The purpose of the structural equation model is to assess the latent constructs from the factor analysis models for causal connections and determine if the factors interconnect though a larger framework. Give the results of the factor analysis models, the measures of organizational structure from the financial variables are more likely covariates to drug development strategies than casual indicators. The failure of the model testing the financial variables to converge supports this interpretation; therefore, these measures were excluded from further model building.
I began the structural equation modeling process for the drug variables by running individual models for each of the 11 factors in Table 5.6. To ensure that there were enough indicators for each construct to reach a solution, all of the variables in a factor with a loading of 0.30 or greater were included in the models. Models that converged were retained and all retained models were added in a stepwise process starting from the innovation research factor, the factor that explained the highest proportion of variance. Figure 5.2 shows the final best fit model.
The goodness of fit statistics for the model in Figure 5.4 are satisfactory: χ2(40) = 44.84, p=0.276, RMSEA = 0.045, CFI = 0.95, TLF = 0.93, and SRMR = 0.10. Analysis of the modification indices indicated that the correlation between orphan drug status and
antineoplastic ratio would improve model fit. This relationship makes conceptual sense because a potential strategy for altering a drug’s market segment is through expanding the patient base by utilizing orphan drug requests.
Figure 5.2 Best Fit Structural Equation Model
The model indicates a path dependency in the drug development strategy of pharmaceutical companies. The inverse relationship between the items loading on the factors of innovative research and alternating current treatments shows approvals for alterations to cancer drugs reduces the number of new molecular entity approvals a company receives. This finding is interesting in comparison to previous research on path dependencies in the pharmaceutical industry.
treatment B that was an alteration of treatment A. While my finding does not contradict this argument, it does question the spillover effect Cook et al. (2011) assumed would occur internally from the development of treatment B. The model in Figure 5.2 instead shows that developing alternations to existing drugs, either through new dosage or new formulation approvals, cost a company through a reduction of innovative approvals. This finding suggest then that a competitor firm rather than the originator firm as proposed by Cook et al. (2011) may realize the spillover effect from an intermediate treatment.
Further analysis of the data supports the presence of a bifurcated research path dependency within the field. Creating variables for innovative research and altering current treatments revealed that corporations with higher rates of innovate research had lower rates of altering current treatment approvals. The mean rate of alteration approvals by corporations above the median for innovative approvals was 0.40, and the mean rate of alteration approvals by corporations below the median for innovative approvals was 0.62. The same trend occurs comparing the rate for innovative approvals of corporations above and below the median for alteration approvals, 0.36 to 0.90.
Another interesting finding from the model is the significant correlation between antineoplastic and orphan drug approvals. While the ratio of orphan drug approvals loads on the innovative research factor along with the ratios for new molecular entities and priority review approvals, as covered in Chapter 4, orphan drug approvals can be granted for existing treatments. Considering that cancer drugs represent the largest therapeutic category of drugs in the development pipeline and that new treatments are among the most expensive drugs on the market, this correlation could indicate a distinct strategy of expanding market coverage for the most profitable drugs in a company’s portfolio. If the
treatment options are more effective for the target population than existing options, then this would still represent innovative research, but it also supports criticism of the Orphan Drug Act as being utilized by companies primarily for financial gain rather than
stimulating novel research for underserved patient groups.
In summary, the model in Figure 5.2 shows that the connections between
organizational research practices are not clearly explainable by the technical or scientific aspects of pharmacology. The trade-off that appears between developing innovative products and alterations to existing products is likely the result of organizational culture emphasizing one research path over the other. A more detailed explanation on this causal relationship requires additional organizational level data on the internal research practices at these companies and would be a good direction for future research.
In relation to the process of field structuration, the structural equation model significantly reduces the number of indicators contributing to measurable latent constructs in the data. This reduction is beneficial because it allows for a more
parsimonious latent class analysis, as opposed to one including all of the variables, which increases the likelihood of finding meaningful class divisions (Collins and Lanza 2010). Additionally, the structural equation model indicates that the financial variables are more appropriately treated as covariates to the research strategies rather than as measures of latent constructs.