How a “Best Dose” Becomes a Headline Number
This chapter follows the number at the center of the aerobic-exercise review: how reported exercise details become MET-minutes per week, how that exposure is modeled against postpartum outcomes, and where the apparent precision stops. The arithmetic is accessible, but the evidence question is not simply “How many minutes should someone exercise?” It is whether a harmonized, partly estimated study-arm dose can reveal a useful pattern without becoming an individualized prescription.
The arithmetic is simple; the harmonization is not
A MET is a unit intended to represent the energy cost of an activity relative to rest. The review expresses weekly aerobic exercise dose as MET-minutes per week. In its basic form, the calculation multiplies three ingredients:
weekly dose = intensity in METs × minutes per session × sessions per week
A program described as four METs for 30 minutes on three days a week would therefore correspond to 360 MET-minutes per week. At the same nominal intensity, 45 minutes on three days would be 540 MET-minutes. These examples show what the metric combines: intensity, time, and frequency. They do not show how much benefit any particular person should expect.
That distinction matters because the three ingredients are not equally direct observations. A trial might specify intensity with a heart-rate target, a percentage of oxygen-consumption capacity, or a perceived-exertion scale. The review converted such descriptions into MET values. When intensity was not explicitly reported, the investigators used the activity type and workload described in the trial and mapped them to the closest value in the 2011 Compendium of Physical Activities. The resulting dose is a harmonized estimate, not a laboratory measurement taken in the same way from every participant.
The mapping is useful because studies rarely describe exercise in a perfectly common language. “Moderate walking” or “supervised aerobic activity” may be recognizable to a human reader but insufficiently precise for a direct comparison. A shared metric lets the analysis place different schedules on one axis. Yet the conversion also introduces assumptions. An activity label may conceal different speeds or workloads. The same nominal activity can feel different to people with different fitness, sleep, pain, or recovery. Individual energy expenditure varies. If the original report is vague, the numerical dose inherits that vagueness.
The review also rounded doses to the nearest 250 MET-minutes per week and grouped them into categories of 250, 500, 750, 1,000, and 1,250. Rounding helped connect the available evidence for dose-response modeling. It did not make the underlying programs identical. A value near one category boundary could be placed with programs that were not meaningfully the same in session pattern, intensity, supervision, or adherence. The categories are therefore a statistical bridge across heterogeneous trials, not a set of finely calibrated rungs for personal progression.
There is another important distinction between prescribed exposure and received exposure. A trial arm may be assigned a particular weekly load, but participants can miss sessions, modify effort, or stop early. The review collected adherence where it was available, yet the dose variable itself is built from the program’s reported characteristics. It does not automatically become a record of what each woman completed. This is the point at which a study-arm metric diverges from an individual dose: the formula describes the intervention on paper, while the lived burden includes effort, recovery, access, support, and the ability to return for the next session.
The formula can also conceal different kinds of work behind the same total. At four METs, 50 minutes on three days equals 600 MET-minutes per week. At five METs, 30 minutes on four days also equals 600. The totals match, but the schedules differ in session length, frequency, and nominal intensity. They may place different demands on time, recovery, transport, supervision, and childcare. The review’s common metric makes comparison possible; it does not demonstrate that every combination producing the same total is interchangeable.
This is a general problem with composite exposures. Increasing the total can mean adding minutes, adding sessions, increasing intensity, or some mixture of the three. Each route changes the experience of the program, even when the multiplication gives the same answer. A person who can manage several brief sessions may face a different burden from someone who can protect one longer block of time. Conversely, a higher nominal intensity may be harder to sustain even when it reduces total minutes. The supplied evidence does not identify which component should be adjusted first for an individual.
The metric is therefore best understood as a coordinate system for the trial literature. It places programs on a common axis so a model can look for a pattern, but it does not describe every dimension that matters to lived exercise. It leaves out the texture of effort, the reason a session was missed, the support that made attendance possible, and the recovery cost that followed. Those omissions do not undermine the arithmetic. They explain why the arithmetic should orient interpretation rather than dictate behavior.
The chain can be made explicit:
Each handoff is a possible failure point for overinterpretation. If intensity is missing, it must be estimated. If the activity description is broad, the estimate may be misclassified. If attendance is incomplete, nominal and received exposure diverge. If the dose is rounded, the model sees categories rather than exact schedules. None of these makes the analysis useless; they define how strongly its output should be stated.
Two curves, two meanings of optimal
Once the dose values have been assembled, the review asks whether the relationship with outcomes is linear or nonlinear. A linear reading would suggest that more exercise is associated with a steady increase in benefit. The selected models instead describe curved relationships. That is why the review can report a middle range that appears more favorable than both the lower and higher ends without claiming that more is always better.
For depressive symptoms, the reported relationship is U-shaped. Improvement was observed across approximately 380 to 790 MET-minutes per week, with the modeled optimum near 570 MET-minutes per week. The authors interpret the pattern as a balance: too little activity may provide too small a stimulus for sustained behavioral or physiological change, while too much may increase fatigue, discomfort, competition with caregiving and sleep, and the risk that the program becomes another stressor. In that interpretation, an intermediate volume is attractive not because 570 is a biological switch, but because it may sit near the point where potential benefit and practical burden balance in the studies included.
The wording “modeled optimum” is doing important work. The analysis did not randomly assign postpartum participants to every possible dose and then observe which exact amount caused the largest response. It used available trial data, some of whose intensities were reconstructed, to fit a dose-response curve. The peak is therefore a summary of the model across the evidence it received. It may move if better trials add different populations, intensities, or adherence patterns. It also does not say where a person should begin, how quickly to increase activity, or what to do when recovery makes the nominal volume unsustainable.
Quality of life produces a different curve. Six studies reported HRQoL, and the review found an improvement with an SMD of 0.45 and a 95% confidence interval from 0.29 to 0.60. The dose-response relationship was described as an inverted U: gains increased up to about 420 MET-minutes per week and diminished beyond roughly 690. The apparent peak is lower than the depression-symptom peak. That difference is not an inconsistency to be averaged away. It demonstrates that “optimal” depends on the outcome being optimized.
The reason is visible in what HRQoL measures. It can include physical functioning, vitality, emotional well-being, social participation, and the ability to manage daily tasks. A moderate amount of activity may improve a person’s sense of capability or daily energy without requiring the same exposure that is associated with the largest average change in depressive symptom scores. At higher volumes, fatigue, time demands, musculoskeletal strain, or interference with sleep and caregiving may erode those functional gains. The curve is a reminder that a broad quality-of-life outcome has a different relationship to burden than a symptom score may have.
The two peaks should not be turned into competing commandments—one number for depression and another for quality of life. They are better read as a warning against optimizing an unnamed outcome. If the decision concerns depressive symptoms, the depression model is the more direct reference. If the concern is daily functioning or overall well-being, the HRQoL model supplies a different signal. If anxiety is the goal, this analysis does not provide a comparable curve at all; the review had too few anxiety studies for subgroup or dose-response modeling. The metric is shared, but the evidentiary meaning is not.
A nonlinear curve is not the same thing as a biological threshold. The model can show that the observed pattern bends, but the bend does not identify a switch at which benefit suddenly begins or ends. It also cannot tell us whether a lower volume would work for a particular person, whether a higher volume would be tolerable after gradual adaptation, or whether an apparent decline at the upper end reflects fatigue, lower adherence, differences between study populations, or other features that travel with higher-volume programs. The review discusses these possibilities, but the curve itself does not test them separately.
Because the dose values describe study arms, the model primarily relates program profiles to average outcome changes. It is not an individual-participant analysis in which the same person is observed across a sequence of increasing doses. That distinction limits the causal story. The peak identifies where the fitted relationship is highest within the available evidence; it does not prove that increasing a person’s volume toward that point will move them along the same path. New trials with clearer intensity reporting, more complete adherence data, and participants at different postpartum stages could change both the shape and the location of the curve.
The authors make the depression and quality-of-life results more tangible by translating an approximate 400–600 MET-minute range into roughly 100–150 minutes of moderate-intensity activity at about four METs. They illustrate this as 30–45 minutes on two to four days per week, or shorter bouts that may fit childcare and recovery. The calculation is straightforward: divide the weekly MET-minute range by an assumed intensity of about four METs. Its value is explanatory. It turns an unfamiliar research unit into a quantity readers can picture.
Its assumptions should remain attached. “About four METs” is an approximation, not a personal measurement. The minute-based example does not establish that one exact schedule was tested as a single protocol, that two shorter bouts are equivalent to one longer session for every outcome, or that anyone should force a nominal target when recovery or symptoms make it inappropriate. The review’s own dose estimate already contains mapped intensities, rounded categories, and heterogeneous programs. A calculation made from it cannot be more precise than the evidence underneath.
The same range can also be reached through different ordinary schedules. Four METs for 30 minutes on four days would be 480 MET-minutes; four METs for 45 minutes on three days would be 540. Those examples fall within the authors’ approximate translation, but they are not interchangeable instructions. One person may have more access to several short windows; another may have one reliable block of time. The evidence gives no basis for declaring one arrangement universally superior, and it does not measure the personal cost of fitting either arrangement around feeding, sleep, pain, or other demands.
This is also why the authors’ practical range should not be read as a pass-or-fail threshold. A person who completes less than the modeled range has not been shown to receive no benefit; a person who reaches it has not been shown to receive the maximum possible benefit. The curve describes average patterns, and the confidence around those patterns matters. The range can orient a conversation about a feasible amount of structured activity, but it does not replace judgment about actual response, burden, adherence, or clinical context.
The most defensible conclusion is therefore modest but useful. MET-minutes provide a common language for comparing programs. The depression curve suggests an intermediate range associated with better average symptom results, and the HRQoL curve suggests a different intermediate range for a broader functional outcome. Both signals support looking beyond “more is better.” Neither establishes a precise dose that causes the same result for every postpartum person. To understand why the difference matters in practice, the next question is not only how much activity was assigned, but which outcome—and which person—was expected to change.