Three numbers circulate in every oxygen-therapy decision in Indian respiratory practice — SpO₂ from a fingertip oximeter, SaO₂ from an arterial blood gas, PaO₂ from the same arterial sample. They are routinely treated as interchangeable at the bedside. They are not. The confusion is responsible for a steady rate of under- and over-prescription of long-term oxygen therapy, and for the familiar misconception that a patient reading SpO₂ 92% is “eight per cent short of oxygen” and therefore desperately hypoxic. This article pulls apart what each number measures, why the pulse oximeter is both the most useful and the most misleading device in the respiratory clinic, and when the ₹900 ABG machine at the next-nearest hospital becomes the only honest source of data.
The target audience is physicians and respiratory therapists making prescription decisions, home-care staff handing over devices to patients, and engaged patients trying to make sense of the readings they see at home.
Three numbers, three physical quantities
PaO₂ is the partial pressure of oxygen dissolved in arterial plasma, reported in mmHg (or kPa in SI units). It is the driving pressure that pushes O₂ from plasma onto haemoglobin in the pulmonary capillary and from haemoglobin into the mitochondrion at the tissue end. PaO₂ is obtained from an arterial blood gas sample — typically a radial-artery puncture, occasionally femoral or brachial — analysed within 15 minutes on a blood-gas machine. At sea level, normal adult PaO₂ is 80–100 mmHg. There is a predictable fall with age; for adults breathing room air, PaO₂ ≈ 100 − (age/3) mmHg is a workable estimate.
SaO₂ is the fraction of arterial haemoglobin binding sites actually loaded with oxygen, expressed as a percentage. It is measured by CO-oximetry — a multi-wavelength photometer that distinguishes oxyhaemoglobin from deoxyhaemoglobin, carboxyhaemoglobin (COHb), and methaemoglobin (MetHb). SaO₂ is the gold standard saturation number in clinical research and in forensic-grade decisions. Normal adult SaO₂ at sea level is 95–100%.
SpO₂ is peripheral capillary oxygen saturation as estimated non-invasively by a pulse oximeter. A red (660 nm) and an infrared (940 nm) LED shine light across a pulsatile vascular bed — usually a fingertip — and a photodiode measures the transmitted signal. Because oxyhaemoglobin and deoxyhaemoglobin have different absorbance ratios at the two wavelengths, a calibration curve maps the measured ratio to a saturation estimate. SpO₂ approximates SaO₂ but is not identical to it. Normal adult SpO₂ at sea level is 95–100%.
The relationships:
- PaO₂ and SaO₂ are linked by the oxyhaemoglobin dissociation curve.
- SaO₂ and SpO₂ are supposed to agree within ±2 percentage points under ideal conditions. They often do not.
The oxyhaemoglobin dissociation curve
The critical physiology every prescriber should carry in their head is the sigmoid relationship between PaO₂ and SaO₂. The curve is not linear.
- PaO₂ 100 mmHg → SaO₂ ~98%
- PaO₂ 80 mmHg → SaO₂ ~96%
- PaO₂ 60 mmHg → SaO₂ ~90%
- PaO₂ 55 mmHg → SaO₂ ~88%
- PaO₂ 40 mmHg → SaO₂ ~75%
The near-horizontal upper plateau means that at PaO₂ 80–100 mmHg, SaO₂ barely changes. The steep mid-slope around PaO₂ 60 mmHg means that a small PaO₂ fall from 60 to 50 mmHg drops SaO₂ from ~90% to ~83%. Most LTOT prescription thresholds live in this steep zone (GOLD Report).
The curve shifts with physiology:
- Right shift (lower affinity — harder to saturate, easier to release at tissue): acidosis, hypercapnia, hyperthermia, elevated 2,3-DPG.
- Left shift (higher affinity — easier to saturate, harder to release): alkalosis, hypocapnia, hypothermia, carboxyhaemoglobinaemia, fetal haemoglobin, stored red cells.
A septic, acidotic COPD patient with SpO₂ 92% has a true PaO₂ that is lower than the textbook curve predicts for a well-compensated patient. Trending SpO₂ alone in a metabolically disturbed patient is not defensible.
”SpO₂ 92% is not 8% short”
The commonest lay misunderstanding — and unfortunately a misunderstanding seen in some clinical documentation too — is that an SpO₂ of 92% means the patient is “8% short of oxygen” and needs urgent correction. That framing is wrong on two counts.
First, SpO₂ is the percentage of haemoglobin binding sites loaded with oxygen, not the percentage of the body’s oxygen requirement being met. A well-perfused patient with SpO₂ 92%, normal haemoglobin, and normal cardiac output is delivering plenty of oxygen to tissue. The oxygen content of arterial blood (CaO₂) depends on haemoglobin concentration, SaO₂, and dissolved oxygen — with haemoglobin being the dominant term.
Second, the dissociation curve means 92% corresponds to PaO₂ roughly 65 mmHg at the sea-level, normal-pH patient — well above the LTOT threshold. The patient is not in respiratory failure. A stable COPD patient with consistent SpO₂ 92% at rest does not qualify for chronic oxygen therapy under any published guideline.
The useful framing at the bedside: SpO₂ above 92% is reassuring, SpO₂ 88–92% warrants investigation (trend, perfusion, ABG if decisions are pending), and SpO₂ below 88% is actionable in a stable patient.
When to trust the pulse oximeter
A fingertip pulse oximeter is accurate to within ±2 percentage points of true SaO₂ when the following conditions hold:
- The patient is well-perfused (warm, pink, adequate cardiac output, PI ≥ 1.0).
- Haemoglobin is within normal range (10–18 g/dL).
- No carboxyhaemoglobin or methaemoglobin is present in clinically meaningful quantity.
- The oximeter is a clinical-grade device with motion-artefact rejection.
- The probe is appropriately sized and sited.
- A clean pulsatile waveform is visible on the device display.
Under these conditions, SpO₂ is a reasonable proxy for SaO₂ and can anchor clinical decisions. This is the vast majority of outpatient and inpatient encounters.
When not to trust the pulse oximeter
The failure modes every prescriber should recognise:
Cold extremities and poor peripheral perfusion. The oximeter requires a pulsatile arterial signal. In a vasoconstricted patient — cold, hypovolaemic, on high-dose vasopressors, or in shock — the signal is degraded and the reading may drift by 5 percentage points or more. Warm the finger under a cloth or a brief warm-water immersion before a diagnostic SpO₂. Check the perfusion index (PI) if the device reports it; PI below 0.4 is a caution flag.
Nail polish, henna, and artificial nails. Dark nail polish and fresh henna — common in Indian female patients — alter light transmission. Remove polish with acetone, or position the sensor sideways across the finger, or use the earlobe. Acrylic nails should be removed or avoided as a site.
Skin pigmentation bias. A substantial body of work from 2020 onward has shown that pulse oximeters systematically over-read SpO₂ in patients with darker skin pigmentation — that is, the device reads a higher number than the patient’s true SaO₂. In US data, the over-reading is clinically material at the LTOT threshold — a patient reading SpO₂ 92% may truly sit at SaO₂ 88–90%. The mechanism is still being characterised, but melanin absorbance at the calibration wavelengths and historically under-representative calibration datasets are both implicated. Indian skin tones span Fitzpatrick IV–VI across much of the population; the direct transferability of US evidence has not been formally established, but the mechanism suggests Indian patients are also susceptible. Clinical implication: at the LTOT decision boundary, confirm with ABG.
Carboxyhaemoglobin. COHb absorbs red light almost identically to oxyhaemoglobin. The pulse oximeter reads it as O₂Hb. In a heavy smoker, chronic biomass-cookfire exposure (common in rural Indian homes), or acute CO poisoning, SpO₂ grossly overestimates true SaO₂. A smoker with COHb 10% reading SpO₂ 98% may truly sit at SaO₂ 88%. ABG with CO-oximetry is mandatory in suspected CO exposure.
Methaemoglobin. MetHb has an absorbance profile that pushes the oximeter reading toward ~85% regardless of true SaO₂. Patients exposed to dapsone (leprosy therapy, Pneumocystis prophylaxis), topical anaesthetics (benzocaine, prilocaine), nitrates, or aniline dyes (occasional Indian occupational exposure in textile dyeing) should have MetHb considered when SpO₂ sits anomalously near 85% and does not respond to supplemental oxygen.
Severe anaemia. The oximeter reports the fraction of remaining haemoglobin that is saturated, not the total oxygen content. A patient with haemoglobin 6 g/dL and SpO₂ 99% has dangerously low oxygen-carrying capacity. The oximeter cannot see this.
Motion artefact. Tremor, shivering, and positioning movement corrupt the waveform. Clinical-grade units (Masimo SET, Nellcor OxiMax) use signal-extraction algorithms to discriminate true pulses from motion. Cheaper consumer units do not; readings taken during movement should be disregarded.
Ambient light. Strong fluorescent or surgical lighting can contaminate the photodiode signal. In bright sunlight on an outpatient ward, cover the sensor with an opaque cloth for a diagnostic reading.
Indian-context oximeter quality
The Indian consumer oximeter market is dominated by imported devices sold under many brand names, some branded, many effectively unbranded at the lower price tiers. Typical street prices in 2026:
- Branded clinical-grade (Masimo, Nellcor modules) — ₹8,000–25,000, hospital channels.
- Branded consumer (Nonin, Beurer, Philips Respironics fingertip, Omron fingertip) — ₹2,500–6,000.
- Mid-tier (BPL, Dr Trust, Hicks, Accusure, Control D) — ₹800–2,500.
- Unbranded or generic fingertip units — ₹300–800.
The lower the price, the more likely that the device’s internal SpO₂ calibration curve was derived from a small, demographically narrow dataset, that motion-artefact handling is absent, and that the LED wavelengths drift over time. At the COVID demand-surge peak in 2020–21, many ₹400 devices entered Indian homes; a meaningful fraction are still in active use in 2026. For a serious LTOT decision, a single reading from an uncalibrated three-year-old consumer device is not an adequate basis. For general wellness trending at home, such devices remain useful — the error is systematic, and trends (up or down) are often more informative than the absolute number.
A pragmatic recommendation: a clinic should use a branded clinical-grade oximeter for diagnostic decisions and accept consumer devices for trending. Patients starting LTOT should be shown their saturation on the clinic-grade device and on their home device on the same visit so they understand the offset between the two.
When the ABG is justified
PaO₂ confirmation via ABG is indicated in:
- LTOT qualification when SpO₂ sits borderline (86–90%) at rest on two or more readings in a stable patient. The GOLD criterion is PaO₂ ≤55 mmHg or SaO₂ ≤88% measured under stable conditions (GOLD Report).
- Suspected CO₂ retention — morning headache, daytime somnolence, plethora, cor pulmonale in a COPD patient. SpO₂ does not detect hypercapnia; ABG reports PaCO₂ directly.
- Acute respiratory failure triage — emergency department, ICU admission. SpO₂ is for trending; ABG defines the failure type (I vs II) and acid-base status.
- Before high-flow nasal oxygen, NIV, or intubation — baseline PaO₂ and PaCO₂ anchor subsequent decisions.
- Suspected CO or MetHb intoxication — the pulse oximeter is actively misleading in these states.
- Severe anaemia, post-transfusion evaluation, high-altitude evaluation — where dissociation-curve shifts or reduced oxygen-carrying capacity invalidate simple saturation-based decisions.
- Any clinical setting where the oximeter reading does not match the clinical picture. A patient who appears cyanotic and distressed with SpO₂ 99% has a reading that is wrong until proven otherwise.
ABG access in India is heterogeneous. Tertiary hospitals, tier-1 private centres, and most tier-2 district hospitals have blood-gas analysers. Smaller centres, rural primary care, and home-care contexts do not. The ethically defensible practice in the ABG-unavailable scenario is to qualify LTOT on SpO₂ with explicit caveats — multiple readings, multiple fingers, perfusion check, device-quality awareness — and refer for ABG confirmation within 90 days when that is feasible.
Putting it together: three vignettes
Vignette 1. A 68-year-old stable GOLD 3 COPD patient, ex-smoker, no cor pulmonale. Home SpO₂ 91% at rest on a ₹500 oximeter. Clinic SpO₂ on a calibrated device reads 89%. ABG: PaO₂ 57 mmHg, SaO₂ 89%, PaCO₂ 44 mmHg. The patient does not meet the PaO₂ ≤55 mmHg threshold. If there were cor pulmonale or polycythaemia, the PaO₂ 55–60 mmHg category would apply. There is not. LTOT is not indicated. Follow-up in 6–12 months.
Vignette 2. A 72-year-old post-COVID patient, 8 months out from a severe illness, exertional breathlessness. Home SpO₂ 96% at rest. On a 6-minute walk test, SpO₂ falls to 85%. This is exertional desaturation. The patient may benefit from ambulatory oxygen even though resting SpO₂ does not meet LTOT criteria. Prescription is for ambulatory use only — a portable concentrator or an ambulatory cylinder sized to the walking-dose requirement.
Vignette 3. A 55-year-old with chronic respiratory failure, heavy biomass-smoke exposure, SpO₂ at rest 94%. The reading does not match the clinical picture — the patient is breathless and plethoric. ABG: PaO₂ 52 mmHg, PaCO₂ 62 mmHg, SaO₂ 86%, COHb 8%. The pulse oximeter was reading COHb as O₂Hb. LTOT is indicated. The high PaCO₂ mandates careful titration to SpO₂ 88–92% to avoid further hypercapnic drive suppression.
Clinical takeaway
SpO₂ is the right measurement for home monitoring, clinic trending, and rapid triage. SaO₂ and PaO₂ from ABG are the right measurements for LTOT qualification decisions, for any borderline case, for suspected hypercapnia, and for any scenario where the oximeter cannot be trusted — pigmentation-driven bias, CO or MetHb exposure, poor perfusion, severe anaemia, motion artefact. The dissociation curve explains why a 92% reading is not a 92% problem. The cheap consumer oximeter is useful for what it is useful for and dangerous when asked to anchor a prescription it was not calibrated to anchor.
Consult your physician before changing oxygen therapy on the basis of any home oximeter reading.