Data points falling within the control limits indicate the process is in statistical control. 4) Data points beyond the control limits suggest the process is out of control and may require investigation and corrective action. 127 VALIDATE CQA OPTIMISATION : STATISTICAL PROCESS CONTROL R Chart: The R chart is used to monitor the process variability or dispersion over time. It displays the range or difference between the highest and lowest values within each subgroup or sample of CQAs. So if there are 50 values in each sample, each sample will have Highest value and lowest value. Range = Highest value - lowest value. So each of 50 samples will also have range value, just like average or mean value. The R chart helps identify any changes in process variability. Higher the variability suggests that CQA values are not stable. Any data points beyond control limits suggest that CQA may not be under control. Remember, we have calculated CQA based on Monte Carlo simulations. Hence, if any data point goes beyond control limits, it indicates that Input values are not set appropriately…hence go back to Prescription-Predictability table and change values for Input factors In Design Phase. If CQA values for both X bar and R charts are within control limits, it validate/ verify that Input values are statistically optimum values. 128 As highlighted in red, you can see X- bar chart with 50 avg. values for 50 samples of 20 CQA values, total of 1000 iterations. Blue line is for avg. values of each of 50 samples & yellow & amber dotted lines are statistical control limits. For both X- bar & Range chart, most of the data points within control limit validate / verify that CQA is stable & under control as Input factors – CMAs/CPPs are at “Statistically optimum” values. Thus, change SOP (std. operating procedural) values for Input factors to above validated values for future settings. 129 To summarize, 1) Define phase gives us – from Qualitative to Quantitative to Statistical problem of CQA /Output. 2) Measure phase gives us – from multiple probable factors to 2 Critical Input factors (CMAs/ CPPs) that impact CQA/Output. 3) Analyze phase gives us- 2 Experimental level /values for both Input factors for experimentation. 4) Design Phase gives us- 1 would be Optimum value for both Input factors. 5) Validate Phase gives us- from 1 would be Optimum value to 1 Statistically Optimum value for both Input factors that improves performance of CQA/Output. That’s the journey Of QbD LEAN SIX SIGMA – DMADV !! Here we are done with Validate / Verify Phase. 130 CONTROL PHASE “ Sustain With the Process Improvement & Stable / Control Process Performance for ongoing basis…” 131 LSS Step QBD LEAN Six Sigma Yellow Belt - End to End Problem Solving Steps Tool Phases # 1 Identify The Current Business Issue (Qualitative problem) VOC Analysis-complaint / Feed Back Log 2 Collect & Study Historical Data (Quantitative problem ) Historical Data Collection DEFINE 3 Calculate Baseline / Pre- Solutioning Pharma Process Capability (Statistical Problem) Capability Index - Parts per Million ( PPM) & Sigma Score (Zlt) 4 Compare Baseline Capability With Industry Benchmark & Set Improvement Goal Improvement Goal Setting 5 Identify List Of 'Controllable', Material Attributes & Process Parameters. CI Matrix & Fishbone Diagram 6 Prioritize 2 Key 'Impactable & Critical' Material Attributes & Process Parameters. And Their Current Settings. CI Matrix & SOP MEASU RE 7 Select 2 Levels For Prioritized Material Attributes & Process Parameters. : (Low (-) & High(+) Brainstorming 8 Design Combinations Of 2^2 (2 Factors & Their 2 Levels ) & Execute The Experiment With 3 Replications. Experimental Design & Data collection ANALYS Multivariate Modeling, Coefficient Of Determination- Data 9 Identify Statistically Significant Process Parameter And

QbD_LSSYB_DOE_PHARMA DEVELOPMENT OPTIMISATION_MCS_STUDY MATERIAL_156933 PDF

Document Details

Tags

Related

Summary

Full Transcript

Upgrade to continue