SURPASS-5 asks a narrower question than the rest of the program: not whether tirzepatide works, but what it does for someone already taking insulin and still above target. That population is the one most likely to be offered a second injection, and it is also the one where the class’s most-repeated safety claim — that these drugs do not cause low blood sugar — stops holding. The trial that swapped tirzepatide for basal insulin is covered in the SURPASS-3 article; this one stacks it on top. What combining the two involves in practice is in the insulin article.
What was run, and what the control arm was allowed to do
A randomized, double-blind phase 3 trial at 45 medical research centers and hospitals in eight countries enrolled from August 2019 to March 2020 and completed follow-up in January 2021. 475 adults with type 2 diabetes and inadequate control on once-daily insulin glargine, with or without metformin, were randomized 1:1:1:1 to once-weekly subcutaneous tirzepatide 5 mg (n = 116), 10 mg (119), 15 mg (120) or volume-matched placebo (120) for 40 weeks. Everyone on active drug began at 2.5 mg and moved up one 2.5 mg step every fourth week until their allocated dose was reached. Mean age was 60.6 years (SD 9.9), 211 (44%) were women, and mean glycated hemoglobin was 8.31% (SD 0.85).[1] Entry required a glycated hemoglobin of 7.0% to 10.5%, a body-mass index of at least 23, and at least three months on insulin glargine.[2]
The comparator is the thing to get right. “Placebo” here means a matching weekly injection added to insulin that was still being titrated. Over the 40 weeks, the daily mean insulin glargine dose changed by +13.0% (SE 7.34), +8.1% (7.03) and −11.4% (5.85) across the three tirzepatide arms, and by +75.0% (11.11) in the placebo arm.[2] The registry’s between-arm estimates for that row are modeled on a log-baseline scale and are not the arithmetic difference of those four figures, so only the arm figures are quoted here.
That single line reframes everything below it. The control group was not untreated and did not stand still; it got three-quarters more insulin, and whatever it achieved on glucose it achieved by taking more of a drug that causes weight gain. Only the 15 mg arm came off insulin at all.
The primary endpoint, and what it actually covered
The preregistered primary endpoint was the mean change in glycated hemoglobin at week 40 — for the 10 and 15 mg doses only. Those two comparisons carry 97.5% confidence intervals, reflecting the multiplicity control applied to them; the 5 mg comparison is a separate secondary outcome reported with a 95% interval.[2][1] A summary that presents three doses as one primary result has flattened the trial’s own hierarchy.
The paper and the registry then report that endpoint differently, because they analyze it differently. In the abstract, glycated hemoglobin fell 2.40 points at 10 mg and 2.34 at 15 mg against 0.86 on placebo, giving differences of −1.53 (97.5% CI, −1.80 to −1.27) and −1.47 (−1.75 to −1.20), both at P < .001; the 5 mg arm fell 2.11, a difference of −1.24 (95% CI, −1.48 to −1.01).[1] In the posted results the same arms fell 2.59 (SE 0.081), 2.59 (0.083) and 2.23 (0.081) against 0.93 (0.079), with differences of −1.66 (95% CI, −1.88 to −1.43), −1.65 (−1.88 to −1.43) and −1.30 (−1.52 to −1.07).[2]
Both accounts belong to this trial and neither is wrong; they condition on different things, in the familiar direction where the analysis that assumes everyone stayed on treatment flatters the drug and penalizes the control. Quote one set or the other, never a figure from each. And note what happens at the top of the ladder in the abstract’s numbers: 15 mg came in 0.06 points behind 10 mg on the primary endpoint. In the registry’s numbers the two are identical to two decimal places. On blood sugar, this trial found no benefit at all in the highest dose, which is a fact the dose ladder discussion in the tirzepatide dose article has to accommodate.
Where the top dose does earn its place
Weight separates cleanly and in dose order. The abstract reports changes of −5.4, −7.5 and −8.8 kg against +1.6 kg on placebo, with differences of −7.1 kg (95% CI, −8.7 to −5.4), −9.1 (−10.7 to −7.5) and −10.5 (−12.1 to −8.8).[1] The registry reports −6.2 (SE 0.58), −8.2 (0.58) and −10.9 (0.59) against +1.7 kg (0.57), with differences of −7.8 (−9.4 to −6.3), −9.9 (−11.5 to −8.3) and −12.6 (−14.2 to −11.0).[2]
A twelve-and-a-half-kilogram spread is not twelve and a half kilograms of weight loss. Roughly ten-point-nine of it is what tirzepatide did and one-point-seven is what escalating insulin did in the other direction. Losing at least 5% of body weight ran 53.91%, 64.60% and 84.62% against 5.93% on placebo, with odds ratios of 17.15 (95% CI, 7.55 to 38.93), 27.24 (11.87 to 62.55) and 79.61 (32.76 to 193.44).[2] A glycated hemoglobin in the nondiabetic range, below 5.7%, was reached by 26.09% of the lowest arm, 47.79% of the middle one and 62.39% of the top one, against 2.54% of the control group.[2] Both of those climb with dose even though the primary endpoint does not.
Reaching below 7.0% is the third place the two sources diverge. The abstract describes 85% to 90% of tirzepatide participants against 34% on placebo; the registry posts 93.04%, 97.35% and 94.02% against 33.90%.[1][2] The control arm is the same in both, and a third of a group on escalating basal insulin reaching target is a working comparator rather than a straw one.
The hypoglycemia column, and what it reverses
Episodes below 54 mg/dL or severe occurred at rates of 0.49 (SE 0.141), 0.66 (0.169) and 0.38 (0.099) per participant-year on tirzepatide 5, 10 and 15 mg, against 0.51 (0.149) on placebo.[2] The 10 mg arm sits above the control arm. There is no dose ordering, no separation, and no reassurance to be had from any of the four numbers. Serious hypoglycemia was reported in one participant at 10 mg and one at 15 mg, and none on placebo.[2]
Read those next to the monotherapy trial and the point becomes sharp. In SURPASS-1, the same outcome measured the same way ran 0.02 episodes per participant-year on every tirzepatide dose against 0.04 on placebo. Here every arm is more than an order of magnitude higher. The property people quote — that these drugs stimulate insulin only when glucose is high, so they do not cause hypoglycemia — is a statement about the drug given alone. Put it on top of injected insulin and the risk belongs to the insulin, which is why the useful result in this trial is not the hypoglycemia rate but the dose column above it. The wider picture is in the hypoglycemia article.
Who stopped, and who left
These are two different figures and the gap between them is the largest in the program. The abstract reports that treatment was prematurely discontinued by 10% of the 5 mg group, 12% of the 10 mg group, 18% of the 15 mg group and 3% of the placebo group, while 451 of 475 (94.9%) completed the trial.[1] The registry posts study non-completion of 7, 4, 10 and 3 across the four arms — 8.3% at the top dose.[2]
So at 15 mg roughly eighteen in a hundred came off the drug and eight in a hundred left the trial: more than half of the people who stopped taking tirzepatide stayed in and kept contributing data. That is exactly the population whose treatment the two analyses handle differently, and it is the mechanical reason the abstract’s numbers and the registry’s differ. Of the ten departures at 15 mg, two were attributed to an adverse event, five to withdrawal by the participant, two to protocol violation and one to loss of follow-up.[2]
Serious adverse events were 9 of 116, 13 of 119, 9 of 120 and 10 of 120 — no dose ordering, and no death in any arm.[2] Cardiac failure was recorded as a serious event in three participants at 5 mg, one at 10 mg, none at 15 mg and none on placebo; the trial was not powered to compare any of this. The symptom columns behave as expected: nausea 15, 21 and 22 against 3; vomiting 8, 9 and 15 against 3; diarrhea 14, 15 and 25 against 12; constipation 7, 8 and 8 against 2.[2] The abstract summarizes the same table as diarrhea 12–21% against 10% and nausea 13–18% against 3%.[1]
What SURPASS-5 did not establish
It adjudicated no cardiovascular outcome and reported no event composite; with 475 participants over 40 weeks and forty-one serious events in total, it could not have. It ran 40 weeks, so it says nothing about the second year or about what happens when either injection stops. It measured no body composition and no imaging endpoint, so the ten-point-nine kilograms is undifferentiated. It compared tirzepatide against placebo and against nothing else: no GLP-1 comparator, no alternative intensification such as prandial insulin or an SGLT2 inhibitor, so it does not establish that adding this drug beats the other things a clinician would consider at that point.
The population is narrow by construction. Everyone had type 2 diabetes, everyone was already on once-daily insulin glargine, mean age was 60.6, and mean glycated hemoglobin was 8.31%. Enrollment was 129 in Germany, 94 in Czechia, 82 in Japan, 57 in Spain, 36 in Poland, 34 in the United States, 31 in Slovakia and 12 in Puerto Rico; 380 of 475 participants were White and 6 were Black or African American.[2] Nothing in the trial describes a person without diabetes, and nothing describes a person not on insulin.
One more limit is about the record rather than the design. The paper is in JAMA and a PubMed Central record exists, but the publisher does not permit full-text retrieval, so only the abstract could be read. Anything the full text reports that is not in the abstract and not in the registry is outside what this page can say.
What was in the pen
Every figure here belongs to branded, FDA-approved tirzepatide supplied free, escalated under protocol at 2.5 mg every four weeks to a fixed assigned dose, injected weekly for 40 weeks alongside a titrated basal insulin managed by a study team, in adults with type 2 diabetes. Sellers listed on the compounded tirzepatide board dispense compounded preparations, which are not FDA-approved and are not reviewed by the FDA for safety, efficacy or quality before they reach a patient.
The number to be careful with here is 12.6. It is the gap between an arm that lost weight and an arm whose insulin dose rose three-quarters, in people whose diabetes was already being treated with an injection. And the sentence to retire when reading this trial is the one about these drugs not causing low blood sugar: in the only tirzepatide trial run on top of insulin, the 10 mg arm recorded more of it than placebo did.