Opinion | Americans Need Generic Drugs. But Can They Trust Them?

As Mr. Baker followed clues, he unearthed devious practices at what seemed like transparently run plants. Technicians used initial hidden tests to get preliminary results, which then guided them as they tinkered with the test settings. Then they retested in the plant’s official system to get the desired results showing that the drugs fell within specifications. Those drugs with altered test results could then be released to patients.

During his 27 months in India, of the 38 drug plants he inspected, Mr. Baker found fraudulent or deceptive data in 29 of them. As he worked, other investigators learned from his techniques. The difference was stunning, one F.D.A. employee recalled. “Like you walk into a dark room and suddenly someone just turns on the light,” the employee said. “It was shocking.”

And there was an additional negative consequence to the F.D.A.’s system of advanced notice. With the companies serving as travel agents, F.D.A. investigators spoke of inappropriate perks: hotel upgrades, for which the investigators would never see a bill; golf outings, massages, and trips to the Taj Mahal. The result was what some F.D.A. employees referred to as “regulatory tourism.” The F.D.A. said in response that “any allegations of improper conduct by F.D.A. personnel are investigated.”

A new head of the F.D.A.’s India office, Altaf Lal, arrived in mid-2013. To tame the twin problems of company fraud and compromised investigators, Mr. Lal made a novel pitch to agency officials. He proposed a pilot program to make all inspections in India either on short notice or unannounced. By December 2013, he had a green light. The results were instantaneous.

In January 2014, the F.D.A. was planning an unannounced inspection at a plant in northern India on a Monday. Fearing that plant officials had heard they were coming, Mr. Baker and his colleague went a day early, unannounced. They proceeded to the quality control laboratory, expecting it to be quiet on Sunday morning. Instead, they were stunned to see a hive of activity. Dozens of workers hunched over documents, backdating them. On one desk, Mr. Baker found a notebook listing the documents the workers needed to fabricate in anticipation of the inspectors’ arrival. There were Post-it notes stuck to some surfaces, noting what data to change.

In large swaths of India’s generic drug industry, the pilot program uncovered a long-running machinery dedicated not to producing perfect drugs but to producing perfect data. At one plant, Mr. Baker went straight to the microbiology laboratory and found the paperwork for testing the sterility of the plant in perfect order: microbial limits testing, biological indicators, all the samples with perfect results. Yet most of the samples didn’t exist. The plant was testing almost nothing. The laboratory was a fake.

At the vast majority of the unannounced inspections, the investigators found things the plants no longer had time to fix: Infestations of birds and insects. A pile of critical manufacturing records, tossed in a trash bin. An employee bathroom near a sterile manufacturing area in one plant lacked drainage piping, so urine puddled directly onto the floor.