[Tom Beauchamp and Yashar Saghai. "The Historical Foundations of the Research-Practice Distinction in Bioethics," Theoretical Medicine and Bioethics 33, no. 1 (2012): 45–56. DOI 10.1007/s11017-011-9207-8. h/t Nathan Emmerich.]
Relying largely on the transcripts and draft materials of the National Commission for the Protection of Human Subjects of Biomedical and Behavioral Research (which employed Beauchamp on its staff), Beauchamp and Saghai argue that the commission excluded medical practice from IRB not because its members thought medicine didn't need oversight, but because it was obeying instructions from Congress "in a context that had political origins and political limits."
The article emphasizes the provision in the National Research Act to consider "The boundaries between biomedical or behavioral research involving human subjects and the accepted and routine practice of medicine." It attributes this provision to the Senate's interest in testimony by Jay Katz, the only Senate witness to have used the language of boundaries in his 1973 testimony.
This may be true on the Senate side, but the article doesn't explore the House hearings, in which the president of the Association of American Medical Colleges complained that "the proposed legislation does not delimit the activities of the commission in a manner whereby it would be clear that the ordinary practice of medicine is not subject to the jurisdiction of this commission." He offered the following:
PROPOSED NEW LANGUAGE FOR SECTION 1202(b) (5)
We would suggest that section 1202(b) (5), which concerns the duties of the Commission, be amended to read as follows:
... defining more precisely the boundary between biomedical and behavioral research involving human subjects and the accepted practice of medicine, provided that, in defining these boundaries, the Commission shall concern itself with organized research support by Federal funds, and shall not have jurisdiction over the ordinary practice of medicine;
[U.S. House of Representatives, Biomedical Research Ethics and the Protection of Human Research Subjects: Hearings before the Subcommittee on Public Health and Environment of the House Committee on Interstate and Foreign Commerce (93d Cong., 1st sess., 1973), p. 204]
Since this language ("accepted practice of medicine") is somewhat closer to the language in the final statute, I suspect the AAMC was more influential than Katz.
Understanding the AAMC's role actually offers greater support for the article's claim that the commissioners crafted the Belmont Report partly in search of "political acceptability." The authors note Commissioner Robert Cooke's concern that the commission not offend "organized medicine." In other words, IRBs don't govern routine medicine because in 1973 doctors were invited to testify before Congress. Social scientists were not, and we all know what happened to them.
The article finds that the concept of generalizability is central to the IRB world even though no one knows what it means:
The criterion of generalizable knowledge, once articulated, did not generate any substantial debate among commissioners as to its meaning or its ultimate adequacy with regard to various types of learning activities. The notion was not then, and has never since been, carefully analyzed in Commission deliberations, in federal regulations, or in the bioethics literature. This fact is striking inasmuch as generalizable knowledge is the cornerstone of the work of the National Commission and in federal regulations for the purpose of distinguishing research from practice. Nothing is more basic.