I once found myself entrapped by a patient as much as she felt trapped by me. It was the summer of 2001, and I was running a small internal-medicine clinic, supervised by a preceptor, on the fourth floor of a perpetually chilly Boston building. Most of the work involved routine primary care — the management of diabetes, blood pressure and heart disease. It was soft, gratifying labor; the night before a new patient’s visit, I would usually sift through any notes that were sent ahead and jot my remarks in the margins. The patient’s name was S., I learned. She had made four visits to the emergency room complaining of headaches. Three of those times she left with small stashes of opioids — Vicodin, Percocet, oxycodone. Finally, the E.R. doctors refused to give her pain medicines unless she had a primary-care physician. There was an open slot in my clinic the next morning, and the computer had randomly assigned her to see me.

We were living, then, in what might be called the opioid pre-epidemic; the barometer had begun to dip, but few suspected the ferocity of the coming storm. Pain, we had been told as medical residents, was being poorly treated (true) — and pharmaceutical companies were trying to convince us daily that a combination of long- and short-acting opioids could cure virtually any form of it with minimal side effects (not true). The cavalier overprescription of addictive drugs was bewildering: After a tooth extraction, I emerged from an oral surgeon’s office with a two-week supply of Percocet.

When S. came to see me, she was pleasant and self-possessed, a former office worker who had lost her job and was living with her daughter. She offered a packet of mints. I asked her about the headaches, and she told me a story riddled with contradictory clues. At times, the headaches arose with no triggers or warnings. At other times, they resembled migraines: There was a premonition — an “aura,” in medical terms — during which the world smelled different and every sound was magnified. The headaches appeared in the front of her head and in the back. Sometimes she felt nauseated; at other times, she was hungry. The only medicines that had ever helped were opioids.

The first call from S. came the next morning. She had woken up with a headache, and the pain was ascending. Could I call in a prescription for Vicodin? I needed more time to think through the case, I stammered, but barely 10 minutes had passed before there were four more notes on my pager. “My whole face is swollen with pain,” she said. She needed medicines immediately.

I asked her to come and see me in the clinic and rushed through patients to clear half an hour of my schedule. By the time she was in the examining room, her body was evidently contorted in pain. Was this a real headache? Or was she in opioid withdrawal? Was this a ruse? I agreed to give her some non-opioid pain relievers on the condition that she would come immediately to the hospital the next time she felt the first twinge of pain.

A week later, she paged me. None of the medicines had worked, she said. I reminded her of our agreement and admitted her to the hospital. By the time the neurology team came to see S. that afternoon, she was hovering over the nurse’s station, spitting with rage. She accused them of not taking her pain seriously. By then, she had already paged me from the hospital — nearly a dozen times. Each time I tried to focus on a new patient, my pager would ring urgently. The hospital’s page operator called to ask what she should do. I consulted my preceptor, who was sympathetic but unsure of the choices. Was it legal to block calls from a patient? Was it ethical? We didn’t know. There were no rules, no guidelines to work with.

The neurology team recommended a head CT and a volley of tests. Nothing was diagnostic. The neurologist told me, shrugging his shoulders, that he couldn’t say whether this headache was real or what might have caused it. But without a diagnosis, we should not give her opioids. Of course, but what was the alternative? Not his area of expertise, he replied, and moved on to his next case.

I sent S. to a pain clinic, run by a seasoned anesthesiologist. She would be given a strict schedule of long-acting pain-relieving drugs, with no extra pills. And he’d slowly wean her off them and substitute nonaddictive pain medicines. It seemed like a workable plan, and I discharged her from the hospital.

I was in the I.C.U., inserting a central venous catheter — the patient was draped in sterile sheets under me — when there was another stream of pages from S. I had to halt the procedure to get on the phone; it was too distracting to continue. She was frothing with indignation. She’d decided to abandon the pain clinic. The doctor was not sympathetic, she said. Her pain was real, and he was minimizing it. She had fired him.

Could I fire her as my patient? I went to see the chief of my training program. He offered to help, but there was little that he could do; part of our work, he said, was to deal with “difficult patients.”

Over the next six months, I received hundreds of pages, phone calls, threats and entreaties from S. I tried prescribing methadone through a hospital program. It helped — it would turn out to be the only thing that helped — but she refused to continue the program. I referred her to a psychiatrist. I read books on headaches, on addiction, migraines, malingering, the ethics of abandoning the ill and the breakdown of a doctor-patient relationship. Late one night that fall, one of the E.R.s in the city paged me. S. had come in with a headache. By the time I spoke on the phone with the doctors, they had discharged her with a handful of Vicodins. A week later, she was in a walk-in clinic in northern Boston. Fed up with her yelling in the clinic, they had sent her home with a week’s cache of Percocets.

I’m not writing this to expiate the guilt of doctors or excuse my complicity but as a testimony of the 12-odd months of mutual assured destruction that characterized my first encounter with addiction. We had no training in this kind of medicine. If the addict was helpless, then so, too, was the young physician: To try to treat addiction was to discover an inverted form of doctoring, in which the patient and doctor turned into wary, suspicious aliens circling and jabbing at each other. Medicine depends implicitly on a therapeutic alliance between a doctor and a patient, but addiction, I learned, distorts that alliance. The doctor shifts from healer to dealer. To the addict, the doctor is contorting the truth; to the doctor, it’s the addict who is constantly inverting reality. The doctor is, at first, the enabler and the supplier, and then the tormentor, the withholder, the liar, the enemy. Perhaps the reason that the methadone clinic had worked for S., albeit fleetingly, I realized, was not only because of the drug’s pharmacological effect but also because the clinic had forcibly re-established the alliance: There was, at least, a transactional transparency in what she was giving and getting. Rather than trying to fix her, here, at last, was a doctor who would willingly give her a fix.

A few months ago, I was a visiting professor at the hospital in Boston where I first met S. I skipped the usual tradition of delivering a long, didactic lecture and joined the interns on their medical rounds. Every one of their medical practices, I learned, had an addicted patient. The residents now possessed the vocabulary and skills that I had lacked. They had learned the street slang for the drugs; they understood how to navigate some of the legal dilemmas of treatment and relapse. Some had visited methadone clinics in Boston, and they had memorized the complications of co-dosing with molly and China white. But many of them seemed to be reliving the fundamental uncertainties that I encountered 16 years before: How do you establish an alliance with a patient when the relationship is scarred by suspicion and mistrust? It was reassuring, yes, to see them trying to understand the impulses that drove an addict. But I wondered whether we also needed to understand the impulses that drive a doctor’s relationship with an addicted patient: the failure of our credibility, our bizarre circumlocutions, our helplessness, the chronic relapses into self-doubt and disappointment. It was as if they had seen a fragment of a torn map, or solved only half of a puzzle.

Sometime in the fall of 2002, S. vanished. The pages stopped; the scheduled appointments were missed. I tried to call her at home; the phone rang insistently, as her pages had, but there was no response. Perhaps she had moved to another city, or found another doctor. I never saw her again. In 2006, I heard that she had died of an overdose. I’m still trying to figure out how I might have treated her differently.