trial analytics ± a tool for clinical trial management

Acta Poloniae Pharmaceutica ñ Drug Research, Vol. 69 No. 3 pp. 523ñ533, 2012 ISSN 0001-6837Polish Pharmaceutical Society

The accent of clinical trials managementprocess is highly critical for any organization per-forming clinical trials. In the present scenario,organizations spend hundreds of millions of dollarson budgets, time frames that range to decades, andthe complexity of these trials has been ever-increas-ing. There has been a lot of effort duplication andwastage on data management, programming and re-programming for these trials. Pharmaceutical firmsnow know the criticality of managing plan accuracy,timeliness, and making better resource allocationdecisions. Adding a single day to the planned dura-tion of a clinical trial may cost around $ 1 million.Similarly, losing a day from business market exclu-sivity due to a delay in submission could cost around$ 10 millions.

Prolonged timelines and large expenses associ-ated with bringing new drugs to the market haveprompted a new focus on improving the operationalefficiency of clinical trials. As a result, industrysponsors of clinical trials have been increasinglyreliant upon a variety of IT-enabled Clinical TrialManagement Systems (CTMS) in order to improvemanagerial control in trial conduct.

A CTMS is a customizable software systemused by biotechnology and pharmaceutical indus-tries to manage large amounts of data involved withthe operation of a clinical trial. It is widely used tomaintain and manage the planning, preparation, per-

formance and reporting of clinical trials, whileemphasizing on keeping up-to-date contact informa-tion for clinical trial participants and tracking mile-stones and deadlines e.g., those used for regulatoryapprovals or issuing progress reports. Currently, allthe CTMS do provide clinical trial sponsors withimproved oversight of a trial conduct. But, these tri-als continue to be routinely over-budgeting, sufferfrom poor communication, lack team awareness,and are found to be unable to forecast and resolveissues quickly.

The problem with the current CTMS is thewidely varying requirements of clinical trials man-agers. Some standard requirements include: budg-eting, patient management, compliance to govern-ment regulations and compatibility with other datamanagement systems. The complete set of require-ment of one manager/sponsor might be completelydifferent from that of another. These shortcomingsresult in delay and, sometimes, failure of clinicaltrials. A recent study suggests that improvementsin a few key areas can result in a dramatic 60ñ90%reduction in total cost of a clinical trial (1). Yet,beyond electronic data capture (EDC), no class oftools was proposed as enablers of the same (1).This paper proposes the usage of an innovative andcollaborative visualization tool ìas a CTMS add-onî to help overwhelm these deficiencies of tradi-tional CTMS.

GENERAL

TRIAL ANALYTICS ñ A TOOL FOR CLINICAL TRIAL MANAGEMENT

ANINDYA BOSE* and SUMAN DAS

Healthcare Practice-Lab Automation Division, HCL Technologies Ltd., Kolkata-700156, India

Abstract: Prolonged timelines and large expenses associated with clinical trials have prompted a new focus onimproving the operational efficiency of clinical trials by use of Clinical Trial Management Systems (CTMS) inorder to improve managerial control in trial conduct. However, current CTMS systems are not able to meet theexpectations due to various shortcomings like inability of timely reporting and trend visualizationwithin/beyond an organization. To overcome these shortcomings of CTMS, clinical researchers can apply abusiness intelligence (BI) framework to create Clinical Research Intelligence (CLRI) for optimization of datacollection and analytics. This paper proposes the usage of an innovative and collaborative visualization tool(CTA) as CTMS ìadd-onî to help overwhelm these deficiencies of traditional CTMS, with suitable examples.

Keywords: CTMS, BI, CTA, CTI, Decision Support System, OLAP, KPI

523

* Corresponding author: e-mail: [email protected]

524 ANINDYA BOSE and SUMAN DAS

OBSTACLES IN SUCCESS OF CLINICAL

TRIALS

Apart from the above limitations of traditionalCTMS, the success of clinical trials is also hinderedby the following obstacles:● enrollment challenges;● inconsistent and uninformed planning; ● poor visibility of enrollment trends; ● inability to evaluate effective adjustments to res-

cue a problem trial;● poor communications/collaboration across global

study teams. These challenges are explained in detail,

below:

Enrollment challenges

Ninety percent of clinical trials donít finish ontime because of increased regulatory requirements,complex protocols, and a global patient populationthat is quite difficult to track. These reasons makethe trialsí recruitment increasingly difficult to man-age, and result in enrollment failures and thus on-time completion. The impact of this is quite signifi-cant. Delays in enrollment completion affect timelysubmissions for regulatory approvals, and subse-quently, the product launch. Budget and cost over-runs combined with launching delays can cost acompany a good amount of fortune.

Inconsistent and uninformed planning

A reliable and feasible enrollment plan is crit-ical to finishing a trial on time. As tools likespreadsheets, business intelligence software, andtransactional CTMS require extensive customiza-tions, so, study teams often rely on intuition, goodfortune, and choked budgets to meet subject andtimeline targets. Also, requirements for enrollmentmanagement vary from one manager to another,making it difficult to share best practices or enforceconsistent business rules across and among organi-zations.

Poor visibility of enrollment trends

Study managers often miss valuable insightsand fail to spot potential problems because theylack the tools to quickly identify clinical studyenrollment trends. Although, the spreadsheet mod-els can provide some assistance, but have limitedability to aggregate data and predict potential prob-lems in an automated fashion. Instead of foreseeingproblems, most study managers rely on contingencyplanning to address enrollment issues after theytake place.

Inability to evaluate effective adjustments to res-

cue a problem trial

Slow site initiation, high screen failures, sea-sonal variability, and unproductive sites can alldelay recruitment. Even good plans need to beadjusted from time to time. Without the ability topredict, simulate, and model different scenarios,study managers canít effectively adjust the plan tokeep enrollment on track. In addition, significantamount of a study managerís time is spent in aggre-gating data from multiple sources, which can be bet-ter spent analyzing the data to make the best deci-sions. Without simulation and modeling tools, studymanagers canít view and drill down into the data todiagnose problems or to track progress. Moreover,they may over-invest in activities that contribute toover-enrollment, such as advertising campaigns,new centers, and other costly initiatives.

Poor communications across global study teams

Many global study teams have difficulty in col-laborating because they lack timely data communica-tion across numerous countries and centers. This lackof transparency impedes communications across mul-tiple time zones and delays decisions. Not only doesthe absence of a transparent system reduce staff effi-ciency and increase enrollment and data clean-upproblems, but it also allows companies to repeat cost-ly mistakes. When companies canít track each siteíshistorical performance, they canít identify top per-formers or weed out underperformers for future trials.

LIMITATIONS OF TRADITIONAL CTMS

IN CLINICAL TRIAL MANAGEMENT

The clinical trial stage of drug discovery, gen-erally, has a timeline range of 6ñ8 years, andinvolves budgets of hundreds of millions of dollars.The very high rate of failure and delays in comple-tion makes this stage a highly risky affair. Followingare the various limitations of traditional CTMS inclinical trial management:

In answering customer business questions

Although the pharmaceutical industry usesexpensive CTMS, many business questions essentialfor the success of clinical trials remains unanswered,e.g.:● What are the demographics and medical condi-

tions of patients and how are they affecting theresults of a clinical trial?

● How does patient compliance affect outcome? ● How to quickly identify candidates for a clinical

trial?

Trial analytics - a tool for clinical trial management 525

● What is the lifetime value of screening and inter-ventions?

● Why are the reasons for a patientís death? ● How quickly can the unknown trends/patterns be

identified and drilled down for details?

In patient enrollment

Management of patient enrollment has alwaysbeen a big limitation for CTMS. An optimalapproach for managing patient enrollment mustinclude the following:● ability to define feasible enrollment plans;● track progress of enrollment;● make informed corrections to the plan as and

when needed.This information visualization is inadequate in

traditional CTMS without the help of an analyticstool. Moreover there is a priceless need of site selec-tion through intelligent identification of diseasezone/population through historical records.

In clinical trial follow-up

After every clinical trial, there exists a veryimportant follow-up stage. Following are the vari-ous limitations of traditional CTMS at various sub-stages of clinical trial follow-up:

Planning stageEvery clinical trial begins with a plan. Life sci-

ences companies require use of predictive analyticsto forecast and visualize different scenarios so theydetermine the best course before taking action.Traditional CTMS lacks this feature to provide thepredictive analytics to forecast and visualize differ-ent scenarios so that the best course of action whileplanning a clinical trial.

TrackingA study manager tracks the recruitment

progress by gathering data from various source sys-tems (e.g., CTMS) to ensure that the enrollmentexercise in on-track. Most of the times, they enterthese information manually from the source systemsinto the spreadsheets to see how the actual enroll-ment process is progressing against the plan. Thedisadvantage of this manual system is that the factsand figures are updating every moment, so it is verydifficult to get the current picture at any point intime.

Problem diagnosisEven the best-laid plans can run off-course. If

this happens, then study managers must quicklyidentify the underperforming areas (e.g., countries

and site investigators), to correct the plan. This func-tion is also absent in traditional CTMS.

OptimizingOptimizing the plan is easy only when the system

is capable of predicting where recruitment is headed inorder to model and implement changes to bring theproject back on track. This final step in the clinical trialfollow-up process is critical and requires businessintelligence (BI) for the success of clinical trial.

Spreadsheet dependency

Custom-designed or commercial CTMS oftenlack the flexibility and forecasting capability tomodel the entire enrollment picture across a world-wide development portfolio. Consequently, spread-sheets are used for modeling and forecasting, thusgiving rise to the following issues:● Manual processes introduce substantial control

risks.● The combination of CTMS for collecting actual

enrollment and spreadsheets for forecasting orscenario modeling leads to problems of dataintegrity and information silos.

● Spreadsheets lack enterprise-class functionalitysuch as managed workflow and the ability to pushtemplate changes out to hundreds of users simul-taneously.

● Keeping multiple systems up to date and syn-chronized is frustrating, time-consuming anderror-prone.

● Spreadsheets create islands of data, making enter-prise-wide reporting a challenge.

BUSINESS INTELLIGENCE (BI) IN

CLINICAL TRIALS: PRESENT SCENARIO

The BI software market has grown at over 10%per year in recent years (2). The most commondomains are sales tracking, marketing, and financialanalysis (3). Penetration into the clinical trials spacehas been limited. Recent industry consolidation hasleft a few large vendors with entrenched platformsalong with many small emerging vendors with nicheapplications. According to Stephen Few and others,are useless charts and distracting visual features thattend to confuse users (4). Further, chart creation isstill only the realm of a privileged few; users stillhave limited facility to compose visualizations,especially with loosely related data. Hence there isan urgent need of a system that empowers end-usersto compose visualizations themselves would breakdown information silos and allow much faster andmore effective analysis and decision making.


CLINICAL TRIAL ANALYTICS (CTA)

Why CTA?

Improving the study execution process is a dif-ficult challenge, as the number of people, processes,and inherent variability involved is quite large.Clinical trial has become a complex process frominventory tracking to enrollment profiling, adverseevent analysis, financial tracking, and personnelmanagement; similar in size to a large systems engi-neering effort. Keys to success of clinical trialsinclude transparency at all levels, smooth coordina-tion of related efforts, and regular early detection ofproblems. The most disruptive situations occurwhen activity is diverted from the plan. For instance,enrollment at a few sites might be lagging, and it isunlikely that they will be able to catch up. Anothervariance from plan would be an unexpectedly highnumber of adverse events in a specific population. Ineach case, the range of alternative solutions is wideand requires inputs from across functional bound-aries, often a difficult proposition. If data, geo-graphic or procedural barriers prevent fluent com-munication and information flow among the broad-er trial team, issues seem to linger for much longerthan desired. Likewise, a different set of experts arecharged with data monitoring and analysis. Whiletrials are required to adhere to regulations and studyprotocols, any variances have both cost and dataquality implications. Late or missed visits donít justmean schedule delays; the resulting data may betainted and have regulatory impacts. So, carefullymonitoring the incoming data for impending prob-lems acts as will be a key capability. Analytics canhelp here, but just detecting the presence of anom-alies is inadequate. Determining the reason foranomalies requires human brilliance at assemblinginformation fragments and clues across multiple dis-ciplines. Tools like CTA allow broad inspection andintuitive exploration and thus empower analystsbeyond statistical and/or reporting packages.

Scope

The CTA solution is designed for clinicalresearch organizations for optimizing the results ofclinical trials by reducing cost and duration as wellas increasing success rate and safety. Its capabilityof data integration makes it the right choice foranalyses of patientsí complex data, response rates,laboratory tests, adverse events, and physical ordemographic data. Analysts are enabled to quicklyidentify unknown trends or patterns in data, drilldown for details, find incomplete records, evaluatethe quality or significance of information and per-

form more complex data queries quickly and intu-itively.

The unique ability to explore a number of mul-tivariate responses, expressed and stored in all kindsof data, help drug makers evaluate the drugsí safetyand efficacy in early phases of development.Clinical trials groups benefit on having the ability toexplore complex information in a collaborative andflexible environment.

It is to be noted that CTA is not a replacementof CTMS; rather it is an add-on to CTMS. CTMS isa clinical trial management and documentation tool.On the contrary, CTA will be used as reporting andbusiness intelligence/analytics tool. The various fea-tures of the proposed integrated system as opposedto the standard CTMS are highlighted in Table 1.

Intended business users

A CTA system can benefit a wide range ofbusiness users. Managers can track milestoneprogress, sponsors can monitor portfolio perform-ance, site investigators can follow patient timelinesand protocols, and data monitors can review datacollected to date.

Visualizations created for outlier detection orforecasting will always use current data. Tools anddata sets not being correlated, can be viewed alongshared dimensions; for example, financial data canbe overlaid with enrollment events and visits, tovisualize pay-for-performance over a given timeinterval. Likewise, patient adverse events, concomi-tant medications, and visits could be overlaidagainst lab results. Incorporating trial planning andbudgeting can help identify which sites are consum-ing the most resources while producing the leastresults. Protocol compliance charts identifyingwhich study events are overdue and at risk of beingnoncompliant, can call managers to action. Manyproblems faced by clinical trial teams are caused bylack of understandable information. When this bar-rier is removed, performance should improve.

Technical description

CTA uses tools and techniques for decisionsupport, data extraction, querying, and data miningto gather and analyze data for clinical research fordifferent categories of business analytics (BA) like:Information and knowledge discovery; decisionsupport and intelligent systems and visualization (5,6). Information and knowledge discovery usesOLAP (on-line analytical processing) ad-hocqueries, data mining, text mining, Web mining andsearch engines. Decision support and intelligent sys-tems include tools and techniques such as decision


support systems, statistical analysis, data mining andpredictive analysis. These tools and techniques canbe used for generating a hypothesis as well as datacollection and analysis. Further, decision supportsystems, specifically clinical research decision sup-port systems (CRDSS), can be created for multiplepurposes supporting the research goals and initia-tives.

Functional description

The stepwise CTA model is shown in Figure 1.It starts with defining business questions/objectivesthrough requirements eliciting (inputs). This step isfollowed with the creation of clinical trial business

intelligence models using dimensions and measures.The final outputs are analytic reports and key per-formance indicator (KPI) scorecards. These tools pro-vide end-users with predefined and ad-hoc reports,help them to measure and monitor progress towardorganizational goals, and discover meaningful infor-mation about the data. Also, patterns and trends canbe identified. When being web-enabled they can beconsumed by any client application and platform.

CTA can provide solutions for various prob-lems encountered in traditional CTMS. Some of itsfunctions include:● Analysis of patientsí complex data, response

rates, laboratory tests, adverse events, and demo-

Table 1. Comparison of standard CTMS and CTA-CTMS integrated system.

Feature in standard Feature inMain Feature Sub-Feature CTMS CTA + CTMS

(Yes/No/Partial) (Yes/No/Partial)

Ability to define feasibleenrolment plans No Yes

Tracking enrolment progress Partial Yes

Make informed corrections to theplan as and when needed No Yes

Clinical Trial Modifying enrolment dataMonitoring dynamically No Yes

Identification of under-performing areas No Yes

Optimization of the clinical trial plan No Yes

Modifying enrolment data dynamically No Yes

Ease of reporting Partial Yes

Flexible reporting No Yes

Spreadsheet dependency Yes No

Reporting Real-time reporting No Yes

Drilled-down reporting No Yes

Trend analysis No Yes

Data monitoring Partial Yes

Data Data integration from Management various systems No Yes

Database ETL (extract, transform and load) feature No Yes

Identification of KPIs No Yes

Monitoring of trial progress Partial Yes

Business Business forecasting No YesIntelligence Business analytics No Yes

Answering capability for customer business questions No Yes


graphic or physical data to quickly identifyunknown trends or patterns in data, drill down fordetails, find incomplete records, evaluate infor-mationís quality or significance and generallyperform even complex data queries quickly andintuitively.

● Correlating clinical findings such as lab valueswith conditions to discover new approach to diag-nosis and treatment.

● Quickly identifying candidates for clinical trialsby study of patientsí clinical data population todetermine study locations.

● Targeting non-compliant population by compar-ing geo-location-wise distributed study sites todetermine the root cause of study failure/delay.

● Determining the frequency of screening proce-dures to enhance diagnostic values and preventdeaths by finding the risk factors so that high riskpatients can be targeted for diagnosis and treat-ment.

● Interacting with database to find patterns, rela-tionships, and correlations.

APPLICATIONS OF CTA

Study compliance monitoring, performance

measurement, quick identification and issue reso-

lution

CTA tools used for remote site monitoring caninvolve the sites in a collaborative manner andenable them to derive benefits from the data theyenter. Empowering sites to actively track their ownperformance against key metrics allows them tocompete against other sites; strong performance canbecome a strong track record, which can lead toactive participation in future trials. Moreover, thesetools will encourage site investigator involvementand data perusal leading identification and resolu-tion of issues much early than the team meetings orscheduled monitor site visits.

Figure 1. Clinical trial analytics overview

Figure 2. Schematic representation for the clinical trial data model example


Table 2. KPI systems of clinical trial analytics.

KPI name Measurement components

Site location potential index Number of diseased patients, their economic condition, age group, sex etc.

Trial compliance index Number of late visits, recrutment target failure, dosing errors, regulatorycomplications, delay in completion etc.

Drug adversity measurement Degree of adversity, number of deaths, time of recovery, number of countermedicines etc.

Drug potential index Non response-response ratio, degree of recovery, time to recovery etc.

Figure 3. Classification hierarchies of clinical trial data dimensions

Figure 4. Example of country-wise distribution of patient population for a particular disease


Figure 5. Example of country- and city-wise distribution of patient population for a particular disease

Figure 6. Example of site, economic condition, age group and severity wise distribution of patient population for a particular disease in aspecific city

Remote site monitoring

According to Eisenstein, the largest area forimprovement of clinical trial management is in thearea of remote site monitoring, estimated at 21% oftrial costs (3). Monitors have the difficult job ofensuring protocol compliance and data quality, soregular review of trial data is critical. Substantiallylarge costs are incurred with prolonged trial moni-toring travels and resolution of data inadequacies.With the CTA tools, this desired information can be

captured in periodic standardized reports or dataviews resulting faster issue resolution.

Freedom from expensive trial amendments

It is a very regretful fact that many originalsites close down during the course of a trial.Analysis of investigator performance across trialsand assessment of their propensity to drop out oftrials, or not able to recruit patient types, can help inensuring that organizations do not have expensive


Figure 9. Example of adverse event report (sex-wise

Figure 7. Example of country- and age-group-wise patient severity to a particular disease

Figure 8. Example of age-group-wise patient response to a particular disease


trial amendments. CTA is capable of resolvingthese issues by identifying the right patients; maptheir concentration across geographies and siteswhere they would be best recruited and even fore-cast the expected volume of patients and its impacton the trial cycle time, using historical data onpatient epidemiology and knowledge of medicalclaims.

Applying collaborative visualization to clinical

trials

Using CTA, it is possible to integrate all rele-vant data into a single environment where no queriesare needed to select and view data. Moreover, thedata is always up-to-date, and charts continuouslychange based on new or altered data. Users canbrowse at different levels of granularity and directlyselect and move data among charts, tables, andreports to track important metrics and informationconcepts. Users can share live visualizations withspecific team members or to everyone involved inthe study. Yet, data security filters prevent unautho-rized access to either data or visualizations.

Clinical trial enrollment forecasting

Patient enrollment is an ongoing challenge forlife sciences organizations. Failure to enroll ade-quate numbers of patients is a primary reason whysome clinical trials fail. The ability to collect andmodel enrollment trends and make decisions to shutdown or set up additional sites is key to managing aglobal development portfolio as study designsbecome more complex and resources become morescarce. CTA empowers sponsors and contactresearch organizations (CRO) with robust forecast-ing and enrollment analytics that involve collabora-tion among managers, analysts and the individualinvestigator sites.

EXAMPLE OF BI MODELING USING

CLINICAL TRIAL ANALYTICS

Let us view an example of schematically creat-ed clinical trial model (Figure 2), which enables cre-ation of clinical trial analytics for further reporting.This model contains clinical trial fact table, dimen-sional elements (such as patient, location, adverse

Figure 10. Example of adverse event report (age-group-wise)

Figure 11. Clinical trial non-compliance analysis with drilledpatient causes

Figure 12. Clinical trial non-compliance analysis with drilled sitecauses


event, trial non compliance and time), and hierar-chies (explained separately in Figure 3). It alsoincorporates business logic through relationships forcarrying various measurements such as site selec-tion, trial performance, adverse event analysis andtrial non compliance analysis. The final outputs are:creation of the front-end analysis applications suchas KPI (key performance indicator) systems, bal-ance scorecards, reporting systems, and data miningsolutions. Examples of some of the visual analyticreports are given in the following figures (Figures4ñ12). We have also indicated some relevant KPIísand their description in Table 2.

CONCLUSION

Gains from popular clinical trial software toolsare achieving diminishing returns, and industry con-solidation is shrinking the toolmakersí innovationpipelines. Connecting all clinical trials stakeholdersthrough shared understanding of integrated data willhelp align their incentives and encourage collabora-tion and prompt issue resolution. Both cost avoid-ance and value creation can be derived from such acollaborative visualization system. Decreased travelcosts, shorter reporting time, and lower query andissue resolution time are common outcomes. Usersare empowered to create understandable, usable, andappealing visualizations to capture and share theirideas. Employing a collaborative visualization sys-tem leverages existing IT projects and assets to cre-ate an integrated data environment capable of grow-ing with your portfolio, while increasing your effec-tiveness and efficiency at managing clinical trials.Business intelligence has become essential in mostclinical research based organizations. BI is not con-strained to individual departments or organizations;rather it is essential at the clinical trial level. Thispaper analyzed the need for clinical trial BI and pre-sented the development methodology that incorpo-rate iterative and cross-functional CTA approach.

The introduced CTA model promotes an approachfor clinical trial modeling and BI design accordingto real business requirements. It reduces the clinicaltrial life cycle and increases its efficiency by bettermonitoring, communication and real time measures.Clinical trial BI reveals opportunities to reduce costsand time, stimulate revenue growth, and enablescompanies to understand the entire clinical trialprocess from the global organizational perspective.

Acknowledgment

The authors are grateful to the LIMS COE,HCL Technologies Ltd., India for providingrequired facilities for this work.

REFERENCES

1. Eisenstein E.L., Collins R., Cracknell B.S.,Podesta O., Reid E.D., Sandercock P., ShakhovY. et al.: Clinical Trials 5, 75 (2008).

2. IDC. Worldwide Business Intelligence Tools2005 Vendor Shares. http://download.micro-soft.com/download/0/7/6/0760A23B-FBAD-4446-AE12-2D03352BD3B8/Worldwide_busi-ness_Intelligence_Tools2005VendorShares.pdf

3. Neiss J.: Next generation pharmaceutical(NGP). 15 (2009). (http://www.ngpharma.com/article/Reducing-the-Cost-of-Clinical-Trials-with-Integrated-Visualization).

4. Few S.: Visual Business Intelligence. (http://www.perceptualedge.com/blog)

5. Turban E., Sharda R., Aronson J., King D.:Business Intelligence: A Managerial Approach,1st edn., Pearson Prentice Hall, Upper SaddleRiver NJ 2008.

6. Keeling T.L.: Issues in Information Systems 11,372 (2010).

Received: 3. 04. 2011

trial analytics ± a tool for clinical trial management

Documents