The results of this survey suggest that EAR in randomized controlled trials in India is low, with an upper bound of 95% confidence interval of 32.1%. This is in contrast to the situation in most developed countries where EDC use is now more widespread. Emam et al.’s survey identified that as far back as 2007, nearly 41% of the trials were using an EDC (12). In 2010, a study of the European Clinical Research Infrastructure Network (2010) identified that nearly 90% of the centers use a clinical data management system (CDMS) - which enables the capture of trial data using electronic case report forms (18). However, they also identified a wide heterogeneity in the systems used across centers and adherence to standards.
Given the demonstrated utility of EDC in ensuring data quality(19) while improving efficiency (20) and reducing the costs of clinical trials (21), it is interesting to note that software costs were considered as a critical barrier towards implementation. This is not surprising considering the high initial capital investment (server and software purchase) required for setting up such systems. Only 51% of the randomized trials in this survey have access to an institutional CTU. However, access to a CTU was not associated with a higher EAR. This may indicate that CTUs are not actively investing in and advocating for EDCs.
Unless institutional funding and technical support are available, it may not be feasible for investigators to individually adopt and maintain EDC systems for academic clinical trials (Appendix II Table 11). Additionally, high bandwidth internet access or hospital local area networks are required for most systems. Even among the free-text responses, lack of funding and difficulties in hiring, training, and retaining personnel who can use and manage the EDC systems were prominent. Several investigators were also unaware that such systems existed. The role of an institutional CTU which has the capability to manage an EDC is, therefore, paramount to improve the EAR.
Data from our survey also suggests that EAR has not changed over the past decade as 95% confidence intervals of EAR tend to overlap for all years. This is not the case in the West, which has seen a steady increase in EAR over the past decade (21). Increasingly direct electronic health record (EHR) to eCRF data collection is being adopted in the West (22). The lack of change in EAR India is likely to be related to systematic barriers that have remained consistent over time.
Between 2013–2015, the Central Drugs Standard Control Organization (CDSCO) instituted several regulations like a requirement for audio-video consenting, mandatory registration of ethics committees as well as the specification of compensation norms (23). This may have resulted in fewer industry-sponsored trials being registered (24). Analysis of our data suggests, that the proportion of industry-sponsored trials fell from 37.7% of the total trials in 2012 to 11.9% of total trials in 2013 (Appendix II, Fig. 3). The decline in industry-sponsored trials may have affected the EAR across time.
The categorization of EDC sophistication levels used in the present study was derived from Emam et al. to maintain comparability across study results (12). These levels can be modeled using a Guttman scale, which is suitable for cumulative functionality levels (25). EDC systems which have implemented level 6 are likely to have implemented level 5 also. Noteworthy level 6 sophistication level corresponds to the ability to track medication inventory at the site, and level 5 is subject recruitment that can be tracked online at each location. Both of these requirements are important for multicentric industry-sponsored studies. Thus it is not surprising that industry-sponsored studies tended to use more sophisticated forms of EDC. However, only 15 of 32 (46.8%) industry-sponsored trials had used an EDC with a sophistication level of 6.
Setting up EDC systems to be available institute-wide requires substantial infrastructure, equipment, and human resources. While commercial EDC systems exist, they are expensive and may be out of reach for most academic investigators. Free and feature-rich, web-based EDC systems like REDCap are available, but increased awareness is necessary (26). Our institute has maintained a REDCap EDC, which has hosted over 300 projects over the past decade. The REDCap consortium website shows that 96 centers in India are using REDCap (27). Cloud-based EDCs may ease the adoption barrier in centers with limited IT support. However, the complex regulatory and patient privacy requirements would mandate that a system is available at a national level with stringent protections for patient privacy and intellectual property.
Meanwhile, developments in the software have resulted in improved functionality achieved by EDC. Some examples are the ability to abstract data from electronic health records through the integration of natural language processing and artificial intelligence, including decision support systems, integration with electronic patient-reported outcomes, and wearables (21). In resource-constrained settings like India, it may make more sense to develop and maintain such advanced EDC systems as a web-based application available to all researchers across the country.
Limitations
The study included only registered randomized controlled trials where contact details were available. However, since RCTs provide the highest level of evidence regarding efficacy (and therefore need high-quality data to be interpretable and usable), the use of EDC will be most important in these studies. We also contacted the investigators instead of the study coordinators as the contact details of the study investigators were only available. Contact details of contract research organizations (CROs) were also inconsistently available on CTRI. Contacting the CROs may have improved the EAR as industry-sponsored studies are more likely to have funds to use an EDC. As shown by the results of this survey, only half of the studies had a clinical trial unit at the center. However, the definition of a clinical trial was not specified in the survey instrument. Finally, random sampling was not used in the study. However, key trial characteristics viz. sponsorship type, number of sites, countries of recruitment, and sample size seem to have a similar distribution in the trials for which we received a response versus those we did not (Appendix II: Table 3).